The need

Document Sample
The need Powered By Docstoc
					 Manna – Case Study

Submitted by:
Ehud Gordin
Tzvi Steingart
Sharon Zuhovitzky
Michael Fink
Oren Steinberg
Ariel Zuhovitzky

 Internet Entrepreneurship Course
               January 18, 2001

Executive summery –––––––––––––––––––––––––                         page 2
Manna - how it all began ––––––––––––––––––––- page 3
Manna – major landmarks –––––––––––––––––––- page 4
Online personalization – the need –––––––––––––– page 5
Current personalization solutions –––––––––––––– page 6
FrontMind - Manna’s solution to personalization                     page 9
The technology enabling the FrontMind solution                      page 33
Market analysis ––––––––––––––––––––––––––––– page 15
Marketing & Sales ––––––––––––––––––––––––––- page 22
Financial analysis –––––––––––––––––––––––––––- page 24
Personalization and Privacy ––––––––––––––––––– page 26
The Manna team –––––––––––––––––––––––––––– page 22
Manna – the future and beyond –––––––––––––––- page 29
Bibliography –––––––––––––––––––––––––––––––- page 32
Appendixes ––––––––––––––––––––––––––––––––- page 33
 Appendix 1 - intro to statistical inference using Bayesian networks 33
 Appendix 2 - ADI (Advanced Distributed Intelligence)               34
 Appendix 3 - Manna Major Competitors                               39
 Appendix 4 - E-Shoppers Choose Personalization Over Privacy        42
 Appendix 5 - Harcourt, Inc. Customer Privacy Policy                44
 Appendix 6 - Privacy Policy                         46
 Appendix 7 - Privacy Statement for                 48
 Appendix (U.S) 8 - DoubleClick Sued for Online Privacy Invasion 49
 Appendix 9 - The European Directive on data protection             52
 Appendix 10 - Manna – management                                   54

Executive summery
The following paper is a comprehensive analysis on Manna Inc. Our choice of Manna
Inc. was by no means accidental. Manna Inc. is a promising company, which is based
on an exceptional technology and development team and run by a highly professional
Manna Inc develops a unique online personalization solution called FrontMind.
FrontMind, Manna’s flag product, introduces a blend of personalization approaches
enabling e-marketers to deliver value to site visitors. In this paper we will try to
illustrate this exceptional start-up company from different aspects. Our goal will be to
highlight the vision of Manna’s founders and to see how this vision is becoming a

Manna’s founders have realized that existing solutions in the personalization market
did not properly address some of the market’s most fundamental needs. In doing so
the business justification for Manna was set. Manna’s exceptional R&D team
developed Manna’s FrontMind solution, a personalization software product, which
combined the ability to use business rule definitions with learning and inference
capabilities. Important aspects of Manna’s solution are automation, speed, control,
single-point-of-integration and lowest total cost of ownership.

This essay will be composed from three major sections:

The first section will contain a short depiction of the way in which Manna was
founded. This section will include a time line illustrating the key events in the
company’s life.

The second section will include a comprehensive analysis on the technological side
of the personalization market. We will try to review the common approaches in the
personalization market. This section will emphasize on presenting Manna’s unique
product, FrontMind and its technological platform.

The third section will deal with Manna’s business modal. In this section we intend to
analyze the different aspects of the business environment in the personalization
market. Our aim will be to examine Manna’s position in this market.

Our company analysis will conclude in our opinion and perspective on the future of
Manna Inc. By that we will try to integrate all the information we have gathered into
one comprehensive conclusion on the prospects and possibilities in Manna’s future.

Manna and online-personalization - how it all began
Three entrepreneurs founded Manna Inc. in 1997: Tal Bar-Noah, Gad Barnea and
Ze’ev Rozov. Prior to founding Manna, the three worked at SEA Multimedia, an
Israeli publisher of sport content on CD-ROMs and through the Internet.
Tal Bar-Noah had previously founded SEA Multimedia (AIM) and served at the time
as the company’s chairman and CEO. Before establishing SEA, he helped found and
manage CDI (TASE), a major Israeli manufacturer of CDs1.

Ze’ev Rozov attended the University of Tel Aviv to study modern European history.
But in 1993 he left school to start SEA Multimedia, the first company to develop CD-
ROMS exclusively for the sports market. Ze’ev and Tal took SEA public in the AIM
stock exchange in London. Ze’ev acted as Manna’s CEO until replaced by Dan Ross
when moving Manna’s HQ to Boston. By the time Manna grew and established its
full product line, Ze’ev Rozov felt that his entrepreneur skills need new stimulation
and he left Manna in order to found Idealive, a company dedicated to bringing artists,
their fans and investors together to help artists fund and successfully complete their
creative projects2.

Gad Barnea was the technical expert of the three and has been building and designing
high-end distributed systems for more than eight years. He is also a composer of
contemporary classical music and has studied philosophy at the Sorbonne. Gad held
the CTO position in Manna until he left the company at 1999 due to difficulties
working with the new CEO Dan Ross. Gad continued to work on a new startup
initiative still in the seed stage called Zebrazone3.
It could easily be seen that each of the three founders has a wide range of interests and
all enjoy the thrill of facing new challenges, clearly three classical “serial founder”

After the consolidation stage of the idea the three approached several venture capital
funds, including Gemini. At first the group didn’t manage to persuade Gemini to enter
as a major investor, but a while later after the group grew to approximately ten
people, and Gemini stumbled across a successful presentation on Manna in a trade
show. The decision was made and Gemini invested the initial 2 million dollars the
founders needed to translate their vision into reality. Gemini’s decision was mainly
base on the founder personality and background combined with the promising
technological superiority of Manna’s “slide-ware”. Gemini also lured its international
partner Advent Corp to invest an additional 1.4 million dollars in the young company.
So the stage was ready with a great idea and technology, a strong development group
and the sufficient funds to actuate Manna’s potential.


From establishment to the present day – major landmarks
     1997 – Manna Incorporated founded at Ramaat-Gan Israel
     Beta site - Streamline Boston (1998)
     April 12, 1999 -Manna Inc. announced that Dan Ross will take the helm as
      chief executive officer.
     December 13, 1999 - Manna Inc. announced the launch of FrontMind 1.1
     January 24, 2000 - Manna Inc., announced its partnership with, and
     February 7, 2000 - Manna Inc. announced an additional $14.1 million in
      second round capital funding
     April 10, 2000 - Manna Inc., announced the release of its latest FrontMind™
      product, FrontMind 2.0.
     August 28, 2000 - Adobe Systems - Incorporated (NASDAQ: ADBE)
      announced an alliance to integrate Manna's FrontMind with Adobe GoLive 5.0
     October 10, 2000 - Manna Inc., announced the shipping of the newest version
      of its high-end personalization solution, FrontMind 3.0.
     OCTOBER 23, 2000 - Manna Inc. announced the appointment of Veronica
      O'Shea as vice president of sales. O'Shea will be responsible for heading up
      the worldwide sales force, creating new sales channels and increasing Manna's
      revenue stream.
     November 13, 2000 - Manna Inc., and I-Impact, Inc. will integrate and jointly
      market the combined software solution to companies looking to leverage
      information generated by their enterprises.
     November 13, 2000 - Immersant, a leading Web consultant and developer for
      complex e-businesses today announced a strategic alliance with Manna Inc
     November 14, 2000 winners of the 2001 Appian Laurels for personalization
      software products

Online personalization – the need
Unlike a local store’s cliental, an extremely heterogeneous crowd usually visits a web
e-commerce site. A young cheerleader from Texas and a Japanese businessman might
shop an e-commerce site simultaneously. This is what turns internet-site-
personalization into an essential tool for e-marketers. Since e-commerce sites often
market a large number of diverse products, presenting all site visitors with the same
site simply implies that only a small fraction of them actually see the optimal offers
suiting their interests and shopping habits (low convergence rates and high level of
shopping cart abandonment’s). This shortcoming translates directly to a critical
reduction in the site’s revenues. The more products a site markets, the more likely the
customers need help in selecting the right product for them.
Even when the number of products is small different clients need to be approached by
different marketing proposals. Although many clients might be interested in the same
product some might be very large clients while others might be very small, some
might be professionals and some might be first time armatures. These considerations
and the enormous competition over profitable clients in the web, entails that in order
to maximize site’s revenues a personalization solution is a necessity.
After defining the need for a personalization solution, we would like to take a look at
the required characteristics of an optimal personalization solution.

The first set of requirements we will define is common to any large-scale e-commerce
solution. The solution must be able to grow with the web site, while maintaining a
24/7/365 availability policy. It should be easy to use and easy to integrate into an
existing site structure. Finally it is expected to pass the return-on-investment test
justifying the time and resources invested by the organization.

The second set of considerations is more specific to evaluating the need for a
personalization solution. It is required that all client-data gathered by the organization
from a wide range of sources, could be used to leverage the personalization efforts.
The heart of a personalization solution is in its inference-capabilities, so powerful and
creative inference is necessary to make the most out of the collected data. The
inference and recommendation process should be fast enough to provide an online
response to newly gathered data. It is crucial that the personalization solution should
not compromise the basic privacy of the site’s clients, driving potential clients to shop
elsewhere. In addition it is expected that long and tiresome questioners usually
causing up to 80% of the site’s visitors to search for friendlier shopping environments,
shouldn’t be a part of the proposed solution.

One last requirement concerns the need for feedback to marketers on the performance
of their personalized campaigns; in order to swiftly make any required changes. Pre-
campaign simulation capabilities aimed at predicting the results of a future campaign
is as another aspect of this important feedback.

The points detailed here summarize the product need for a personalization
solution, the market need for a personalization solution will be analyzed in the
market analysis section later in this draft.

Current personalization solutions

Current solutions – general
Almost every Internet visitor will encounter personalization in one form or another: at
portals, retail sites, online banks, even from news and information providers. The
following section presents a few of the most popular personalization techniques,
employed by thousands of Web sites — including some of the most visited ones —
Customers choose what they want to see With this method, first-time visitors are
asked to complete a questionnaire, specifying information such as age, address,
gender, occupation, marital status, and other demographic data, as well as some
general preferences and areas of interest. The site then uses this information to
customize content for each user and to target him or her with relevant content.
Visitors often can further personalize the site by choosing information (e.g., airfares,
news, stock quotes, weather) –they wish to see.

Setting business rules In this approach, business logic is embedded in conditional
“if/then” statements to help control and monitor sales for selected items. For example,
an online retail-clothing store that has overstocked white t-shirts may employ a pre-set
business rule of “If a customer purchases khakis, then offer them a white t-shirt at a
50% discount.” Similarly, an online bank may use a business rule to detect when a
customer needs to order new checks: When the customer is at check number 985, he
or she will receive a message to reorder. These forms of personalization attempt to
make use of traditional business knowledge to generate rules of relevance to online

Community-based personalization Community-based personalization matches
customers with others who have similar tastes and preferences to make
recommendations, as in the following scenario: Julia enters an online bookstore,
intent on buying the latest John Grisham novel. Just as Julia is about to complete the
purchase, she is shown a list of four other recommended action/thriller novels she
may want to add to her shopping cart, based on buying preferences of her
demographic peers. This technique is ideal for selling low unit-price products like
books, CDs or videos.

Current personalization technology
The back-end technology that makes online personalization possible falls into two
main categories: rules-based engines and collaborative filtering engines. Both can
detect user preferences and build customer profiles through manual surveys and
questionnaires, and/or by monitoring a customer’s online behavior. However, the
similarities end there.

Rule-based engines
Rule-based engines use traditional merchandising rules and business logic embedded
in conditional “if/then” statements to create content display. Through a graphical user
interface, pre-set business rules are used as the framework for customer interaction.
As illustrated in the online retail clothing store example above, rule-based engines are

able to directly link organizational strategies to customer interaction. Most rule-based
solutions are limited in this process, however, as they do not look for data that is
dynamic in nature or can be leveraged for future use. Static rules are not capable of
learning over time or making inferences based on partial or changing information.
Thus, an exclusively rules-based solution will not be able to use dynamic data (e.g.,
clickstream, purchase history or ‘empty shopping carts’) and cannot look for or find
trends that will change over time. And so, when e-marketers notice and want to
respond to changes in the marketplace, in a rule-only environment database changes
will need to be made to continually leverage new information. Also, with a static rule-
based solution, the e-marketer must deploy hundreds of business rules to capture all
possible combinations of offers – not only is this difficult and time consuming for the
e-marketer to manage, the technology can not scale to handle this many variables,
causing the technology to break down. As a result, with a static rule-based solution,
marketing and site analysis reports must be constantly reviewed and analyzed. The
resulting analysis must then be transformed into rules that are manually fed back into
the system and must then be carefully monitored for efficiency. This requires a
massive manpower investment throughout the life of such a solution. Worse,
however, is the simple truth that a static rules system ipso facto cannot offer true
individualized personalization.

Collaborative filtering engines
Collaborative filtering engines — the technology powering community-based
personalization — record actual user preferences to algorithmically make customer
recommendations predicted on what like-minded customers would want based on
their buying patterns. Site visitors are segmented into groups based on their
preferences. Through the use of preference databases the engine makes its predictions.
However, there are definite drawbacks associated with this technique. Collaborative
filtering vendors cannot leverage any personal profile and demographic data other
than purchase history. Collaborative filtering implementations are most powerful
when the customer actually puts in the time and effort of an explicit ‘rating process’
through a Web site, to in turn receive meaningful results. In fact, the customer must
provide enough data points for the collaborative filtering technology to match his or
her data with that of enough similar customers’ to provide a useful comparison and
recommendation, not an efficient process at all.

