Personalization for Web Applications Jon Jenkins jjenkins netperceptions com Net

Personalization for Web Applications Jon Jenkins jjenkins@netperceptions.com Net Perceptions, Inc. Agenda Evolution of personalization technologies The personalization value proposition Implementing personalization for web applications The future of personalization Evolution of Personalization: Personalization has been used for decades – Direct mail – Telemarketing It was usually implemented offline in batch mode Classic personalization was based on demographics – Segmentation – Profiling – Categorization Old personalization used static, predefined categories Old personalization required prior knowledge about customer Common factors to consider in this type of system include – Gender – Zip Codes – Income Levels Traditional Personalization Next Generation Personalization Focuses on individuals not segments: one-to-one Learns in real-time based on behavior Dynamically infers user behavior based on others Weights various behavioral attributes Gets smarter with every use Works well with extremely limited information Makes accurate predictions on the first visit Recognizes that personal affinities overlap demographic categories Paradigm Shift between Traditional & Next Generation Personalization Old personalization focuses on "likelihood to sell" – Mainstream demand – The "average" customer – External categories New personalization focuses on "likelihood to buy" – "Average" customer doesn't exist – Behavior can only be predicted on individual level – Individual behavior is best predicted by other individuals Value Proposition from Customer Perspective Save time by finding relevant products/content first Learn about new products effortlessly Personal attention and recognition is comforting Personalization Value Proposition: Value Proposition from Business Perspective Differentiation from competitors Better customer retention Turns browsers into buyers (conversion) Increased incremental revenue per customer Fosters interdepartmental cooperation Improves efficiency Improves productivity Inventory control Product management and acquisition Employee satisfaction Advent of the Personalization Centric Enterprise Gather information at all inbound touch points Provide recommendations at all outbound touch points – Web site – Call center – Direct mail – Outbound email – Telemarketing – Customer service – Statement stuffers – Shopping lists Personalization as a corporate philosophy Pervasive Personalization Walk – Kiosks – Checkout lines (coupons) Talk – Call centers – Outbound telemarketing Click – Personalization via the web Beep – PDAs – Cell phones Implementing Personalization: Four Personalization Scenarios Scenario 1 – Simple name recognition Scenario 2 – Checkbox/Preferences based personalization Scenario 3 – Segmentation and rules-based personalization Scenario 4 – One-to-One (Collaborative Filtering) Simple Name Recognition Technique – Simple database query Advantages – – – – This is the simplest implementation of personalization Placing the user's name on the screen can build trust e-tailers such as Amazon use this technique No special technology is required for this functionality Caveats – Don't make you application "creepy" with too much of this type of personalization – Be careful not to compromise your users' privacy Code Sample Scenario 1 Code sample in any language – Like Amazon Demo Checkbox-Based Personalization Technique – Provide the user with application configuration options (checkboxes) – Build the application to use these preferences Advantages – The perfect way to personalize for specific situations – Stock quotes – Weather – News – Provides rapid access to customized information – Easily implemented using a few database queries Caveats – Too many options can be overwhelming – Options that don't reflect the proper range of choices can be confusing – Be careful to respect privacy Code Sample Scenario 2 Code sample in any language – Like Bloomberg Demo Segmentation Based Personalization Technique – Retrieve information from enterprise data stores – Categorize users based on the data – Build the application to use this information Advantages – Companies already have the data necessary to implement this type of personalization – Provides rapid access to information – No special technology is required (but it helps) Caveats – Segments must be extremely carefully defined & redefined – Be extra careful not to compromise your users' privacy in this case you are using information they have not explicitly provided for the purpose of personalization Code Sample Scenario 3 Code sample in any language – Code for dynamically building segments – Code for personalizing application Demo One-to-One Personalization Technique – Collect behavioral data (purchases, views, clicks) – Implement Collaborative Filtering (CF) solution – Build the application to use this solution Vendors of collaborative filtering solutions – Macromedia (LikeMinds) – Microsoft (Site Server) – Net Perceptions – Personify An Inside Look at Collaborative Filtering Information about customer is gathered A neighborhood of similar users is built The neighborhood is analyzed Recommendations based on the neighborhood analysis are presented One-to-One Personalization Advantages – CF solutions learn on the fly – CF solutions work well with sparse or inadequate data – CF solutions require little maintenance Caveats – Requires special software (and probably more hardware) – Application design requires more skill – Data flow is crucial Code Sample Scenario 4 Code sample in any language – Overview of necessary data structures – Code for personalizing application Demo Non-Traditional Apps of Existing Technology Find users most likely to buy a specific product Predict purchase orders required to satisfy demand Decision support for product acquisition Future of Personalization: Features of Upcoming Technology Microsecond response times Working with very large data sets Adaptive pricing Other stuff that has to be approved before I can talk about it Questions? Jon Jenkins jjenkins@netperceptions.com Net Perceptions, Inc.