Muninn Technologies ®, LLC
Real-time Interaction Management
August 13, 1999
INTRODUCTION E-commerce is rapidly changing the face of today's marketplace as more and more consumers move away from traditional "brick and mortar" sales channels towards sales channels that primarily rely on the Internet. According to the International Data Corporation (IDC), Ecommerce is growing at a rate of 205% per year and a recent study by the University of Texas estimates that the "Internet Economy" generated $300 billion in revenues in 1998. These explosive trends are expected to intensify, as Forrester Research predicts that the total ecommerce market, including both the business-to-business and the business-to-consumer segments, will exceed $1.3 trillion by 2003. Fueled by such grand expectations, E-commerce sites have proliferated, serving up a vast array of goods from children's toys to routers, from chintz curtains to complex financial products. As more mainstream customers replace the early movers (high-tech enthusiasts), greater emphasis needs to be placed on meeting customer expectations and increasing customer satisfaction. This is especially important in the new economy where competing shopping sites are just a few clicks away. In order to meet such expectations, companies must place a greater focus on building strong customer relationships and developing customer loyalty. They must manage customer interaction and ensure customer satisfaction throughout the browsing and purchasing process. These thoughts related to customer satisfaction are not new. An entire industry - Customer Relation Management (CRM) - is devoted to managing customer relationships. CRM amalgamates all the information a vendor knows about a customer to provide better service and build strong relationships and loyalty. CRM attempts to automate much of the "customerlifecycle" including generating targeted campaigns and focus segments to providing customer information to personnel in call centers that are providing after-sales service and support. Using CRM systems, personnel interacting with customers have information regarding the entire customer-life cycle and thus are more likely to provide the support and information the customer needs. More efficient customer support greatly increases customer satisfaction and retention. There are numerous companies within the CRM space, including specialized CRM companies such as Siebel Systems, Epiphany, and Trilogy and larger ERP companies such as SAP and IBM. A key fact to note regarding CRM is that these tools support the management of customer relationships in a wide variety of markets, not just in E-commerce. In fact, CRM was first developed to provide customer service support within call centers, a market that still remains important, even in the Internet/E-commerce space. The problem, however, is that CRM by itself is not sufficient to manage customer relationships in the E-commerce space. The one critical missing piece is the ability to interact with an ecustomer, in real-time, and impact her shopping experience with an intent to maximize the likelihood of an eventual purchase in the current shopping visit. This issue, i.e., increasing the likelihood of an eventual buy in any given visit, is of significant current interest. There is emerging concrete evidence in the online marketing literature, that "buy probabilities" are strongly correlated to the level of personalization that a site offers - this is referred to variously as "stickiness", "click maximization" and other such terms. The key behind such a personalization effort is the ability to accumulate vast amounts of information, and knowledge, about customers, and the ability to serve this up in real-time. CRM has primarily been used by
call centers to pull up individual customer records or to build targeted marketing campaigns. In the first function, complex querying of the data is not required; in the second function, the complex querying is done off-line where the total time of querying is not as critical. In short, CRM systems cannot handle the complex, real-time querying that is required for E-commerce. What are the new requirements of E-commerce? Success in E-commerce is predicated on one critical element - the "stickiness" of the web site. The "stickier" a web site is, the more attractive it is to online consumers and advertisers, the longer customers stay at the site, and the more likely they are to buy products and services. Real-time Interaction Management (RIM) is a new category of software that enables organizations, both in the B2B and B2C spaces, to build and deploy E-commerce solutions that are "ultra-sticky" and that provide the user with a hyper-personalized shopping experience. The core technology behind RIM, enables real-time querying of vast amounts of historical data which allows E-businesses to anticipate consumer needs and to provide content that is tailored for the consumer. RIM increases the utility of the web site and the displayed content to the consumer, reduces search frustration, builds brand loyalty, and increases probability of purchases - all directly leading to an increase in sale revenue and advertisement revenue generation. WHAT IS Real-time Interaction Management (RIM)? Companies collect massive amounts of data of various types -- transactional, navigational, and data provided by third parties. There also exist technologies that can delve into this data and generate knowledge in the form of correlations, associations and clusters. RIM uses fast, realtime technology to create a hyper-personalized experience for each individual customer by bvcdr tapping into this vast information and knowledge stores. In other words, RIM provides true oneon-one marketing for each site visitor. To illustrate the possibilities of RIM, consider two examples from the real world: Business to Business: Dell Computers Dell Computers has implemented a system that generates customized web pages for business customers. These "Premiere Pages" allow customers to configure and order products, check the status of orders, and access help desk information - typical CRM functions. While this system is beneficial to both Dell and to its customers, a greatly more responsive system is possible using RIM. For example, consider a customer who, several months earlier, purchased several computers via his "Premiere Pages". This customer returns to the Dell site to purchase additional memory modules to upgrade the systems. However, when he places his order, he chooses modules that are incompatible with his computers. RIM software, detects the incompatibility with his previous purchase, alerts the customer and suggests the appropriate modules, thus avoiding the inconvenience and expense of returning inappropriate hardware. Clearly, the ability to interact with the customer in this manner requires the ability to delve into the vast customer records, correlate the current purchase attempt to previous purchases and to