Because of the need to compare behavior of hundreds of individuals, many
collaborative filtering technologies lose their abilities as large data volumes of items
to be recommended are introduced, making its use unfeasible for most large e-
commerce sites. Further, most collaborative filtering technologies are not designed to
embrace or execute static business rules to implement explicit business.
A combination of running the two solutions simultaneously is hardly a practical
answer. Apart from the expense and resources required, there is additional effort
involved in maintaining the collaborative filtering system, as well as a separate static
rule-based system, yet somehow keeping the two completely synchronized. No one
disputes that collaborative filtering and rule-based engines are the main technologies
fueling the most trafficked Web sites today. Yet it is equally obvious that both have
built-in limitations and drawbacks, and that neither offer maximum leverage of data.
And the most important question of all still remains: How personalized are today’s
personalization efforts?

Shortcomings of current technologies
Despite their prominence in the media, most of today’s personalization efforts are
fairly simplistic, and take into account only limited customer information. For the vast
majority of current web personalization tools, it is a customer’s next visit — not the
current one — that will result in a more customized offer. By that time, with most
visitors spending less than 30 seconds on a site, it is often too late. Naturally, the more
information known about a customer, the better the chances are of presenting him or
her with attractive offerings. But current personalization technologies require
significant amounts of upfront information from a customer, running the risk of
aggravating visitors with lengthy questionnaires and surveys. Today’s efforts may be
sufficient for selling “simple” products and services like books or CDs, instances
where it is difficult to attach a real potential value to a given customer. But for more
complex, feature-rich products, such as computer hardware or credit cards, other
methods are needed. Furthermore, as noted above, using either of these methods
yields very little understanding of what worked and what did not work. The “lessons”
that could and should be learned and applied in an iterative basis are not retrievable,
leading to continual — and expensive — “trial and error” attempts.
Ironically, most of today’s personalization tools lack the two crucial factors in
successful Internet relationship management that they are supposed to support: the
ability to provide real-time control and learning relationships.

FrontMind - Manna’s solution to online personalization
Manna Inc. developed a unique online personalization solution, FrontMind; to address
the market needs presented above. FrontMind introduced a blend of personalization
approaches enabling e-marketers to deliver value to site visitors. This results in an
increase the likelihood of visitors becoming customers, and ultimately, purchasers.

How Manna meets the online personalization challenge – general
Manna Inc. has synthesized attributes of each online personalization technology and
developed a unique solution, FrontMind. Manna’s FrontMind solution combines both
learning and inference capabilities with business rule definition. Important aspects are
automation, speed, control, single point of integration and lowest total cost of
ownership. It is a personalization solution combining three approaches. FrontMind’s
self-updating customer behavior models help e-marketers understand and predict the
needs of individual customers. Its intuitive, easy-to-use Business Command Center
(BCC) gives e-marketers instant control over their marketing initiatives, without
requiring support from IT. And FrontMind’s comprehensive reporting and analysis
capability delivers feedback on both specific marketing initiatives and overall return
on investment. This singular personalization approach enables e-marketers to know
their customers, deliver precisely targeted offers, and learn and improve marketing
initiatives. FrontMind utilizes the following methodology to deliver true online

FrontMind – overview
Figure 1 depicts FrontMind’s step-by-step delivery of these abilities:

Create/Update: Business rules are created, launched and updated immediately based
on automatic feedback from the customer behavior models. FrontMind’s BCC permits
the creation of two types of business rules: static and self-updating. Static rules use
straightforward “If/Then” statements (e.g., “If a customer enters the fruit aisle, then
offer him bananas at half price”). To create true personalization, self-updating rules
use Learning Engine to determine the best product, promotion or content to be
displayed to the appropriate customer.
Test: Once a business rule is created, FrontMind can pre-test any marketing initiative,
before launching it. An expected rate of acceptance is generated, along with what
attributes contributed to the expected rate of response. This iterative process
contributes greatly to reducing the costs and time of failed marketing programs.

Interact: By creating a holistic profile of the customer, FrontMind generates a
recommendation at a given moment. By forwarding each new click to the self-
updating customer behavior models, it delivers varying recommendations to reflect
customer behavior.
Learn: FrontMind’s Learning & Inference Engine models customer behavior and
refines these models as it learns from interactions. FrontMind possesses the ability to
build self-updating customer behavior models by leveraging all existing data (e.g.,
demographic, psychographic, past-purchase history). These models, which are used
by the Learning & Inference Engine, coupled with “behind-the-scenes” recording of
real-time customer activity, give FrontMind the ability to make personalized
recommendations to each customer, from the very first visit.
Analyze/Report: FrontMind’s in-depth reporting and analysis of all published
marketing initiatives provides insight into site’s revenue generation and rate of
acceptance, so marketers can make intelligent decisions.

Process flow of the FrontMind solution
The FrontMind Solution integrates with an enterprise’s e-commerce solution, tying
into its Web Server for real-time customer click-stream behavior, as well as into
existing database information such as customer demographics and buying behavior.
The process flow operates as follows:

The Marketer:
1. Creates and updates business rules using the Business Command Center (BCC)
2. Pre-tests rules for acceptable response rate (Simulation)
3. Publishes rules to the FrontMind Server

4. The customer accesses a Web Site through the Web Server.
5. FrontMind Client sends events to and receives results from the FrontMind Server.
The FrontMind Client can be deployed using ActiveX, Servlets or Sockets, according
to the system tools and platform.
6. The Core, FrontMind’s “operating system”, is “listening” for occurrences of
selected events resulting from the customer’s click-stream activity.
7. Through the Rule Evaluator, a component of the Core, the event is sent to all active
rules to determine which rules are relevant for the given customer.
8. The Rule Evaluator calls on the Learning & Inference Engine as needed, to
determine how to apply a rule for the given customer. In addition to click-stream data,
the Learning & Inference Engine can utilize Operational and Historical Databases for
specific customer/product data.
9. Following rule evaluation, the Core sends the recommendation (personalized
products, services or content) to the FrontMind Client.
10. The FrontMind Client communicates with the Web Server and delivers the
recommendation to the Customer.
11. In the background, the Core executes basic functions such as load balancing, rule
creation and messaging. It reads and updates data including rules and configuration
information, which are stored as XML files in the FrontMind Database.

FrontMind solution components
The FrontMind Solution is made up of five main components:
       1. Business Command Center
       2. FrontMind Server
       3. FrontMind Client
       4. FrontMind Database
       5. Administration Tools
The Business Command Center (BCC) gives marketers the ability to create, pre-test,
update, and evaluate the impact of their intended marketing initiatives using business
rules with browsers and Wizards, in Windows environment. The BCC also provides
in-depth reports and analyses offering marketers the information necessary for making
effective decisions in real-time mode.
Both the Core and its main component, the Rule Evaluator, as well as the Learning &
Inference Engine are key components that run on the FrontMind Server, the
infrastructure that provides operating system-level and other services to FrontMind
applications. The FrontMind Server is responsible for distributing events to the
appropriate Service and to the FrontMind Database to be logged for future reporting

Plug-ins are executable components of the Core created for a specific business
function. They can be created, invoked, updated, and removed in real time without
interrupting the system operation.

Load Balancing
Load balancing is the term given to FrontMind’s ability to control the distribution of
Sessions over virtual machines. Load balancing determines optimal deployment for
any given hardware configuration.

Query Request Contexts (QRC) is the FrontMind internal messaging agent. Messages
are sent to system components external to the FrontMind Server in XML format, and
transmitted internally it in XML DOM format.

The system integrator, in conjunction with the e-commerce system administrator,
establishes events. Each event is recorded in a configuration table, which is stored in
the FrontMind Database. Any visitor action on the Web Site can trigger an event (e.g.,
a visitor registering his name).

Rule Evaluator
A rule is a specified business scenario that can occur online, including an interactive
event that is triggered by a scenario. The Rule Evaluator, a component of the Core, is
the mechanism that evaluates information generated by an external event (e.g., a
customer purchase) according to a business rule. The Rule Evaluator receives the

rules established in the BCC, and evaluates each using stored data or the FrontMind
customer behavior model combined with other available data. The result is routed to
FrontMind Client in real time, using the Front Mind Server’s messaging facility. One
business rule can serve many Sessions and can be in one of two states: Published or
Unpublished. The state of the rule is controlled through the BCC.

FrontMind Client uses protocols to mediate between all leading Web applications and
FrontMind. It can support a single Web application or adapt itself to work with Web
applications distributed over a number of hosts, as well as non-sticky sessions.

The FrontMind Database stores the configuration tables required for system setup,
Business Objects created in the BOD, and business rules created in the BCC. Log files
created by FrontMind for future BCC reporting purposes are also stored in the
FrontMind Database. The data is stored in XML .

The FrontMind Solution includes two administration tools for implementing and
maintaining a personalized Web Site, the Business Object Developer (BOD) and
Console Manager.

Business Object Developer
The Business Object Developer (BOD) is an automated tool for creating business
objects, later to be used as building blocks by the BCC to create rule components. The
BOD incorporates a browser and “wizards” into its design to further facilitate its use.
Each operation can access a different database. Business Objects are designed for
three of the four stages of the Rule Evaluation process, namely, Situation, Profile and
Within FrontMind’s Core, the Rule Evaluator determines where to go to evaluate
Business Objects. This may require invoking click-stream data stored within the
FrontMind Database to see if the customer has just completed a specific action on the
site, or using stored demographic or past buying data (stored in company databases or
the FrontMind Learning & Inference Engine) to determine the best Profile or Result
to target.

Console Manager
The Console Manager provides the tools used to configure the FrontMind Solution,
track system utilization of resources and set security and other system controls. With
the Console UI, the system administrator can configure hardware system components,
FrontMind Server parameters and system security. In addition, the system
administrator is able to the view system status and ensures proper functioning of the

The technology enabling the FrontMind solution

Since Manna’s advantage in the very crowded market of personalization solutions is
in its powerful distributed-intelligent-inference technology, we will dedicate the
following part of this case-study to try and understand the uniqueness of the
foundations this technological supremacy lays upon. In a crowded market dominated
by several large players, with rather questionable technological solutions Manna is
justifiably concerned about revealing how these foundations were actually
implemented in the creation of FrontMind. As a consequence this chapter will refrain
to giving only a high-level overview of the various agent technologies and Bayesian
network methods used in Manna’s FrontMind product.

Since we will present only a high-level overview of the technologies incorporated in
FrontMind we will include the more “techish” parts as appendixes (appendixes1 and 2
on pages 31 and 34). We urge all readers to read these appendixes and try to pick up
the general orientation of the technological aspects of FrontMind (even if
understanding every technical detail or three-letter abbreviation is not possible).

As described earlier in the previous chapter FrontMind is a high-end, mission-critical
server for Internet Relationship Management that permits commercial sites to conduct
online learning relationships with their customers through intelligent business rules.
We will open this chapter with a short introduction to statistical inference using
Bayesian networks. This is the core of FrontMind’s inference capabilities. In the
second section of this chapter we will describe the technological intelligent agent
approach of the FrontMind solution called ADI (Advanced Distributed Intelligence).
It is important to note that Manna’s technology is extremely versatile. This versatility
is due to the combination of strong inference capabilities using an extremely flexible
and reliable platform. This combination could easily be adopted to solve a wide range
of needs emerging in today’s network environments; therefore positioning Manna
with the crucial flexibility needed in the internet-market.

A short introduction to statistical inference using Bayesian networks
Graphical models are graphs in which nodes represent random variables, and the lack
of arcs represent conditional independence assumptions. Manna uses a directed
graphical model called Bayesian Networks. The method of Bayesian Networks
enables the calculation of probabilities of certain events represented in the random
variables mentioned above. Bayesian Networks can incorporate actual observations of
surfers’ click-stream behavior, background data and demographics in order to predict
the optimal site composition any individual surfer should see. This optimization leads
to the capability of maximizing the web site’s revenues. A detailed analysis on
Manna’s management can be found on appendix no 1 and 2 – pages 31, 34.

Advanced Distributed Intelligence - The technological platform
Manna’s FrontMind personalization solution is applied using a programming
approach call ADI (Advanced Distributed Intelligence). This approach emphasizes the
use of artificial agents that serve applications and agents that serve other agents
forming together FrontMind’s federation of agents. FrontMind Implements agents that
can learn independently, for the purpose of improving themselves, and digesting new

data into a form from which they can infer when queried by a human or an

There is a high degree of communication among agent communities in FrontMind.
The whole system is designed such that it creates a federation of agents in which all
the different agent communities speak the same language (based on DOM and an
XML dialect), while specializing in very different tasks. The cooperative metaphor
allows FrontMind to achieve exceptional reliability and stability, since, although the
interactions and tasks are certainly complex, the system itself is strikingly simple.

There’s a full pseudo-biological circle in the system in which every agent community
relies on the other and at the same time serves it. The load-balancing agents rely (for
certain tasks) on the learning and inference agents. At the same time, the learning and
inference agents—dealing with large amounts of data, in real time, need to rely on the
load-balancing agents to handle the amount of load a live e-commerce system expects
to take. The learning and inference agents communicate more “intellectual” data
among them. Obviously, this form of cooperation uses the messaging agents for its
intergraded communication. Because of its inherent modularity and flexible
componentization this form of design leads to systems that are easier to maintain and
more ready to evolve.

