This presentation describes a framework for describing and categorising personalisation as experienced within eCommerce. It explores its application to a number of examples, and discusses the implications (e.g. instances that could in theory exist but don’t).
2. Contents Definitions Personalisation vs. customization Dimensions of Personalisation Profiling Data Profiling Method Target Scope Persistence Personalization Patterns Examples 2 UXLabs - User Experience Research and Design - www.uxlabs.co.uk
3. Definitions Personalisation vs. customisation System-led vs. user-led Levels of personalisation What is minimum? Entry of postcode Recently viewed / purchased (e.g. Amazon, Asda) Simple on-page display settings (e.g. Argos, Asda) Ability to provide an individual user experience Contrast with static site experience 3 UXLabs - User Experience Research and Design - www.uxlabs.co.uk
4. Dimensions of Personalisation Profiling data what data is acquired? From whom is it acquired? Profiling method how is the profile data acquired & applied? Target where are effects of personalisation experienced? Scope what is the extent of the personalisation experience? Persistence what is the focus of the personalisation approach? 4 UXLabs - User Experience Research and Design - www.uxlabs.co.uk
5. Profiling data (what data is acquired? From whom?) User-provided demographic data, interests, location, etc. e.g. iGoogle, BBC Behavioural buying / viewing history e.g. Amazon, Virgin Individual Your personal buying / viewing history Collective Aggregate viewing / buying patterns 5 UXLabs - User Experience Research and Design - www.uxlabs.co.uk
6. Profiling method (how is the profile acquired & applied?) Explicit Preferences + interests (e.g. BBC, Monster, B&Q) Implicit Role-based Segment (e.g. new vs. repeat visitor) Behavioural (e.g. Amazon, Virgin Wines) Content-based Suggestions based on product similarity e.g. Amazon User-based Suggestions based on user similarity e.g. Last.fm 6 UXLabs - User Experience Research and Design - www.uxlabs.co.uk
7. Target (where are the effects experienced?) User Interface Layout of tools & widgets, theme, colour scheme Content Widget configuration, content Display defaults Implicit pre-configuration of interface + content, e.g. greater support / richer content for certain types of user re-ranking of search results Merchandising Recommendations Related items: cross-sell, up-sell, etc. 7 UXLabs - User Experience Research and Design - www.uxlabs.co.uk
8. Scope (what is the extent of the experience?) Site-wide settings (apply across whole site experience), e.g. BBC e.g. location Page-specific display options, e.g. Arrow, Farnell, Argos e.g. basic/advanced filtering, column pickers, list/gallery view, etc. No independent user model 8 UXLabs - User Experience Research and Design - www.uxlabs.co.uk
9. Persistence (what is the focus of the approach?) Short-term temporary interests Long-term stable interests Default is long-term, across sessions 9 UXLabs - User Experience Research and Design - www.uxlabs.co.uk
12. Personalisation Examples User-driven customisation Branded Content Portals Personal Web Portals Collective recommendations Most ‘popular’ (purchased, viewed, emailed, etc.) Search results re-ranking e.g. Google Collaborative filtering e.g. Last.fm (Music) http://www.last.fm User-created alerts e.g. Google / Yahoo Alerts Implicit + explicit personalisation e.g. Amazon Deep linking e.g. Google search, email promotion 12 UXLabs - User Experience Research and Design - www.uxlabs.co.uk
13. User-driven Customisation Branded Content Portals Drag & drop arrangement of widgets / content panes Examples: BBC (News) http://www.bbc.co.uk Tutsplus (Education) http://tutsplus.com/ 13 UXLabs - User Experience Research and Design - www.uxlabs.co.uk
14. UXLabs - User Experience Research and Design - www.uxlabs.co.uk Personal Web Portals Layout & content of tools & widgets Theme / colour scheme, widget features, configuration, content 14
15. Collective Recommendations Other items bought by purchasers of the target item At the same time Over a longer period 15 UXLabs - User Experience Research and Design - www.uxlabs.co.uk
16. Search Results Re-ranking User can explicitly promote or remove results Changes preserved for same query Could be used to re-weight related queries 16 UXLabs - User Experience Research and Design - www.uxlabs.co.uk
17. UXLabs - User Experience Research and Design - www.uxlabs.co.uk Collaborative Filtering Suggestions based on user similarity Best suited to content that is taste-oriented Films, music, etc. 17
18. User-created Alerts User builds profile for topic of interest e.g. set of terms for monitoring news & web sites e.g. Yahoo, Google, Amazon 18 UXLabs - User Experience Research and Design - www.uxlabs.co.uk
19. Implicit + Explicit Personalisation Amazon Implicit Buying / viewing history Buying history is strong endorsement Explicit Improve recommendations Turn off browsing history Opt out of 3rd party personalized ads 19 UXLabs - User Experience Research and Design - www.uxlabs.co.uk
20. Personalisation through Deep Linking Implicit data Referrer (Google search) Unique ID (promotional email) Bypass internal search Pre-qualified buyers UX depends on priorities e.g. rapid transaction, stickiness, return visit, other? Challenge is to indicate breadth of content, branding, reliability, service, etc. 20 UXLabs - User Experience Research and Design - www.uxlabs.co.uk
21. Conclusions Alignment with overall business/marketing strategy What are the priorities? e.g. customer acquisition vs. customer retention long-time loyalists vs. newer customers frequent shoppers vs. biggest spenders individual vs. segmented Personalisation strategy should fitwithin overall strategy Different conceptual model Navigational model: Where am I? What is here? Where can I go next? Personalisation model: Who do you think I am? (profile data) Why is this here? (rationale / business rules) What am I missing?(default experience) UI needs to answer these questions 21 UXLabs - User Experience Research and Design - www.uxlabs.co.uk
22. Conclusions Highly differentiated journeys risk alienating misclassified users Making accurate predictions is difficult Needs and goals change over time Can lead to inconsistent UX, missed opportunities Confidence in user segmentation is crucial What proportion of site users are logged in? What does their registration data tell us? How accurate is it? Defensive strategy is to apply suitable defaults e.g. highly visible support vs. hidden (but accessible) support Results page: images, columns Line level: image, Technical reference (+ data sheet), Attributes, Overview, Range Overview, etc. Additional UI controls required Excel-style hide columns / button Accordion controls, etc. 22 UXLabs - User Experience Research and Design - www.uxlabs.co.uk
23. Conclusions Balance needed: What the merchandiser wants to sell vs. what the user wants to buy Margin vs. relevance Personalization should not constrain information access User must always be able to exit the personalized experience Challenges in implicit personalisation Offline channel interactions Each purchase degrades the training set Crude product relationship modelling Popular items tell us little Directed purchasing behaviour Recommendations may be of limited value to a buyer with no purchasing discretion Challenges in explicit personalisation Many users just accept the default Default design must be appropriate + scalable Privacy concerns Too much explicit user involvement can be counter-productive 23 UXLabs - User Experience Research and Design - www.uxlabs.co.uk