react dynamically. Business to Consumer: Amazon.corn Amazon has taken a customer-centric approach to book-selling by implementing a customer preference and purchase tracking system used to generate a list of recommendations for a returning customer. Although this system has proved worthwhile for Amazon (58% of orders come from repeat customers), it is possible to dramatically improve upon this system using RIM. Since Amazon generates recommendation lists off-line, the recommendations presented to the customer do not reflect navigational data generated during the currentsession. Nor can Amazon detect, based on online navigation patterns, whether a customer is about to churn (i.e., leave without buying). RIM can incorporate such data in real-time, reflecting up-to-the-moment data about a specific customer. Imagine a college student who regularly orders both textbooks and pleasure reading from Amazon. Perhaps she has just graduated, and logs in to the Amazon web site looking for light summer reading. Having perused several categories, such as historical romance and science fiction, she excitedly checks the recommendation list and is disappointed to find textbooks from her degree in Quantum Physics. Clearly, real-time analysis of both buying history and navigational data would improve the recommendation lists generated, and further improve customer loyalty in the form of repeat customers. Also, current CRM systems do not help Amazon push a buy decision for any given customer. For instance, if it can be determined that a customer is highly likely to leave at the current frame or a subsequent frame, it is possible to take drastic steps to reduce this likelihood, such as providing a significant discount on something likely to interest the user. Such dynamic pricing models cannot be supported using current technologies. The real-time analysis described in the examples above is difficult to achieve using current storage, retrieval, and analysis technologies. The sheer volume of data storage required, as well as the time needed to perform such analyses, makes online, real-time, interactive profiling impossible with existing technologies. The technology underlying RIM is a high performance storage and retrieval engine that provides extremely fast querying and data access, while simultaneously keeping storage and system maintenance costs extremely low, when compared to existing storage and retrieval mechanisms. Some highlights of the benefits of RIM software include: detailed profiling and efficient access of user profiles >' intelligent catalog management targeted advertisements and promotions >' targeted content delivery
high-performance ad-hoc reporting chum modeling and analysis How does RIM differ from CRM? Although CRM and RIM seem to be quite similar in that they both seek to improve customer satisfaction and loyalty, there are significant differences between CRM and RIM. It is important to highlight some of the key differences between these technologies: Table 1: RIM vs. CRM R !M Revenue generation Online Individual-centric Developed for the Ecommerce Dynamic content CRM Campaign generation Offiine Group-centric Developed for the call center Static content
•
Revenue generation vs. Campaign generation: RIM is a "click generator" -- it is specifically designed to increase the "stickiness" of a web site; that is, to increase the buy probability of a given visitor, thereby increasing revenue. RIM achieves this by analyzing the customer profile and presenting a highly personalized experience to the consumer. In contrast, CRM provides analysis to support the generation of marketing campaigns based on, essentially, the same data RIM uses to generate individualized customer interactions. Online vs. Offline data analysis: RIM provides real-time data analysis of an individual customer's profile using the most up-to-date information available, while analysis in CRM takes place off-line. lndividual-centric vs. Group-centric: Based on real-time analysis, RIM is able to achieve one-on-one marketing for individual customers, while CRM, using off-line analysis, categorizes customers into groups for marketing purposes. Internet vs. Call Center: RIM software is designed specifically for E-commerce, while CRM was originally designed for customer support call centers. Dynamic vs. Static content: Using real-time online analysis, RIM is able to dynamically update web site content for a specific user as that user's profile changes. In other words, content is generated while a customer is currently active at a site, allowing for the system to respond to specific customer actions, such as choosing a different navigational path than the one chosen in the last visit. CRM, on the other hand, produces static content for each marketing group, rather than for individual customers, resulting in "canned"
•
•
•
•
navigational experiences. Another important impact of RIM systems is the potential to maximize advertisement revenues, an important revenue component of online commerce sites. Essentially, RIM systems will allow much more effective advertising by analyzing and reacting in real-time to customer profiles and navigation patterns in order to target advertising to each user's individual interests. This in turn will be of immense benefit to the advertisers, as they will be exposed to the users who are most likely to be interested in their products and services. This feature will allow RIM-enabled Ecommerce companies to charge a premium for advertising on their sites. One fact to note is that although RIM and CRM differ in several significant ways, they are actually complementary technologies. The joint effect of the deployment of both CRM and RIM technologies yields a powerful synergistic combination: Total Customer Experience Management (TCEM).