Market analysis

On-line personalization – the market need

There is no longer any doubt that e-commerce is not just the wave of the future for
business, but increasingly of the present as well. Forrester Research estimates this
intensely competitive sector will grow to $3 trillion in sales by 2003; with B2C sales
at $108 billion and B2B at $1.92 trillion. With the shift to the Internet for a fast,
convenient shopping experience becoming more pronounced daily, e-marketers must
do more than simply satisfy visitors’ buying needs with basic online transactions. For
e-businesses to become truly profitable and sustain a competitive edge in this
multibillion-dollar business, they need to give visitors a reason to return to their sites
by creating a productive and efficient online experience. In short, to turn the
occasional visitor into a loyal customer.
Online personalization, the concept of presenting individualized Web content for site
visitors, is now recognized as a key component in fostering these customer
relationships, building online brands, and improving sales and profit margins; in
essence, building a successful e-commerce Web site.
The benefits of personalization are quite obvious: For vendors they include more
sales, larger sales, more frequently returning customers. Companies usually have
three important objectives for using web personalization:
1. Retention -- The more time customers spend on the site, the more they will interact
or transact on your web site. Web personalization has improved retention because
customers invest themselves in the system. And as the system recommends relevant
alternatives, users select more and more web pages and buy more products.
2. Repeat rates -- For customers, the benefits of Web personalization include easier
access to products they care about, and a better overall experience of their interaction
with the company — in fact, this is why they buy more and come back more often.
3. Purchase rate -- In general, the average purchase rate on the web is about 2
percent. But a recent report showed that web personalization pulls in purchase rates
somewhere in the neighborhood of 10 to12 percent."
And while there is consensus that most e-businesses feel their personalization efforts
are yielding results, only 16% of e-marketers surveyed in a recent Forrester report
claimed they actually measure the direct impact of their efforts. Most e-marketers will
agree that with many of today’s personalization technologies, it is difficult to measure
results, let alone translate them into verifiable data.
Despite the prominence accorded it by the media, most personalization products
offered today are still simplistic, and do not provide the real-time control or learning
relationships e-marketers need to fully manage online marketing initiatives. Most
“state-of-the-art” technologies cannot scale for complex recommendations or high
volumes of customers, and detailed pre-purchase information from the visitor is often
required. Worse still, most offer little understanding of what initiatives worked, what
did not, and why.

Potential clients
Manna Inc is constantly looking to expand and diversify the number of companies
using the FrontMind personalization solution. The increase in customers beside the
benefits upon the revenue side, help Manna to establish itself as a technology leader
in the personalization market. From talks we held with Ms Tali Aben, senior partner
at Gemni Capital Management Fund, it seems that many potential clients are
companies that were using other personalization solutions. Many times these solutions
are part of a comprehensive e-commerce solution provided by consulting firms. These
suites do not provide a personalization solution which matches the one provided by
Manna. In the highly competitive market of today it is absolutely crucial to use to
most advanced and efficient solution. Many of Mann’s newest customers are
companies that were disappointed from the level and performance they received from
other personalization software in the market. With each new customer Manna is
positioning itself as one of the best personalizing solution in the market an by that the
company is overcoming it’s basic disadvantage as technology based company and not
a marketing driven one. We see Manna management decision to become what is
known as the “best of breed” strategy very helpful in accruing new clients and by that
expanding Manna’s market share.

Personalization market size

In the past three years the e-commerce market has emerged as an important channel
for marketing, sales and support. Online personalization is now recognized worldwide
as a key component in fostering customer relationship, building online brands, and
improving sales and profit margins for e-commerce. Both “” and “clicks and
mortar” companies alike are investing ever-increasing amounts in these efforts.
Online retailers are investing heavily in personalization efforts to increase sales and
make their sites easier to use.
Appian Corporation predicts the online personalization industry, including custom
development and independent consulting, will reach $1.3 billion this year, and $5.3
billion by 2003.

The predictions made by the Appian Corporation were made prior to the “Dot.Com”
crash in U.S since April 2000. Every day more and more e-commerce sites are
“closing shop”. As these “Dot.Com” are a big part of the personalization market the
future is gloomier then was predicted at the beginning of 2000.

It is our opinion that the current emphasis on immediate revenues business
model might actually act as a boosting effect on the personalization market.

Manna has many competitors most of whom are much bigger and have more market
experience than Manna. Manna’s weak spot, in comparison to her competitors, is the
lack of complementary products to the personalization solution that most of the
competitors provide. These complementary products include database management
products, billing systems and integrated solutions for phone center personalization. A
manager of an E-commerce site that is seeking a comprehensive solution including
database management and personalization capabilities might wish to purchase a
complete package from one company. This solution will be cheaper and enjoy a
superior integration process.
Who’s big and who’s small in the personalization market:
                                      Table A-1
      Comparative analysis of major players in the personalization market

     Company                 Sales                Net Inc              Employees
Manna                      2.5-5M$
Net Perceptions              15M$                  12M$                    328
Broadvision                 115M$                  19M$                    652
Blue Martini                 11M$                  10M$                    235
ATG                          61M$                   5M$                    469
Vignette                     89M$                 (42)M$                   769
E.piphany                    19M$                 (22)M$                   279
Personif                    1.2M$                  (9)M$                    82
net.Genesis                   6M$                 (16)M$                   160

Personalization – a world of alliances
A word on the alliances in the personalization market…
In the past year the personalization market had witnessed a string of acquisitions and
strategic partnerships among personalization companies. For example, Vignette, a
leading provider of e-business application servers, recently announced it is acquiring
DataSage, which provides analytical tools to support large-scale, e-marketing efforts.
BroadVision, which competes with Vignette, has formed partnerships with Broadbase
and E.piphany, both of which provide analytical tools for e-business. Accrue acquired
NeoVista, NetPerceptions purchased KD1 and Macromedia acquired Andromedia. In
addition, large vendors such as Microsoft, IBM, MicroStrategy and NCR are bringing
together far-flung operational and analytical product lines to deliver sophisticated e-
marketing and personalization capabilities.
Since numerous companies and alliances flood the personalization market we chose to
briefly present only Manna’s main competitors in the Appendix 3 on page 39.

Manna in the market
"Manna is poised to be the leader in the personalization software industry”, states Dan
Ross, CEO of Manna Inc. "Manna is well-positioned to capture a large portion of the
growing online personalization market," said Patrick Kenealy, general partner at IDG
Ventures. "Manna's solution represents the state-of-the-art in real-time personalization
technology, and the Manna executive team is a great combination of senior marketing
and technical experts. The team's big accomplishment has been to make such
powerful software usable by marketing people, not technical people, and that will help
Web site owners, visitors, and shoppers."
"Manna possesses all of the key elements for market success - talented management, a
strong product concept and an in-depth understanding of the unmet needs in today's
personalization marketplace."

SWOT- analysis
 Technology – Manna has a extraordinary personalization approach which enables e-
marketers to know their customers, deliver precisely targeted offers, learn and
improve marketing initiatives. The basic technological advantage Manna has over her
competitors is a clear and valuable asset. This asset is based on the exceptional
research and development team in Israel.
 Management - Manna had the ability to attract a team of top notch professionals in
management and in development areas. The Manna Team is by all accounts a great
asset. The experience and expertise of CEO Dan Ross and VP R&D Moni Manor (see
appendix 10 on page 54) are an assurance that the company will know how to
confront new challenges and make the most out of new opportunities.
 Pioneers – Manna is enjoying a pioneering status in many of the technological
solutions to the personalization need. This status if can be maintained is a major
advantage point and a huge breakthrough opportunity.

 Market of giants – as can be seen Manna is facing numerous companies and
alliances in the personalization market. Manna’s main competitors are giant in the
market that posses not only large amounts of capital but also clutch a large part of the
total market. Mann’s necessity to compete with some of the biggest and strongest
software companies in the world is clearly a weak spot.
 Long sales cycle – The nature of Manna’s market creates a long period of time
between first contact with the client to closure.
 Best of breed strategy – Manna is by no doubt concentrating on a market niche.
This strategy is risky and to our opinion exposes Manna to shifts in her core market.
Even though Manna’s best of breed strategy helps focus resources in the company it is
a disadvantage when confronting competition that supplies complete e-commerce

 Mergers and Acquisitions activity - one of Manna’s greatest opportunity lies in the
possibility that Manna will acquired by a bigger player in the market. This prospect is
quite feasible when taking into account Manna’s strong technological foundation and
market proven solution.
 Expanding and mature market – the e-commerce business is evolving in tremendous
speed even in the current times when the Dot.Com environment is losing ground. The
e-commerce is a revolution that is expanding rapidly. Manna has an excellent
opportunity to grow with this expansion of its core business. Another side of the e-
commerce market change is the sophistication and expertise of these sites. The more
sophisticated the site wishes to become the higher the emphasis it will give to

 Funding & Revenues – as in any company liquidly is a major concern. Manna as a
company has not crossed the profitability line yet. Until it does so Manna is
financially non-independent. This situation represents a viable liquidity threat.
 Market shift – one of the biggest threats Manna is facing in the near future is the
change in Dot.Com environment. As more and more companies go belly up the e-
commerce market is becoming more and more concentrated and thus competition is
 Regulations – the rising concern about privacy can bring forward the legislations of
rules and regulation that will limit the ability to gather and use personal information.
The threat is from 3 types of legislations:
Government, market (i.e. consumers avoiding sits with personalization tools), and self
regulations (i.e. regulation created by companies active in personalization market in
order to crate public trust in privacy issues)

Marketing & Sales

Manna – Marketing strategy
Manna has a limited marketing budget due to the early stage the company is at and to
the need to invest heavily on R&D. The marketing budget has to be used efficiently.
Manna incorporates three marketing channels that fit the budgetary limitations while
providing new clients and vast exposure in the personalization market. Manna
employs three channels of marketing:

Manna’s seminars - Manna’s Executive Breakfast Series are intended to help
prospective clients understand how personalization can be used to answer real on-line
business challenges. In these seminars potential customers can hear an in-depth
presentation led by Mitchell Kramer, Senior Analyst/Consultant from The Patricia
Seybold Group, the people who brought followed by Manna’s Chief
Executive Officer, Dan Ross.

Trade shows - Another channel of marketing is through industry trade shows. The big
trade shows attract high level marketers and decision-makers who are looking for
more effective and innovative ways to use personalization technologies to improve
their business.
The trade shows are a great way to put Manna’s identity in the spotlight at the
Personalization market.
Here are some of the industry trade shows manna will take part in the year 2001:

                      BEA eWorld 2001 The Sixth Annual BEA Users Conference.
      Gartner CRM Event March 19, Chicago
      E Personalization March 21, San Francisco

                         Personalization Summit April 1, New York

Professional Magazines – Manna is focusing on achieving maximum exposure within
the professional circuits. One of the most efficient ways to reach this goal is by
advertising in professional magazines that are read by IT professionals and CTO’s in
the personalization and e-commerce markets.

Selected customers
On January 2000 Manna announced its partnership with two fast-growing online
retailers - (OTC:BB:SLET), an e-commerce business that offers the
outlet shopping experience online, and, a new Web site geared
toward outdoor enthusiasts. Both and will be using
Manna's FrontMind personalization solution to help plan and implement online
marketing strategies and provide site visitors and customers with a helpful, more
enjoyable online experience to keep them coming back.
                    was established in 1999 and is based in
                             Berkeley, California. is a portal site
                             designed specifically for outdoor enthusiasts. Launched
                             in April 2000, the site offers information about outdoor
sports, including helpful suggestions and tips on recommended destinations. Visitors
will be able to purchase a wide range of outdoor gear and apparel on the site.
"We are pleased to be working with Manna to bring a new level of customer service
to the online arena," said Adam Meron, CEO of "With Manna's
FrontMind solution, we can ensure that our visitors receive a highly personalized,
instantly responsive experience each time they visit4."

                  Saleoutlet is a New York based online outlet center (established in
                  1998); a dynamic site where you’ll find brand-name merchandise at
                  deeply discounted prices. Saleoutlet has a powerful search engine,
                  which lets the customer specify exactly what he wants, or lets him
                  browse through areas of interest. There are a lot of categories like
clothes, music, books, sporting good, gardening tools and china. Although Saleoutlet
inventory changes all the time but one thing stays the same: every thing is on sale,
every day.
Saleoutlet has a “Personal Shopper” which helps the customer find specific items and
even e-mails him directly when new items his been looking for are available.
Saleoutlet provides customer Service to assist him in his shopping experience. The
customer can even maintain a check out address book, so Saleoutlet can ship his
purchases to all his friends with just a click. has chosen to work with Manna’s FrontMind software in order to
effectively cross sell and up sell related products to customers in real time, as well as
assist customers with gift-giving ideas.
"We are pleased to be working with Manna to create the best possible customer
service for our Web site visitors.'s goal is not only to attract new
customers, but to build a base of loyal customers who are continually drawn to the
most individualized shopping experience possible online5," said Michael Aronowitz,
President and CEO of

    Israel's Business Arena on January 25, 2000
    Israel's Business Arena on January 25, 2000

                           Harcourt is a collaboration of a team of leading publishing
                           companies joins together to provide lifelong learning
solutions for all people, all ages, and all walks of life. Harcourt mission is to develop
products that inspire learning.

Harcourt Web site opens it’s doors to those who wish to browse the materials
available in all areas of expertise within the Harcourt family.
Harcourt gathered educational resources, articles, and ideas for customer to use
throughout his journey of lifelong learning.

Harcourt site is committed to traditional learning and to providing outstanding
materials to our four major markets: Education (K-12); Higher Education; Scientific,
Technical and Medical; and Corporate and Professional Services.

At the same time, Harcourt recognize that new technologies are rapidly transforming
how people learn today. Lifelong learning is moving out of the classroom and
becoming available to all. is the place where Harcourt's years of
educational excellence meets the latest technologies to deliver new ways of learning.

The costumer can choose from any of the following categories to see what Harcourt
and its partners have to offer:

                           Kids, Teens, and Parents
                           College & Grad Students
                           Science & Health Professionals
                           Business & Legal Professionals Educators
                           Personal Enrichment & Continuing Education.

Manna’s Partnerships Program
Manna believes that its success can only be achieved through the success of its
Partners, measured by satisfied mutual clients using the FrontMind solution.

Manna provides expert advice to ensure the successful, seamless integration of
FrontMind, educating its partners on technology. The overall goals are to help
partners successfully augment their core business with Manna’s enabling
technologies, as it extends the knowledge and reach, and increase the scope of
business, for both companies.
Candidates for Manna's Personalization Partner (MPP) Programs

Manna's Programs are tailored to fit the skill sets and capabilities of Partners who
provide services, applications, and complementary technologies across the spectrum
referenced above. That would include:

Systems/Solutions Integrators who design and implement e-Commerce and/or eCRM
solutions, or who recommend technologies for such implementations
Service Providers (CSPs/ASPs) who provide hosted solutions for clients in the e-
Commerce and e-Marketing space
VARs/ISVs interested in embedding Personalization technology into a vertically
oriented bundled application solution
Technology Vendors who provide platforms for e-Commerce solutions and see value
in insuring compatibility and ease of integration with state of the art Personalization

The Programs
    Alliance Partner Program - This Program (available to all) allows Partners
       to align themselves with Manna for the purpose of joint marketing and
       business development. An optional Referral program allows the Partner to
       register sales opportunities and co-engage with Manna Sales teams at
       prospective end users. Through the Referral program, Alliance Partners may
       receive referral fees as a percentage of license revenue, or such fees can be
       used for training credits, marketing co-op, etc.
    Implementation Partner Program - Designed for SIs with highly qualified
       implementation skills and personnel, this Program assures Certified
       Implementation Partners a fully supportive and non-competitive posture from
       Manna's Professional Services team on any registered opportunities.
    Developer Program - This Program is designed for Partners (Service
       Providers, VARs, ISVs) who desire to bundle FrontMind into an integrated
       service offering and/or application solution. Developer Partners are granted
       the right to sublicense FrontMind when sold with its approved value added
       service or application. Such Partners establish the pricing with the end client
       and is obligated to provide client support.
    Technology Alliance Partner Program - This Program is designed for
       Technology vendors with products complementary to FrontMind who are
       interested in insuring compatibility, ease of integration, or enhanced

Financial analysis

Investment rounds

Manna’s first capital raising exercise, at the seed stage, amounted to $500,000,
received from private investors (Angels).

Manna’s first round of major capital funding was held on May 1999. Manna raised
$3.7 million from venture capital funds Advent (international) and Gemini (Israeli), at
a company value of $10 million after money6. The investment served Manna for
recruiting manpower for company management, and for putting the company’s first
product on the market.

Manna, after the two early stages, was owned by the Gemini and Advent funds (30%),
SEA holding company (30%), and employees (20%), with the balance in the hands of
private investors.
On February 2000, Manna announced an additional $14.1 million in second round
capital funding from several noted firms including lead investor IDG Ventures,
Robertson Stephens' Bayview Investors, Advent International, Gemini Capital
Management Fund, Apax Partners and Bluewater Capital Management 7. This infusion
of capital enabled Manna to expand its national sales force and professional service
teams, and to fuel its growing market share in the online personalization industry.

It should be noted that Both Advent International and Gemini Capital invested in
Manna during its first round of financing. "We continue to be impressed with Manna's
FrontMind software and the company's aggressive approach to giving e-marketers
tomorrow's personalization solution, today," states Marcia Hooper, partner at Advent

Form talks we had with Ms Tali Aben, senior partner at Gemni Capital Management
Fund, Manna is considered a valuable and promising company. Ms Aben noted that
Gemni has very close relations with Manna’s management team and that Gemni will
positively look upon investing in Manna’s next round.

    Israel's Business Arena May 18, 1999
    by Israel's Business Arena on February 8, 2000

Revenue model
Pricing – median price for the FrontMind application is marketed to Fortune 500 and
large e-commerce companies is $250,000 for license and installation Expenses or on
an annuity model.
From talks we had with Mr. Moni Manor Manna’s VP R&D he stated that FrontMind
clients report an exceptional improvement in the web sales conversion rates.
Conversion rates is the percent of shoppers in your web site who actually make a
purchase. As can be seen in table 2 personalization has enormous impact on the
conversion rate of a web site. Mr. Manor has stated that Manna’s solution has been
proven and implemented in its customers web sites. Those web sits have shown a
remarkable improvement in the conversion rate.

“In one study that we did we compared 6 months of a retail web application
without our product and then 3 months with it. We did measure a decrease of
50% in shopping cart abandonment, with translated to almost doubling of the
average purchase. In addition we improved the conversion rate (amount of first
time visitor that actually buy) by 70%” Mr. Moni Manor Manna’s VP R&D.

                                     Table 2
                Web Sales Conversion Rates and Personalization 8/998

                           August Conversion Rate         Personalization?
     Amazon                         8.30%                        Yes                  7.90%                        Yes
     QVC                            7.10%                        Yes
     Lands’ End                     6.30%                        Yes
     CDNow                          6.20%                        Yes
     Ticketmaster                   5.70%                        Yes
     BMG Music                      4.80%                        Yes                     4.50%                        No
     J. Crew                        3.40%                        No                       3.50%                        No

    Source: Nielsen//NetRatings, August 1999

Personalization and Privacy
Privacy vs. Personalization9
When a company, not unlike Manna, is active in the personalization market it must
consider different legal issues regarding privacy. The personal information that flows
through the Internet is used by companies like Manna and thus gives them power and
control. With the increase of volume in the e-commerce market, the need to protect
the personal information is growing. The collision between personalization and
privacy is a “hot” legal issue in the Internet era.
One can advocate two different approaches for tackling this dilemma:
        1. Preferring personalization to privacy.
        2. Giving privacy the lead.
This approach states that there are two kinds of information which a marketer can
gather on a consumer: active information which is gathered mainly through
questionnaire; and passive information which is gathered by the choices the customer
makes as he surfs the web and the specific site (i.e. gathering information from the
interactions of the consumer with the site). The first kind of information should be
protected as it is protected in the “real” world (outside the Internet). But the second
kind of information has different qualities to it and thus requires a different type of
protection. A common claim is that the Internet is a reflection of the real world and
that the regulation of Internet transactions should derive from regulations of the real
markets. Since the second kind of information gathering is not protected or limited in
the “real” world it shouldn’t be limited on the Internet. This argument doesn’t claim
that such information should be given to third parties but rather that it should be
allowed to gather this data for usage in that specific site.
Another strong argument in favor of personalization is of an empirical nature. It states
that consumers prefer personalization to privacy. While shoppers are very concerned
about their privacy -- particularly the sanctity of their identification-related
information – there is evidence that a majority of online shoppers do not mind their
behavior being watched if it allows their shopping experience to be customized.
(However, these customers expect to be notified and to have the ability to opt out).
According to a survey conducted by the research and information firm Privacy and
American Business, using Opinion Research Corp.'s weekly CARAVAN survey, 68
percent of the Internet users surveyed said that they would provide personal
information in order to receive tailored banner ads if notice and opt-out are
provided.10 (For more information about the survey see Appendix 4 on page 42).
Now, since the legal and regulatory systems are aimed at setting the best rules, which
reflect the preferences of the market itself, while deviating from those preferences
only in cases of market failures, and since this case doesn’t present any such failure,
then the free market approach should be preferred here. The lack of market failure is
based on the fact that the “weaker” side here is the consumer, and that the majority of
the consumers rather receive personalized service than protect their privacy (under
certain conditions as mentioned in the survey).
  This section was written with the assistance of: Dr. Tamar Gidron, Dean of Law School at The
College of Management; and Nimrod Kozlovsky who teaches a course on Law & Internet. In addition
throughout this section there has been extensive reliance on a book by Swire & Litan “None Of Your
Business”, Brooking 1998.

On the other hand there are those who claim that privacy considerations should take
preference over personalization ones. In response to the arguments raised before one
can say that first of all the Internet is not a reflection of the real world, the Internet
enables information gathering that was not possible before. Secondly the net enables
much easier transfer, usage and manipulation of that information with much less
ability to supervise such data flow and usage. These unique problems didn’t exist in
the pre-Internet age, and they require new legal solutions.
In addition, the fact that a survey showed a certain preference doesn’t automatically
mean that this is the “correct” solution. One can draw the paternalistic card and say
that we should protect the consumer even if he doesn’t want to be protected;
alternatively we should set regulation protecting the minority of such a survey.
In addition to rebutting the pro-personalization arguments, one can claim that a major
reason for the slow growth of the e-commerce markets is the privacy violation fear of
the consumers. It is not enough to promise the e-shoppers that their personal
information is safe (under the privacy policy of the web site). In order to help e-
commerce leap forward there should a feeling of safety. Even if the information is
protected under the strictest privacy policy in a certain site, the average surfer is still
accompanied by a sense of a “big brother” watching.
In order to deal with this issue e-commerce companies have a strong interest in
creating an appearance of privacy for e-shoppers. This interest runs on a direct
collision course with wanting to provide a more personalized site.
Another argument for protecting privacy could stem from the power given to
marketers from gathering and storing personal information. As mentioned before the
Internet enables gathering of information that wasn’t available before. This
information is equivalent to power held by those who control such information.
Having such power requires imposing limitations and restrictions on those holding the
power. Those limitations and restrictions take the form of protecting privacy.
Ways of protecting privacy11
Assuming that there can be significant problems in the protection of personal
information, the next question is to ask what institutions in society should be relied
upon to address such problems.
One institutional solution is to rely on the market. The basic idea is that the reputation
and sales of companies will suffer if they offend customers' desires about protecting
An opposite institutional approach would rely on government enforcement. The basic
idea is that enforcement of mandatory legal rules would deter companies from
abusing people's privacy.
A significant element of current thinking about privacy, however, stresses "self-
regulation" rather than market or government mechanisms for protecting personal
information. Numerous companies and industry groups have promulgated self-
regulatory codes or guidelines for the use of personal information.
Two additional options are: the media, and the technology. The media is helpful
especially in deterrence from negative publicity, but it has its drawbacks like the need
of attractive stories and victims. The technology can be useful in letting people do
business without revealing their identity, but cannot help much once the data are
already revealed.


Manna’s approach
From analyzing the privacy policies of Manna’s customers, and the lack of such a
policy by Manna itself, we learnt that the whole privacy issue is something that
Manna doesn’t deal with directly. Manna leaves that dilemma to its customers (the
marketers at the e-shopping web sites). This approach has its pros and cons. On the
pros side Manna’s solution is more generic because it doesn’t impose a specific
privacy policy on its customers, but rather allows them to create their own. On the
other hand, if we accept the argument that the personalization companies have an
interest in creating a privacy protected image or appearance, then Manna’s none
policy is not serving that collective goal.
Looking at Manna’s customer’s privacy policies we find a common principal of
promising not to disclose personal information to third parties without previously
getting the customer’s approval. More over we find that most of Manna’s customers
mention in their privacy policies that they preserve the right to use the information
(including disclosing to third parties) in the form of general and statistical information
(for example: 50% of our shoppers are women etc.). But we also see some grave
differences between the policies, for example some policies are more elaborate and
provide better protection for the consumers, while others are brief and open for
interpretations. Also one customer preserves the right to change its policy (they
probably got the same legal advise as DoubleClick got – above). For more
information you can see copies of the privacy policies in the Appendixes 5, 6, 7 on
pages 44, 46, 48.

The Manna team

Manna - management structure

                                            Dan Ross

     Moni Manor           Gary McMeekin                 Mike Murphy             Veronica O'Shea
      VP R&D           VP Professional Services    VP Business Development   Vice President of Sales

   Dr. Joel Ratsaby
  Group Manager, AI

A detailed analysis on Manna’s management can be found on appendix 10 on page


Manna – the future and beyond
We will conclude our case analysis presenting our expectations as to where Manna
Inc. is expected to proceed in the coming years. Manna’s future is inseparable from
two major shifts in the e-commerce industry and the personalization market. The first
is the major shift in the Internet environment entailed by the collapsing of the
“Dot.Com” bubble. The second is the fierce acquisition activities and mergers turning
the personalization market into a playground for giants. We will separate our
perspectives for the near future from our long-term predictions concerning Manna’s

In the immediate future Manna has to adapt to the new situation in the “Dot.Com”
environment. This includes minimizing the company’s “burn rate” in preparation for
the expected capital-drought. The company should also reconsider its pricing policy
and realize that the competition in the field will grow fiercer.

In the long term Manna has to develop a strategic plan as to how the company might
position itself in the personalization’s evolving market-of-giants. Manna’s cutting
edge technology is her main asset, and as such Manna will probably strive to join
forces with an e-commerce comprehensive solution giant. This cooperation will
probably take the form of acquisition rather than merger or strategic alliance. In order
to achieve this goal Manna will focus on building up the company’s value and
attractiveness by emphasizing on the following points:
             Developing and maintaining technological superiority
             Presenting clients success stories
             Expanding and improving the selection of customers with a major
                emphasis on cliental quality vs. quantity as to not tarnish the
                company’s success record
             Expanding Manna’s presence in the market. Building a brand name in
                the professional circle. Making Manna a synonym for excellence in
                personalization by using advertising and P.R.
             Become a financially independent company

Evidence to all of these trends could be seen by the present company’s policy and was
stated explicitly by Ms. Tali Aben, as one of Manna’s future goals. If Manna will
achieve these goals the e-commerce arena will enjoy a leap in the personalization-
standards of future web sites. This will conclude Manna’s story as the triumph of a
great technology in subduing the personalization market. If the company’s
management will fail to achieve the goals stated above, will evidence another
promising Israeli technology failing to transform into an actual financial and business
success story. We believe and hope for Manna that this will never happen.

                             Bibliography - data on competitors    - data on
Bayesian networks

Globes - Israel's Business Arena

Firm Privacy and American Business, using Opinion Research Corp.'s
weekly CARAVAN survey, 68

Swire & Litan “None Of Your Business”, Brooking 1998.