••_j._:: :_i_Real.time _: I.teractlo_Managpmen_:..:" ,.,:_._:::., _,, ...... :.:.!:_i: . !_.. ':_ ....... ":..??_ .......i?" _:_:!_-'._ J .. _:_Customer Relattonslilp ManagemeNt i .....To_i_C u_tomer Ex_.pe_rien_e;: _.,_ M_age'_eht :;S ._.i .;:._ ........ :_.. .. :_!_:. . .;. __ :i':_. I ;:. ,;_i:i;> .... :'1 i: ._.....
RIM TECHNOLOGY AND ARCHITECTURE RIM Technology Why have other vendors not done RIM? Until now, the technology required to enable RIM has not existed. The critical functionality required in RIM systems is the ability to store huge volumes of detailed information about customers and the ability to rapidly access this vast warehouse and retrieve key pieces of information in real-time. No current commercial technology can perform such tasks in real-time on such large data warehouses. To give an example of how large these data warehouses can become consider the following: A site servicing a million customers a day, with 300 bytes of navigational information per customer, will produce 10GB of raw data in a month. Assuming a modest yearly growth of 20% and a seven-year storage window, the data warehouse will balloon to over 1.3 terabytes for the raw data alone. With meta information, such as indexes and preaggregated information, the data warehouse will balloon to 3-10 times the size of the raw data. The above example just refers to navigational information. When customer profiles, third-party information and stored knowledge are added, the data store expands to a multi-terabyte warehouse, requiring millions of dollars per year in storage cost alone (usually one dollar spent in purchasing storage results in about $7/year in maintaining this storage).
No existing commercial technology can access and query this type of data to support real-time ad-hoc retrievals. Real-world applications reduce the amount of data collected, perform querying off-line, and generally provide non real-time interactivity with the data warehouse. In response to the limitations of current technologies and the needs of E-commerce, Muninn Technologies has developed an ultra high performance storage and query engine that enables RIM. This engine, code named FastPump, is an open, standard-based, traditional data warehouse highly pre-optimized for RIM. FastPump is designed to house vast volumes of customer, navigation and third party information, in addition to generated knowledge, to guide the dynamic generation of individualized content. Current data storage and retrieval technology would fail to provide this functionality for two primary reasons: (a) storage of vast amounts of data would be prohibitively expensive (in dollar terms),and more importantly, (b) retrievals for such huge volumes of information would be impossible to achieve in real-time. FastPump, based on proprietary technology known as ILR, allows for very compact (no expansion over raw data size) storage, while simultaneously achieving blazingly fast retrieval to support RIM functionality. An important fact needs to be noted at this point. Business intelligence applications such as data mining software and report generators work only as fast as the underlying data pump that feeds them. Often it is the case that the performance of the base functionality of these tools, most of which occurs in memory, is quite efficient. The bottleneck is the acquisition of relevant input data from underlying data warehouses. The CDW in Muninn's RIM solution, based on ILR technology, provides an unprecedented co-branding advantage to these tools by serving as an ultra-efficient data pump. RIM Architecture The RIM architecture is highly modular as seen in Figure 1 below.
Customer Profiles,Navigation Records, etc. {
Web interface
I
i
•.
Customer Detail Warehouser.
._
_..
"_
Figure 1: RIM Architecture The Customer Detail Warehouse (CDW) is the foundation of the RIM solution and is comprised of a query engine and a data store. The CDW feeds a Data Mining Module (DMM), which in turn generates knowledge both on the fly (used by the dynamic content generator), and offline (for such purposes as churn modeling, which does not require real-time responsiveness). A particular strength of Muninn's solution is that generated knowledge can be stored, and thereby, reused and modified. This dramatically reduces the cost of generating knowledge. The Dynamic Content Generator (DCG), using technologies such as Java, generates personalized web pages on the fly, based on content and customer data in the CDW. The Reporting Engine (RE) responds to ad-hoc queries for strategic planning and high-level decision support. The result of all these components working in concert, anchored by FastPump, is a hyperpersonalized web experience for the customer. This experience in turn is likely to result in greater buy decisions than current levels, resulting in increased revenues. Also, because of the highly targeted content delivery, e-commerce service providers can command greatly higher advertisement revenues by virtue of the ability to provide advertisers more "bang for their buck".