                                    Appendix 1
introduction to statistical inference using Bayesian networks
Graphical models are graphs in which nodes represent random variables, and the lack
of arcs represent conditional independence assumptions. We will focus on directed
graphical models (which cannot have directed cycles), also called Bayesian Networks
or Belief Networks (BNs). One can regard an arc from A to B as indicating that A
“causes” B. This can be used as a guide to construct the graph structure.
In addition to the graph structure, it is necessary to specify the parameters of the
model. For a directed model, we must specify the Conditional Probability Distribution
(CPD) at each node. If the variables are discrete, this can be represented as a table
(CPT), which lists the probability that the child node takes on each of its different
values for each combination of values of its parents. Consider the following example,
in which a page of an e-commerce web site specializing in
selling last centuries designers clothes from the 60’s to the 90’s. This simple network
might help analyze the optimal offers to a client shopping on the site. In this example
all nodes are binary, i.e., have two possible values, which we will denote by T (true)
and F (false).

                             P(C=F)         0.5
                             P(C=T)         0.5

                                      Client enters
  E P(C=F)       P(C=T)                                    P P(C=F)       P(C=T)
  F   0.5           0.5                                    F   0.8           0.2
  T   0.9           0.1                                    T   0.2           0.8

                    Present item                    Present item form
                   from Sale list                    60’s Retro list

                                    Item Within
    S R P(W=F) P(W=T)               shopping cart
    F F 1.0      0.0
    T F 0.1      0.9
    F T 0.1      0.9
    T T 0.01     0.99

We see that the event "Item Within shopping cart" (W=true) has two possible causes:
either the presentation of an item on Sale (S=true) or presentation of an item on the
Retro list (P=true). The strength of this relationship is shown in the table.
For example, we see that Pr(W=true | S=true, R=false) = 0.9 (second row), and hence,
Pr(W=false | S=true, R=false) = 1 - 0.9 = 0.1, since each row must sum to one. Since
the C node has no parents, its CPT specifies the prior probability that a client might
Enter this web page (in this case, 0.5).
The simplest conditional independence relationship encoded in a Bayesian network
can be stated as follows: a node is independent of its ancestors given its parents,
where the ancestor/parent relationship is with respect to some fixed topological
ordering of the nodes.
By the chain rule of probability, the joint probability of all the nodes in the graph
above is
P(C, S, R, W) = P(C) * P(S|C) * P(R|C,S) * P(W|C,S,R)
By using conditional independence relationships, we can rewrite this as
P(C, S, R, W) = P(C) * P(S|C) * P(R|C)               * P(W|S,R)
where we were allowed to simplify the third term because R is independent of S given
its parent C, and the last term because W is independent of C given its parents S and
We can see that the conditional independence relationships allow us to represent the
joint more compactly. Here the savings are minimal, but in general, if we had n binary
nodes, the full joint would require O(2n) space to represent, but the factored form
would require O(n 2k) space to represent, where k is the maximum fan-in of a node.
And fewer parameters make learning easier.

The most common task we wish to solve using Bayesian networks is probabilistic
inference. For example, consider the web site network, and
suppose we observe the fact that an item is in the shopping cart. There are two
possible causes for this: either the presentation of the Sale item, or the presentation of
the Retro item. Which is more likely? We can use Bayes' rule to compute the
posterior probability of each explanation (where 0==false and 1==true).


is a normalizing constant, equal to the probability (likelihood) of the data. So we see
that it is more likely that the item is Within the shopping cart because it was presented
on the Retro list: the likelihood ratio is 0.7079/0.4298 = 1.647.

Explaining away
In the above example, notices that the two causes “compete” to “explain” the
observed data. Hence S and R become conditionally dependent given that their

common child, W, is observed, even though they are marginally independent. For
example, suppose that an item is within the shopping cart, but that we also know that
it was presented on the Retro list. Then the posterior probability that the cause was the
presentation of a sales item goes down: Pr(S=1|W=1,R=1) = 0.1945
This is called "explaining away". In statistics, this is known as Berkson's paradox, or
"selection bias".

Top-down and bottom-up reasoning
In the web site example, we had evidence of an effect (item is
Within the shopping cart), and inferred the most likely cause. This is called
diagnostic, or "bottom up", reasoning, since it goes from effects to causes; it is a
common task in expert systems. Bayes nets can also be used for causal, or "top
down", reasoning. For example, we can compute the probability that an item will be
within the shopping cart given that a Client entered the web-page. Hence Bayes nets
are often called "generative" models.

Efficient Inference
A graphical model specifies a complete joint probability distribution over all the
variables. Given the joint probability distribution, we can answer all possible
inference queries by marginalization (summing out over irrelevant variables), as
illustrated in the introduction. However, the joint probability distribution has size
O(2n), where n is the number of nodes, and we have assumed each node can have 2
states. Hence summing over the joint probability distribution takes exponential time.
There are many methods to try and reach the goal of efficient inference, and we
refrain to only naming a few of them, including: Variable elimination, Dynamic
programming, Approximation algorithms and Variational methods (Monte Carlo
methods, "Loopy belief propagation", Bounded cutset conditioning and Parametric
approximation methods).

                                    Appendix 2

               ADI (Advanced Distributed Intelligence)

Bots and Agents have been part of Web technologies since the very beginning.
They’ve been called the first indigenous species of cyberspace. In fact, they’ve
roamed the network in anything from mailing lists to newsgroups and helped us in
everything from Web page indexing to comparison-shopping. Although agents are
commonly linked to Artificial Intelligence, they are rarely “intelligent.” Most
cyberspace agents are working at simple and repetitive tasks, following HTTP links,
and performing simple arithmetic calculations. We tend to perceive agent
technologies mostly as programs that index the Web or gather some information on
the behalf of the Web user (on-behalf computing). In fact, agent technologies can be
broadly defined as independent components of an application that have a task to
perform—typically on remote machines or on the network. The most common
examples of agent technology are programs that serve human beings by automating
Web searches and other tedious tasks. In presenting Manna’s approach, however,
we’ll look at agents of a different breed— agents that serve applications and agents
that serve other agents. We will also see agents that are really “intelligent.” Intelligent
soft-ware means that the program can learn independently, for the purpose of
improving itself, and digest new data into a form from which it can infer when
queried by a human or an application. This is the machine equivalent of a (very)
educated guess.

On-Behalf Computing
The essence of agent-based computing can be viewed as a natural extension of object-
oriented programming concepts: in particular, the isolation (encapsulation) of an
object in the system and the genealogy of objects. Agent-based computing adds the
concept of assigning tasks to objects. In effect, the agent acts on behalf of an entity
that assigns it to a certain task. This is a very powerful metaphor. The task is
decipherable only to the agent, which solely deals with its execution. Tasks are
usually engineered to access data. For example, a simple task could take the following
form: Look at the database and count the number of customers whose birth
date occurs in the next month—return that number to me. In this case the agent
behaves as a database interface (much like a business object) that executes tasks on
behalf of the requester. Agents are typically more generic than business objects—they
can deal with different data sources based on the task at hand, and even physically
move to remote machines to perform tasks locally. It’s important to understand that
the agents themselves need only know whether the task has been completed or not.
They can ignore the inner workings of the task. It’s convenient to have the task itself
be an object and as highly reusable as possible so that it can be shared, persisted, or
updated. Typical tasks call upon agents to deal with remote address spaces—either by
communicating with them through the network or by physically moving to them. This
is a natural effect of on-behalf computing—it has always been a challenge to execute
tasks on remote resources. In this case, the agent performs its task on the remote
machine or process. Such agents (or agent frameworks) need to deal with networking,

security, and failure-recovery in a specialized and expert way and, like their human
counterparts; they need to cooperate to be as efficient as possible.
Cooperative Computing
Agents can live in communities and communicate tasks or other information among
them. Also, in complex scenarios, agents will delegate subtasks to other agents and
create an agent subculture for complex processing. Cooperation between agents is
important, especially in complex, mission-critical scenarios. As you might guess, it’s
best to limit an agent’s expertise to very specific tasks, but to design the community
such that, as a whole, task management occurs in the most effective way. Also, it’s
prudent to divide the work on certain tasks among several agents so that if one agent
fails, other agents can help complete the task. As you can see, another interesting
concept here is that agents can be maintained independently of the tasks they have to
perform. This allows tasks to evolve over time without affecting the agents. The
FrontMind system was designed to be as simple as possible while allowing maximum
modularity. Having a system that completely separates the object’s data from its
behavior enables the creation of very generic objects that could be configured at
runtime. As we will see later, the agent’s data is passed around in DOM documents.
The data is stored in fields, whereas the behaviors are configured through a separate
Task object. The FrontMind system implements several sets (communities) of agents
whose cooperation works in favor of executing intelligent business rules in real time,
in large, mission-critical settings. In effect, the product includes four sets of agents
that are used for messaging, system maintenance, inference and prediction, and
learning. These cooperative communities work together in an organic manner that
gives FrontMind a very high degree of adaptivity and maintainability.
Distributed Agent Messaging
We’ve explained that community-based agents need to communicate on many
different levels. Communication is critical to any living community, including the
agent community— especially as it pertains to tasks. Since tasks tend to evolve and
change, agents need to agree on a common syntax that would be understood at many
different levels (provided you have the right dictionary). Over the past several years,
there have been several attempts at defining a common language for agents. These
languages have focused on communicating knowledge among agents and have a
number of advantages as well as clear disadvantages.
The main advantage is a proposed standard for communication of logic between
agents that can be written by different vendors — this is very important since on the
network, you never know to whom you’d need to talk.
However, there are also disadvantages. First, agents communicate much more than
mere knowledge. In mission-critical and complex agent systems such as FrontMind,
agents can communicate load and performance information, statistics, tasks and
subtasks, and localization and networking information, as well as error-handling
information and reports. Secondly, these languages are hard to parse and are not yet
standardized. Also, these languages are difficult to maintain and extend. Another
important aspect to consider for messaging is mobility. Agents need to communicate
with distant address spaces, and messages need to be constructed so that they can be
delivered reliably, persisted, and shared throughout the network.
In FrontMind, agent communication is performed through novel use of XML and
DOM—both W3C standards. XML-ized messaging can easily capture a lot of the
messaging functionality needed in FrontMind. XML applications such as XSL and
XQL offer logical processing to XML content, which is very powerful. The
FrontMind communication system allows tasks and other runtime information to be

defined independently as DOM documents and passed along to other objects or
agents. An agent holds and maintains a DOM document that defines its state and that
document can be updated, persisted to XML (for example, into an XML database),
shared with other agents, or propagated throughout the system. DOM was an excellent
candidate for the definition of tasks, since it defines a generic object model for data
that is independent of any application-specific object model. DOM, in effect, is a data
structure. The use of XML lets us adhere to a standard that’s supported by all major
vendors and for which many fine utilities exist—most of them free of charge. In
addition, tasks can be defined using common tools, such as Internet Explorer 5.0, and
stored in databases that support XML.

The FrontMind messaging agents are movable; they get a document that defines an
itinerary for them to visit (stopping by various docking objects) where they can
dispense fragments of their document as needed. This itinerary is physical—that is, it
can span multiple address spaces. The messaging agents also take a central part in the
systems-failure recovery mechanism and let the system respond immediately to
execution problems. Messaging agents are also used by other agent communities
outside their native, low-level communities. These other communities, in FrontMind,
are a community of maintenance and housekeeping agents and two intelligent species
of agents: learning agents and inference agents. These agents share a common set of
Bayesian models, outlined in the previous chapter, which constitutes their common
knowledge base.

Maintenance Agents
The ‘‘working class’’ in FrontMind is a set of agents (called engineers) that take care
of keeping the system available and open for business 24/7/365 - no down time. These
agents can install a specific set of plug-ins (a set of application-level Java classes) into
the FrontMind runtime, remove such a set, or update it all without any effect on
system performance. The agent can get real-time information on the state of the
system and make an informed decision on how to best perform its task. This task,
again, is defined in XML and can be of varying complexity. For example, an engineer
(like its sibling, the messaging agent) can receive an itinerary and travel throughout a
FrontMind system (on its physical and virtual machines) to complete its task. A
typical task for engineers is to maintain the intelligent business rules. Here, the
failure-recovery and error-reporting features are critically important. The agent’s task
must be 100- percent recoverable at any time, and there must always be an agent
ready to replace a failed agent. Other forms of housekeeping agents are those that deal
directly with load balancing in the system. These agents monitor the load on the
system (at the virtual machine level as well as at the physical machine level). This
will be discussed again later, but I’ll say for now that these agents communicate with
the learning agents, creating a unique and intelligent learning load-balancing system,
where load patterns are learned and applied to better prepare the system for load

Learning Agents
The learning agents digest data to a form that can readily be used to update the
statistical models used in FrontMind. For example, a statistical model can be used to
model a section of a Web site or the whole site. Take a Web site that specializes in the
retail of electronic appliances. The system can model this site to have different model
mappings. The site itself would be the “top” (or global) model that contains all the
other models. Immediately beneath it in the model chain would be a model for kitchen
appliances, another for entertainment appliances, another for household appliances,
and so on. If we continue and choose a sub model of entertainment appliances—say,
TVs— this model can contain a model called Digital TVs. In this case, we deal with a
hierarchy of statistical models that’s mapped to a hierarchy of agents—a TV agent
and a Digital TV agent. Let’s focus on the TV agent, which also contains the Digital
TV agent. The TV agent contains and learns statistical information on HDTVs,
Digital TVs, Stereo TVs, and so on. The Digital TV agent is more specialized—it
knows and studies its business of Digital TVs, and holds statistical information on
different Digital TV brands (Sony, Hitachi, and so on). This is very much like
salespeople in the real world, in which each salesperson specializes in an area of the
department store. There is an obvious propagation pattern here of task messages
among the agents. What is being transferred is data used for calculations on one hand
and statistical updates on the other. You can see that every agent is highly specialized
and its domain of knowledge is limited to what it knows best. Every agent deals with
learning data that relates to its learning task and propagates that data through the
hierarchy chain. The agent needs to perform fairly complex calculations while taking
great care not to abuse system resources such as CPU time and memory. In addition to
parent/child relationships among models, there’s also a relation-ship among siblings.
The Digital TV agent can relate to the HDTV agent if certain conditions hold.

Inference Agents
The inference agents can be viewed as an interface to a special kind of database that
holds only statistical data. In effect, these agents reflect the way in which learning
agents are organized. They are mapped exactly to the same model hierarchy. In other
words, if the learning agents are the passive part of the person’s intelligence—the
learning part—the inference agents are the active part—they act on the data to
produce accurate predictions based on statistical data. The inference agents can
receive queries from either humans or the FrontMind application-server itself and
return a prediction on the value requested. These agents make business sense. For
example, an inference agent can be approached with the following query: Predict
which is the best TV set to propose for a certain customer given that he is
currently visiting the HDTV section of the site. This query will traverse the agent
graph from the top agent to the one whose expertise is in HDTV and can concurrently
communicate with the agent or agents that specialize in customer profiles. In
FrontMind, the return values of such a query can be anything from a rich set of values
to choose from to a single value indicating that it is the best value for your query. The
FrontMind inference agents work in real time. This means that they respond to
extremely timely information and the results of the query are very accurate. This also
means that they need to know how to deal with a high load volume and perform very
fast. This breed of agents not only needs to deal with fairly complex calculations but
at the same time optimize its own use of resources and, if necessary, migrate portions
of its calculations to other machines with more resources.

The FrontMind Federation of Agents
Agents offer a rich metaphor for designing business applications. It makes sense to
perceive agent communities as organic and natural components in an application’s
flow of logic. Specifically, we’ve seen that there’s a high degree of communication
among agent communities in FrontMind. The whole system is designed such that it
creates a federation of agents in which all the different agent communities speak the
same language (based on DOM and an XML dialect), while specializing in very
different tasks. The cooperative metaphor allows FrontMind to achieve exceptional
reliability and stability, since, although the interactions and tasks are certainly
complex, the system itself is strikingly simple. There’s a full pseudo-biological circle
in the system in which every agent community relies on the other and at the same time
serves it. The load-balancing agents rely (for certain tasks) on the learning and
inference agents. At the same time, the learning and inference agents—dealing with
large amounts of data, in real time, need to rely on the load-balancing agents to handle
the amount of load a live e-commerce system expects to take. The learning and
inference agents communicate more “intellectual” data among them. Obviously, this
form of cooperation uses the messaging agents for its interagent communication.
Because of its inherent modularity and flexible componentization this form of design
leads to systems that are easier to maintain and more ready to evolve. The agent
metaphor achieves a dissociation of application data from its object-oriented behavior.
Web standards such as XML and DOM and powerful object-oriented languages like
Java have allowed us to create a unique multi-community, multi-agent system that
answers real-world needs while being stable and reliable to meet enterprise-level

                                      Appendix 3

                           Manna Major Competitors
                                   Net Perceptions

Company Background
The original GroupLens product launched in 1996 included a scheme for capturing
explicit user ratings and preferences and an API for integrating the recommendation
and prediction engine into a commerce web site. was an early adopter of
the GroupLens toolkit, a tremendous victory for Net Perceptions that helped generate
a good measure of publicity and momentum for the start-up company and the
fledgling web personalization industry as well.
Net Perceptions is still in the early stages of trying to establish a healthy indirect sales
channel. Partners contributed just 15% of total revenues in the first quarter of 1999,
up from 8% of total revenues for 1998. Net Perceptions partners include,
Aisle5, Anderson Consulting, Camworks, IBM Global Services, iXL, Organic, and
US Web.
Net Perceptions has signed a partnership to integrate its new Net Perceptions for Call
Centers product with CommercialWare, a provider of call center order management
and fulfillment software for catalog companies.
Net Perceptions financial picture is typical of many Internet software company start-
ups: heavy losses incurred to rapidly grow the business offset by increasing revenues
and a tremendously successful IPO. The April IPO raised an impressive total of
$53,207,000 for the company and, as a result, Net Perceptions reports holding a
healthy $43.3 million in working capital.
Net Perceptions reported total revenues of $2.8 million in the second quarter of 1999,
Net Perceptions for E-commerce is priced based on the number of profiles stored in
the engine's database. A basic system starts at $80,000 for up to 100,000 registered
Product Description
Net Perceptions now offers a range of products that incorporate its core
recommendation engine technology: Net Perceptions for E-commerce, Net
Perceptions for Call Centers and Net Perceptions for Marketing Campaigns.
Net Perceptions for E-commerce™ is primarily a collaborative filtering engine and a
set of tools designed to simplify integrating collaborative filtering with an e-
commerce web site. Collaborative filtering is the process of evaluating and
recommending content based on the combined preferences of an individual and other
similar individuals. "Content", in this context, can be anything from news to movies,
books, brands of cat food, or even people. Users typically interact with the system by
registering with a web site, rating a set of content, and then asking for
recommendations. Net Perceptions software then attempts to predict what content will
be of interest to the user and displays the content to the user.
The technology does this by learning about each individual’s preferences. There are
two primary methods for discerning user preferences with Net Perceptions
technology: explicit and implicit. Explicit learning involves directly asking users to

rate content, typically through an on-line form. Implicit learning involves letting the
system infer ratings by observing behavior.
Net Perceptions claims more than 110 customers for its real-time recommendation

Company founded in 1993 .BroadVision offers an extensive product line of tools and
applications to build complete, database-driven businesses on the Web. BroadVision
4.0 integrates user profiling, online catalog creation, transaction handling, rules-based
personalization, and content management. BroadVision applications are built on a
logical framework of customizable "components", "templates", and "rule sets" that are
created, managed, and deployed with three primary interface tools: The Design
Center, the Command Center, and the Publishing Center. The Design Center is a
Windows-based visual Web development environment. The Design Center integrates
Macromedia's Dreamweaver design tool and offers a series of wizards to aid in the
creation of dynamic applications. The Command Center is also a Windows-based
application designed to allow business users to make changes to the Web site, update
and categorize content, and define personalization rules. The Command Center was a
primary differentiator when it was first introduced, but in today's distributed Web
world, its complexity and fat-client architecture are significant liabilities. BroadVision
has the unusual distinction of being a profitable company in the Internet space.
BroadVision competes with other commerce and content platform providers like Art
Technology Group, IBM Net.Commerce, the Sun/Netscape Alliance, and Vignette
BroadVision develops and delivers an integrated suite of packaged applications for
personalized enterprise portals. Global enterprises and government entities use these
applications to sell, buy, and exchange information over the web and on wireless
devices. The BroadVision e-commerce application suite enables companies to become
more competitive and profitable by establishing and sustaining high-yield
relationships with customers, suppliers, and employees.
BroadVision reported financial results for the quarter ended September 30, 2000.
Revenues for the quarter were $120.2 million.

V/5, Vignettes’ flag product is a framework for component based content creation and
management. Components are database-driven, dynamically driven content. Vignette
calls its V/5 product an Internet Relationship Management application platform. V/5
incorporates three major features: content management, improved lifecycle
personalization, and decision support. The three agents that enable lifecycle
personalization -- Presentation Agent, Matching Agent and Recommendation Agent --
are each useful at a particular stage of the customer relationship lifecycle, from arrival
to the stage where the visitor becomes a trusting customer. The Recommendation
Agent uses a collaborative-filtering technology based on Net Perceptions
Recommendation Engine. Related to personalization, Open Profiling Services adds
three key capabilities: content catalog, visitor registry and the observation manager.
These features manage and populate a centralized repository of visitor profile and
content information. The observation manager gathers and stores information about
visitor activity and behavior. V/5 also includes complete developer and user interface

tools called the Production Center, the Business Center, and the Development Center.
Vignette's visual tools for building, managing, and changing content components are
some of the there are for enabling users to rapidly accomplish simple tasks. Vignette
does not, however, offer the transaction handling capabilities that BroadVision has
integrated into its Web application platform, choosing instead to create an open
interface to transaction software.

                        Art Technology Group (ATG)
ATG's Dynamo Personalization Server enables the creation of dynamic web
personalization through rules-based content targeting and delivery. The Dynamo
Relationship Commerce Suite consists of a Personalization Server and a Commerce
Server, each of which extends the functionality of the Dynamo Application Server.
Art Technology Group partners with collaborative filtering vendors to offer this
functionality. The administrative interfaces for the personalization control and the
developer's workbench are both well-designed for their intended users. The
developer's workbench provides an html editor and a component manager to create
and modify Dynamo Server Pages. The personalization control center, where
marketers and business user's create profiles and target user groups, is simple to use
and should pose no problems for the technology-averse user.

An early pioneer in the e-commerce marketplace, ATG has consistently set itself apart
as a leader in delivering Web applications built specifically for the needs of the
growing online commerce market. Since its founding in 1991, ATG has enabled
Global 1000 companies to revolutionize how they establish and maintain business-
critical relationships with Web site customers.
Through its Dynamo family of products, ATG helps companies build and manage
Web environments that use personalized content to promote customer satisfaction.
Dynamo's personalization system is based on a high-performance, server-side Web
application platform and targeting engine, and it offers a complete set of commerce
applications for building and managing online storefronts, promotions, billing, and
advertising. It is used by such clients as BMG Direct, Eastman Kodak, John Hancock
Funds, Newbridge Networks, Scudder Kemper Investments, Sony, and Sun

Customer Management is the essential practice of establishing and maintaining
profitable, long-term relationships with customers. The ATG Dynamo e-Business
Platform is a major step forward for businesses, suppliers and partners that want to
manage their customers better and significantly improve their e-commerce efforts. It
offers proven personalization and a wealth of advanced e-commerce features.

                                  Appendix 4
    E-Shoppers Choose Personalization Over Privacy
By Paul A. Greenberg
E-Commerce Times
January 4, 2000
                  The battle lines have been drawn for one of the great e-
                  commerce clashes of 2000. Proponents of online privacy and
                  innovators of site-specific personalization are set to struggle
                  with the big question: At what point does personalization
                  become an invasion of an online shopper's privacy?

                 It is a dispute not likely to be settled soon, as technology is
creating ever-more sophisticated methods of gathering data about online
consumers, while proponents of privacy push for laws that govern the online
shopping experience.

Shoppers Opt For Personalization

While shoppers are very concerned about their privacy -- particularly the sanctity
of their identification-related information -- evidence is beginning to develop that
a majority of online shoppers do not mind their behavior being watched if it
allows their shopping experience to be customized. However, these customers
expect to be notified and to have the ability to opt out.

According to a survey conducted by research and information firm Privacy and
American Business, using Opinion Research Corp.'s weekly CARAVAN survey, 68
percent of the Internet users surveyed said that they would provide personal
information in order to receive tailored banner ads if notice and opt-out are

The survey was underwritten by a grant from Net advertising giant DoubleClick
and was conducted by randomly interviewing 471 adults who identified
themselves as Internet users.

Sixty-one percent said that they favored receiving banner advertisements tailored
to their preferences versus receiving random banner ads, while 58 percent agreed
to having their visits to Web sites used to personalize banner ads to them, if
notice and opt-out were provided.

Fifty-one percent surveyed agreed to having their online purchase information
used to personalize banner ads to them, if notice and opt-out choice were
provided, while 53 percent of users would be willing to have their offline purchase
information from catalogs and stores used to personalize banner ads to them, if
notice and opt-out choice were provided.

Fifty-three percent of those surveyed said they would not mind if the combination
of personal information, Web site visits, and online and offline purchases were
used to personalize banner ads to them, if notice and choice were provided.

Shoppers Seem To Back E-tailers

The study seems to back e-tailers and independent research firms who are
watching online behavior to provide or facilitate customized shopping
experiences, and not privacy advocates who say that this process is an invasion
of privacy.

While shoppers do not seem to have a problem, there are many analysts who do.
Forrester Research analyst Christopher Kelley is skeptical. "The whole idea is very
Orwellian, just the sense that Big Brother is watching over you," he said. "The
problem with the Internet is becoming, for some customers, that the technology
allows people to look over your shoulder at every turn."

Personalization Versus Invasion of Privacy

The purpose for gathering the information seems to be the key determinant
between acceptable personalization and invasion of privacy. The majority of
consumers seem to like the practice if information is gathered to provide a
custom online shopping experience.

It is also important to many shoppers that the site has a privacy policy that
explains what information is gathered and how it is being used.

Should Government Join the Fray?

As for governmental intervention, the Federal Trade Commission (FTC) continues
to exercise a hands-off approach, but that could soon change.

U.S. Congressman Edward Markey (D-Massachusetts) believes that it is time to
show greater respect for the privacy of American consumers. "Ordinary
Americans say 'I want my privacy, and I don't know why the government or
corporate America doesn't give it to me,'" Markey said. "There's a looming
confrontation on the cyber-political horizon that will ultimately determine the
extent of the full privacy protections of all Americans."

Toward that end, Markey introduced an electronic privacy bill of rights last
summer that sets guidelines for e-businesses that gather data about their
customers. The bill has three provisions:

The first grants individuals the right to know that information is being gathered
about them. The second grants users the right to know whether information is
going to be reused by the organization that is gathering it. The third gives
individuals the right to block the gathering of information about them

                                          Appendix 5

                  Harcourt, Inc. Customer Privacy Policy12

             1. Privacy Statement
             2. Collection and Storage of Information
             3. Use of Information
             4.   Your Rights

1. Privacy Statement

Harcourt, Inc. recognizes the importance of protecting the privacy of our customers and
visitors to our Web site while permitting us to conduct legitimate business by providing
services and information of interest to our customers and visitors. Our privacy policy is stated

2. Collection and Storage of Information We will collect certain personal information
when you register for our services or purchase our products. During registration we will ask
you to set up a user name and password to establish secure access to your personal
information. We will also ask you for your name, address, phone number, and email address.
This contact information that you provide at registration allows you to:

       Buy items without having to re-enter your personal information every time you
purchase a product
       View all your purchases and shipments from the Harcourt Store 24 hours a
day, seven days a week
When you purchase a product, we ask you for your name, billing address, shipping address (if
different from billing address), phone number, email address, and payment information. The
payment and contact information is used to complete and verify the transaction, and your
email address allows us to contact you regarding the receipt, completion, and shipping of your
transaction and product. We maintain this information in our secure customer database. From
time to time, we may also collect additional information, such as in connection with special
offers, contests, or other promotions.

We may use "cookies" to recognize you and your access privileges on our site, as well as to
trace site usage. A cookie is a piece of information that is sent to your Web browser from a
Web site and stored on your computer's hard drive. Each time you return to our site, we are
able to identify you as a previous or registered customer. Cookies also enable us to measure
our site traffic, including the number of repeat versus new visitors, the time users spend
online, and their navigation behavior. Collecting this information allows us to make changes
that improve our site.

3. Use of Information
The information that we collect is used to provide the service or product requested; to enable


billing and shipping; to provide you with information about related products and services; to
improve our Web site, for development of new products and services; and for systems
administration and troubleshooting purposes.

We will not share your email address with parties outside the Harcourt family of companies,
and we offer you the option to not have your other information shared with parties outside the
Harcourt family of companies. Personal information, such as mother's maiden name or credit
card numbers, will not be shared with third parties unless you are notified at the point where
the information is collected, such as during a promotion or special event. Information
concerning individuals' usage of our sites will not be disclosed to third parties except in
aggregate form (e.g., "45% of the customers visiting our site purchase at least one product").

4. Your Rights
You may review and approve the information about you collected upon registration that is
housed in our secure customer database by accessing "Manage My Account" from the main
menu. Upon your request, we will correct personal information that you state is erroneous or
remove any information from our customer database. Requests for the correction or removal
of information should be directed to In addition, please understand
that information about you in our databases may come from a number of sources, including
orders placed by you and third-party mailing lists.

Privacy Contact: Questions regarding this policy should be directed to Harcourt, Inc. at Harcourt is not responsible for the privacy policies of Web sites to
which this site may link.

Due to the diverse nature of our businesses, companies within the Harcourt family may have
privacy policies that differ from this policy. In such cases, the privacy policy of the particular
Harcourt Company will be posted, and will supersede the terms of this document.

                                         Appendix 6
                 Privacy Policy13 respects your right to the privacy of the personal information you
provide us on our Web site. To that end, unless you designate otherwise or we state
otherwise in this policy or at the time information is collected, any personal
information you provide to will be known to three parties: you, and the vendor that the order is placed with. The following policy is
intended to explain how your personal information will be treated as you make use of
our site and its features. Personal information includes your name, age, gender, street
and e-mail addresses, billing information, click-through activity and any other
personal information you may provide here. This policy may change from time to
time, so please check back periodically.

Traffic Data
 Each time a visitor comes to, our servers (like most on the Web)
collect some basic technical information, including, for example, the visitor's domain
name (e.g., whether the user is logged on from or, referral
data (e.g., we record the address of the last URL a user visited prior to clicking
through to a Web site) and browser and platform type (e.g., a Netscape
browser on a Macintosh platform). We also count, track and aggregate the visitor's
activity into our analysis of general traffic flows at our sites (e.g. tracking where
traffic comes from, how traffic flows within the sites, etc.). To these
ends, we may merge information about visitors and visits into group data, which may
then be shared on an aggregated basis with our advertisers; but we will not disclose
your individual identity or personally identifiable data without your permission. When
we do present aggregated information to outside companies, no one will be able to
identify you or contact you. Shopping and its shops are located at In the course of
processing orders, we will collect certain personal information for the purpose of
fulfilling customers' orders. We will save this information as needed to keep
responsible records and handle complaints. We do not share this information (other
than in aggregate form, which does not reveal your personal identity or contact
information) with any other companies or individuals except those third-party vendors
that are directly involved in fulfilling the order. We have no control over what a third-
party vendor does with a customer's information once an order is placed with them.
We are however, in agreement with these vendors that they will not abuse a
customer's rights of privacy. If you feel that you are unduly being targeted by a
particular vendor via e-mail, and are not able to opt-out of their promotions, contact
us at and we will ensure that the issue is properly rectified. website security
Any orders that are processed through the website (including credit
card information) are protected by encryption software and Secured Socket Layers
(SSL) technology. is committed to continue to update it's technology
in order to protect any critical data that is processed through it's website.


                                                 46 Membership collects personal information from visitors who become registered users, including first and last names, e-mail address, basic
demographic information, street address, and phone numbers. After signing up as a
registered user and opting in to our promotions, you will receive various Sale E-Mails
and saleoutlet promotional offers from time to time. These e-mails are delivered on a
daily or weekly basis, as well as infrequent other announcements from
We collect member's e-mail addresses in order to distribute those newsletters and
announcements. We do not currently share this information with other companies and
will not do so (except in aggregate form, which does not reveal members' identities)
without first obtaining a member's consent. Users can choose to unsubscribe from at any time by following the unsubscribe instructions found in each e-
mail or by e-mailing us at

From time to time, may sponsor sweepstakes, contests or other events,
which require the collection, processing and storage of individual contact and
demographic information (e.g., your age and state of residence) to be used for the
event. Saleoutlet may also co-sponsor these events along with other companies, in
which case the user's individual contact and demographic information is likely to be
shared with other participating sponsors, who may use this information as they wish.
A registrant's personal contact and demographic information will not be released to
any third-party organization other than named sponsors without the user's consent,
and will request that such named sponsors uphold the privacy of the
registrants' information by not distributing or sharing such information with any other
businesses or organizations.

Our Commitment to Children's Privacy
Protecting the privacy of the very young is important to us. For that reason, we do not
collect or maintain information at our Web sites from those we actually know are
under 13, and no part of our Web site is structured to attract anyone under 13.

Other Disclosure of Personal Information will not otherwise use or disclose your personal information without
your consent except (i) as described in this privacy policy, or (ii) as required by law,
court order or as requested by other government or law enforcement authority. This
policy does not protect information you post to any on-line bulletin board, chat room,
newsgroup or other public forum within our Web sites.

Unsubscribe and Data Removal Policies
If you would like to unsubscribe from a e-mail list, please follow the
instructions at the bottom of the e-mail itself, or e-mail us at

Other Sites contains many links to other sites on the Web. We cannot control and
are not responsible for the privacy practices or content of such other Web sites.

                                         Appendix 7

                 Privacy Statement for GetOutdoors.com14
At, we make our members' concerns about security and privacy on
the Internet our top priority.
We promise to do the following to ensure your privacy:

          We never sell, release, or make public in any way, your personal and private
          We provide only general group statistics about our membership base to
           GetOutdoors partners and affiliates.
For example, we may say that we have a 50:50% ratio of men to women who get
outdoors with us, but we never publish your individual name or personal information
without express permission from you.
Our site does have links to other sites. is not responsible for the
content or privacy practices of such Web sites.

Public Forums and Events
This site makes event information, forums, message boards, reviews, and/or groups
available to its users. Please remember that any information that is disclosed in these
areas becomes public information and you should exercise caution when deciding to
disclose your personal information.

This site has security measures in place to protect the loss, misuse and alteration of
the information under our control.


                                   Appendix (U.S) 8
         DoubleClick Sued for Online Privacy Invasion15
                                          By Robert Conlin
E-Commerce Times
                       January 28, 2000
                       As consumer fears about privacy invasion threaten to stall the
                       considerable growth of e-commerce, a lawsuit was filed in
                       California Superior Court Thursday alleging that Internet
                       advertising firm DoubleClick (Nasdaq: DCLK) has obtained and
                       sold the personal information of online users.

The suit, filed on behalf of Harriet Judnick and the citizens of California, claims
that the New York-based DoubleClick tracks Internet users and obtains personal
and financial information such as name, age, address, and shopping patterns
without their knowledge.

Additionally, the lawsuit contends that DoubleClick has previously stated that it
does not collect personal information and that it strives to protect the privacy of
Internet users.

DoubleClick has declined to respond publicly to the lawsuit.

DoubleClick Versus the Consumer

Many observers believe that "third-party" ad banner delivery companies like
DoubleClick are among the most egregious of privacy offenders, because they
seem to prefer corporate marketing interests over the interests of online
consumers. Internet ad banner serving -- while ubiquitous -- is still largely
unregulated and devoid of standards, resulting in a lack of consumer protection.

DoubleClick operates in a fiercely competitive industry, with competitors like 24/7
and others involved in a wave of corporate buyouts and consolidation. Ultimately,
these efforts by DoubleClick to harvest user information could quite possibly be a
desperate and ill-advised policy aimed at differentiating their services from those
of their competitors.

Bad News for Consumers

At the heart of the matter is DoubleClick's recent $1.7 billion (US$) acquisition of
Abacus Direct Corp., a direct-marketing services company that maintains an
extensive database to catalog the purchasing power of American households.
Internet privacy advocates objected to the acquisition at the time, saying that it
would create the opportunity for the exploitation of online users.

DoubleClick did acknowledge earlier this week that that it is forming alliances with
Web sites to create a network that allows it to track users' online data and
shopping habits.

The company, which distributes advertisements on about 1,500 Internet sites,
recently started linking tracking features called "cookies" to people's names and
addresses in an effort to compile data. It does allow users to "opt out" of
providing personal information on its own Web site, but some critics say that


dense language and difficult access make it possible for few users to ever
navigate to the feature.

Seeking Injunction

Judnick's lawsuit is asking for an injunction against DoubleClick that would
require it to stop using its technology to collect personal data without prior
consent of the user. It is also asking that the company be required to destroy all
records that it has previously obtained without a user's consent.

DoubleClick reported revenue of $98 million for the nine months ending in
September, while it booked escalating losses of $17.9 million for the same period.

The EPIC complaint16
The Electronic Privacy Information Center today filed a complaint with the Federal
Trade Commission concerning the information collection practices of DoubleClick
Inc., a leading Internet advertising firm, and its business partners. The complaint
alleges that DoubleClick is unlawfully tracking the online activities of Internet
users and combining surfing records with detailed personal profiles contained in a
national marketing database. EPIC is asking the FTC to investigate the practices of
the company, to destroy all records wrongfully obtained, to invoke civil penalties,
and to enjoin the firm from violating the Federal Trade Commission Act.

The EPIC complaint follows the merger of Doubleclick and Abacus Direct, the
country's largest catalog database firm. Doubleclick has announced its intention to
combine anonymous Internet profiles in the Doubleclick database with the
personal information contained in the Abacus database.

EPIC's complaint alleges that the DoubleClick merger of the two databases
violates the companies' assurances that the information it collects on Internet
users would remain anonymous, and that the data collection was therefore unfair
and deceptive. EPIC also charges that the company has failed to follow its revised
privacy policy and that this is also unfair.

Marc Rotenberg, Executive Director of EPIC, said, "This complaint against
Doubleclick is a critical test of the current state of privacy protection in the United
States. We are looking to the Federal Trade Commission to see whether
companies that break their promises and collect personal information in an unfair
and deceptive manner will be held accountable."

David Sobel, EPIC's General Counsel, said that "today's complaint raises
fundamental issues involving electronic commerce." He noted that "much of the
information collection that occurs on the Internet is invisible to the consumer,
which raises serious questions of fairness and informed consent."

The Electronic Privacy Information Center is a public interest research organization
in Washington, DC. EPIC's activities include the review of governmental and
private sector policies and practices to determine their possible impacts on
individual privacy interests.
The text of EPIC's complaint against DoubleClick is available online at:


Privacy in the Telecommunication sector (NTIA)17


As the National Information Infrastructure (NII) is built, more and more
individuals will use it for a wide range of transactions. In the course of
using the NII, individuals will create information trails that could provide
others, in the absence of safeguards, with the personal details of their

In this White Paper, the National Telecommunications and Information
Adminstration (NTIA) hopes to contribute to the broader privacy debate by
addressing the privacy issues related to a specific sector -- the
telecommunications sector. Specifically, this paper focuses on the privacy
concerns associated with an individual's subscription to or use of a
telecommunications or information service. The overall purpose of the paper is to
provide an analysis of the state of privacy in the United States as it relates to
existing and future communications services and to recommend a framework for
safeguarding telecommunications-related personal information (TRPI).

The analysis provided herein reveals that there is a lack of uniformity among
existing privacy laws and regulations for telephony and video services. In fact,
similar services are governed differently depending on how they are delivered.
And, other communications services like those available over the Internet are
almost entirely unprotected. Furthermore, NTIA believes that it will become
increasingly difficult to apply existing privacy laws and regulations to
communications service providers as services and sectors converge, and as new
technologies evolve.

To rectify limitations in existing telecommunications privacy law and to provide
consumers with a uniform privacy standard for TRPI, NTIA proposes a framework
that draws upon the Information Infrastructure Task Force's NII Principles for
Providing and Using Personal Information. This framework has two fundamental
elements -- provider notice and customer consent.

Under this proposed framework, telecommunications and information service
providers would notify individuals about their information practices, abide by
those practices, and keep customers informed of subsequent changes to such
practices. Service providers would be free to use information collected for stated
purposes once they obtain consent from the relevant customer. Affirmative
consent would be required with respect to sensitive personal information. Tacit
customer consent would be sufficient to authorize the firm to use all other

NTIA believes that establishing minimum privacy protections across the
communications industry would ensure that consumers are provided with a
reasonable level of privacy protection. Uniformly applied, a common "base"
standard could also prevent some industries from gaining an unfair competitive


                                              Appendix 9

       Overview of the European Directive on data protection18
The European Union Data Protection Directive (the "Directive") was adopted in 1995 and
takes effect on October 25, 1998. The Directive is sweeping in scope, applying to all
"processing" of "personal data," with only limited exceptions.(56) Processing is a broad term
that means "any operation or set of operations which is performed upon personal data,
whether or not by automatic means."(57) Personal data is a similarly broad term, meaning "any
information relating to an identified or identifiable natural person ('data subject')."(58)

Pursuant to the Directive, each E.U. member state must adopt a strict privacy law that
provides clear rights to data subjects. When collecting information from an individual, those
processing data (known as the "controllers") must disclose their identities, the purposes for
the processing, and other information.(59) Data can only be processed for the announced
purposes,(60) contrary to the common U.S. practice of permitting a company to use personal
data for unlimited purposes. Before data can be provided to third parties for direct marketing,
the individual must be informed and have the right to opt out free of charge. (61) Those
processing personal data must guarantee that individuals have access to their own personal
data and the opportunity to correct that data.(62) Other rules apply, such as special restrictions
on the processing of sensitive data, including information about racial or ethnic origin,
political opinions, or the processing of data concerning health or sex life.(63)

In considering enforcement of these rights, it is vital to recognize that the Directive does not
itself apply to any behavior; instead, the Directive requires each E.U. member state to
promulgate a law that complies with the Directive's terms. Actual enforcement will thus take
place under the law of a particular member state.(64) Each country must establish one or more
data protection agencies, known as "supervisory authorities," to help implement privacy
rights. Supervisory authorities are required to have investigative powers, "effective powers of
intervention," and the power to engage in legal proceedings or to bring violations to the
attention of judicial authorities.(65)

In practice to date, supervisory authorities have usually worked informally with controllers
when complaints are filed. In many instances, the controller explains why the practice in fact
complies with applicable standards, or else agrees to modify the objectionable practice. This
non-litigation approach is likely to predominate under the Directive as well. Nonetheless,
more formal sanctions have been and will be used under national laws, including the ordering
of the erasure of data and bans on transfers of data to jurisdictions with weak or non-existent
privacy laws.(66) In addition to administrative remedies, member states are required to provide
for the right of every person to a judicial remedy for breach of privacy rights.(67)

Article 25 of the Directive, governing transfers of data out of the E.U., has drawn special
attention. Article 25 allows transfers to third countries (i.e., non-member states) only if the
third country ensures an "adequate" level of protection.(68) Although the meaning of
"adequacy" will only be clarified with time, many European officials believe that the U.S.
lacks adequate protection, at least for some important sectors.

Where there is not adequate protection, flows of personal information from Europe to the U.S.
would be permitted only under one of the derogations (exceptions) in Article 26. One
important exception is if the data subject has given consent unambiguously in advance to the
transfer.(69) Another is where the transfer is necessary for the performance of a contract, such
as providing the name and address for shipping a purchase into Europe.(70) A different type of


exception is where a supervisory authority believes there are "adequate safeguards" of
privacy, such as where the transfer takes place under a contract that ensures that European-
style rules will apply in the third country.(71) Unless one of the derogations is satisfied,
transfers of personal data are not permitted to countries that lack adequate protection of

                                  Appendix 10
                           Manna – management
Dan Ross, CEO
Dan, a twenty-year veteran of the high tech industry, brings to Manna a proven track
record in developing startup companies into successful industry leaders. Dan joined
Manna after an extremely successful tenure as vice president of worldwide sales at
Open Market, Inc. Dan grew Open Market from a small startup in 1995 into the
dominant global leader in the marketplace with a 30 percent market share in the e-
commerce industry (Dataquest). In addition, Dan contributed heavily to Open
Market's being recognized as one of the top IPO successes in 1996 according to The
Wall Street Journal. Prior to Open Market, he served as vice president and general
manager of the Americas at PictureTel Corporation. Before PictureTel, he was at
Digital Equipment Corporation for sixteen years where he held several senior
management positions. Dan holds a BS from the Wharton School of Finance.
Moni Manor, VP R&D
Moni, a 16-year engineering veteran, was for the past five years VP Development at
Jacada (a 130 person software company), and chief architect of their 8 products.
Moni's delivery of high-quality products has led to Jacada's leadership in the graphical
interface market. Prior to Jacada, Moni was a software engineer at Elron Electronic
Industries. Moni received an MBA from Kellog Business School's Recenati Program
(Tel-Aviv University), and a Masters in Applied Physics from Stanford University.
Dr. Joel Ratsaby, Group Manager, AI
Prior to joining Manna, Joel was an independent consultant/developer on projects in
Data Mining, Pattern Recognition, and Learning Algorithms. He received his PhD in
1994 from the Moore School of Electrical Engineering, University of Pennsylvania, in
machine learning and artificial intelligence. From 1995 to1997, Joel did Post
Doctorate work at the Electrical Engineering Department of the Technion in the area
of advanced machine learning algorithms. He worked at AT&T Bell Laboratories on
the design and development of high speed communication systems. He has published
several papers in the area of Computational Learning Theory and Machine
Intelligence. Joel received his BSEE in Computer Engineering and his MSEE from
Brooklyn Polytechnic Institute, New York.
Gary McMeekin, VP Professional Services
Gary joined Manna from Marcam Solutions, a Boston-based developer of Enterprise
Resource Planning software. At Marcam since 1990, Gary held several management
positions, most recently Software Development Manager. He has extensive project
management and customer interaction experience, and has led large teams in the
implementation of ERP-level software for major clients. Gary has implemented
solutions throughout the US Europe, and played a major role in building Marcam UK
from a start-up into a major regional division of the company.
Mike Murphy, VP Business Development
Mike joined Manna after serving as Vice President of Sales for SDL INTL a multi
lingual content management and localization company based in the UK. Before
working at SDL INTL, he served as Vice President of Portal Sales for OrderTrust, an

Internet e-Commerce order management networking company. Prior to that he
worked at Open Group, a standards company that developed custom technology for
its investors and provided consulting on a variety of technologies, as the Vice
President of Sales/North America. Before working at Open Group, Murphy spent 3
years at IBM as the Eastern Area sales Director for their services division. He came to
IBM from a 13 year tenure at Digital Equipment Corporation in various sales, sales
management and marketing management roles including; District Sales Manager,
Group Marketing Manager for the Applications Industry Group, Group Channels
Manager, Business Development Manager for Digital's Multi-Vendor Services Group
and Vice President of Sales for a Digital owned software subsidiary company, 800
Software. Murphy received his MBA from Northeastern and attended a 3-month
business management program at Harvard.
Veronica O'Shea, Vice President of Sales
Veronica, a 17-year veteran of sales in various software vendor environments, has
consistently demonstrated her sales leadership in building and scaling organizations,
implementing creative sales strategies and methodologies and managing explosive
revenue growth. Veronica joined Manna from EGain Communications, a provider of
customer service software for the Internet, where she served as the vice president of
sales. At EGain, O'Shea was responsible for a 200+ person sales and marketing
organization, growing revenues over 90 percent during her last quarter. Previously,
O'Shea held various sales management positions for Vantive Corporation (acquired
by PeopleSoft), IQ Software Corporation, Oracle Corporation and Dun & Bradstreet
Software - all providers of customer-focused solutions. Veronica holds a BA from
Boston College.

Manna – the board
Tal Barnoach, Chairman
Tal is an angel investor and consultant to a number of Israeli start-ups in the areas of
enterprise software and the Internet, and holds several board positions. Previously, he
was Chairman & CEO of SEA Multimedia, a publisher of sport content for CD-ROM
and the Internet. Tal managed SEA for five years and took the company public on the
AIM in London in 1996. Before establishing SEA, he helped found and manage CDI,
a major Israeli manufacturer of CDs, which is traded on the Tel-Aviv Stock
Exchange. Tal held numerous positions at CDI over a five-year period, finally
becoming Deputy Managing Director of the company. He is a founding member of
the Israeli Multimedia Forum, as well as a member of the Ministry of Industry and
Trade's Council for the Development of Start-Ups in Israel. Tal has a BA in
Economics and Management and studied for his MBA at Tel-Aviv University.
Pat Keneally, Director
Pat is Managing General Partner at IDG Ventures, the $180 million U.S. venture
capital fund of International Data Group, the $2.5 billion global media conglomerate.
Pat supervised IDG Ventures' successfully liquidated investments in Andromedia
(NASDAQ: MACR), BabyCenter (NASDAQ: ETYS), FutureTense (NASDAQ:
OMKT), Service Metrics (NASDAQ: EXDS), and (NASDAQ: AOL),
which collectively returned IDG Ventures 30 times its invested capital in those
companies. He currently represents IDG Ventures on the boards of Career Central,
Manna, Next Planet Over, Online Partners,, Quova, and

Pat founded IDG Ventures in 1996 after a dozen years as profit center manager in
IDG's publishing group. From 1990 to 1996, he was Publisher of IDG's PC World
Magazine and CEO of PC World Communications. Inc. During his tenure PC World
quadrupled revenues and profits and, launched numerous new publications and
products including Multimedia World, The WEB magazine, PC World Online, and
joint ventures with Forbes, Newsweek, Child Magazine, and others. During his last
four years at PC World, he ran IDG's PC World Global Support Center, which
supported locally-published PC Worlds in more than 50 countries.
Before joining PC World, Pat was founder, President, and Publisher of Digital News,
IDG's Boston-based newspaper for the VAX computer market. Before joining IDG,
he was founder, Associate Publisher, and Editor-in-Chief of Digital Review Magazine
at Ziff-Davis Publishing, and Senior Editor of Mini-Microsystems at Cahners
Publishing. Pat holds a bachelor's degree from Harvard University.
Dan Ross, Board Member
Please see above bio under Management Team.
John Losier, Board Member
John Losier, former President and CEO of Philips Electronics North America, brings
over 20 years of experience in information technology to Manna in addition to an
impressive understanding and breadth of knowledge in distribution strategy and
channel development, integration of acquisitions, and strategic alliance development.
Prior to Philips Electronics, John was Vice President of global accounts at Compaq
and Senior Vice President of worldwide sales, marketing, services and support at
Tandem. Prior to Tandem, John's experience spans many of the largest, most
successful communications and information technology companies including Bell
Atlantic, Northern Telecom, AT&T Information Systems, and IBM. Johns also sits on
the Boards of MarketSoft Corporation and Prism Venture Partners. John holds a
Bachelor of Arts degree in Economics from LeMoyne College in Syracuse, NY
Tali Aben, Board Member
As Vice President, Tali joined Gemini Capital Fund from RadView, where she served
as Marketing Director. Prior to joining RadView, Tali was a Sr. Product Marketing
Manager at SunSoft, Inc. From 1990 to 1992, she served as Marketing Director for
Mercury Interactive Corporation, where she established the company's U.S. Sales and
Marketing organization. From 1984 to 1990, Tali held Software Engineering positions
at Scitex Corporation and Aitech Systems. Tali holds a B.Sc. in Mathematics and
Computer Science and an MBA, both from Tel Aviv University.
Marcia J. Hooper, Observer Seat
Marcia received a BS in Chemistry, with honors, from Brown University, an MA in
the same field from Columbia University and an MBA from Harvard Business
School. She began her career as an engineer and marketing representative with IBM.
In 1985, she joined PaineWebber/Ampersand Ventures, where she focused on early-
stage technology investments. In 1993, she left to co-found Viking Capital, where she
managed a dedicated venture fund for R.R. Donnelley & Sons. Marcia joined Advent
in 1996 and focuses on information technology investments. She is a Director of
PolyMedica, Wang Global, LioNBRIDGE Technologies, Captivate, WorldGate,
SegWay Internet Technologies and Interleaf.

Eugene M. Weber, Observer Seat
Eugene M.(Gene) Weber is managing partner of Weber Capital Management, LLC in
San Francisco, the predecessor of which he founded in 1995. Weber Capital is a
crossover investor in both later stage ventures and micro-cap public companies. Since
inception, the firm has managed investments of over $120 million, and currently has
positions in 34 companies, 14 of which are private. Weber Capital focuses on
information technology, communications and healthcare, and its successful
investments include Commerce One, Intertrust, Interwoven, Broadbase Software,
Galileo Technology, Fusion Medical, Optical Coating Laboratory, Polycom, and
Veritas Software. Mr. Weber is on the Boards of NYSE-listed Chyron Corporation
and Amex-listed Computron Software.
Mr. Weber graduated from Cornell University in 1972 with a BS in Electrical
Engineering and worked for 3 years as a design engineer and in product marketing.
He received his MBA in finance from Wharton in 1977, and joined McKinsey &
Company in Los Angeles. He was a founding member of McKinsey's worldwide
electronics industry practice group. In 1983 Mr. Weber joined the investment
management firm of Weiss, Peck and Greer in San Francisco, and became a partner in
its venture capital group in 1984, and a partner in the parent firm in 1987. During his
11 years with Weiss, Peck and Greer, he made venture investments in 16 companies
and was responsible fro 11 other investments.
Oren Zeev, Observer Seat

Oren Zeev received an MBA with distinction from INSEAD and a B.Sc. cum laude in
Electrical Engineering from the Technion Institute of Technology. Upon graduating
from the Technion, Oren joined the IBM Science and Technology Center as a
Research Fellow in the Multimedia Group. In 1991, Oren was appointed a Research
Staff Member and joined IBM's newly-founded Semiconductor Design Center, where
he managed several significant R&D projects. In 1995, Oren joined Apax, where he
covers investments in the area of Internet/IT. Oren represents Apax on the Boards of
eMation,, GoRefer,, Inspectech, Manna (observer),
and TdSoft Communications.


Shared By: