New Face of Persona:Behavioral Analytics for Design    Sanzhar Kettebekov       October, 7 2011
User Experience for a Media Site   Monetizing vs. Usability• A fine line between   –   Lots of ads – no ads   –   “Pop-up"...
Elements of UX Design Process1       requirements       2    personas         3   user needs    4     scenarios    5      ...
Key Requirements for Personas• Adequately represents a user population• Captures functional and critical needs• Serves as ...
Problems with Classical PersonasHard to come up for an unstructured audience(vs. enterprise roles)• Lacks quantitative bas...
Example: Demographics-Based Segments                                       6
Why Not Demographics Segmentation?• Less relevant for design   – Does not represent what users actually do online   – Appr...
Current Media Pain Points• Which audience segments can be monetized and  how to find them• How to sell ad space for a prem...
Why Behavioral Analytics?• Directly relates to usage and monetization metrics   – Enables measurement and tracking of bran...
50 000                                                                                                          100 000   ...
Site Redesign Example                        11
Research Methodology                                                            marketing personas                        ...
Towards Integrated Product Cycle                                  Business                                Intelligence    ...
Personas Research: Intercept SurveyBootstrap personas model from consumption behaviorperspective• Goals:   –   Capture ent...
Results: Poor Content FitQ3: What are you looking for today? (yes/no answers)Q8: What are you interested in? (4-point scal...
Deriving Taxonomy                    16
Personas Research: Focus GroupsDefine generalizable and extensible online consumptionbehavioral models applicable for othe...
Personas Research: Focus Groups Methods•   8 focus groups of 8 people each.     – 2 focus groups in each location.•   2x4 ...
Focus Groups: Consumption Behaviors •   Three primary behaviors. Individuals may exhibit all three behaviors.        –    ...
Actionable Behavioral Analytics                          web analytics                              data           Behavio...
Fundamental Behaviors (grind-scan                                                     TM    )Engagement &                 ...
Behavior ShiftsUVs, Homepage and News Sections800,000                                                          6.3 %      ...
Local Vs. National Audience                                                              100%                             ...
Use of Navigation and Hot Topics                               Use of Navigation                  Use of                  ...
Return Behavior              Return Visit Behavior           Inducements to ReturnGrind         Regular. May be daily,   G...
Return Behavior                        Return Visit Behavior                              Inducements to ReturnScan      A...
Traffic Sources ( a local media example)                               In-Market       In-Market (ext)       Out-of-Market...
Channel Reliance (a local media example)                                             Average                              ...
Engagement Analysis (a local media example)                          Pages per Visit Day   Engagement        Grind        ...
Content Fit Analysis (a local media example)                       SITE                      YOUR                         ...
Behavior-Centric IA 1   Easy access to past dates               1           2     designed for: Trackers, Grinders        ...
Behavior and Interests (a national media example )                                           Grinders Scanners   World & N...
Behavior-Centric IA: Article Page                                    • Different for                                      ...
Behavior-Centric IA: TaxonomyLocal News   U.S. & World News   Entertainment   Sports   Business & Technology   Travel   He...
Extending Personas                demographics                grinder>55, affluent             55-40, affluent            ...
Audience Segments Revised                                         Primary: Grinders                                       ...
Audience Strategy (example)                 Homepage                        Section Fronts            Story-level Page    ...
Audience Risks                   Grinders                Trackers                Scanners                Subject Devotees ...
Results                          new IA          2008     2009        2010• Nearly doubled UVs• Nearly doubled PPVs       ...
Limitations• Need existing site to generate data   – A developed audience should exist• Mechanics of behavioral segmentati...
Where Does It Fit?• This in no way means you don’t need other UX  research methods   – In fact, it is a complimentary meth...
Behavior Analytics Today (Ad Industry)                     current  click history                     interest            ...
Next Level of Behavioral Analytics    measurable behavior               computed intent                 behavioral targeti...
Display Ad Perception/ImpactEngagement &  Source Reliance  Open to consume                            Track suggested cont...
Audience Engineering• Powerful behavior-centric analytics   – Maximum online audience understanding and     tracking of mo...
Вопросы?Sanzhar Kettebekov, Ph. D.sanzhar@segmentinteractive.com
Social Networks Behavior (Dwell-Scan™)                                                                         • Extravert...
Online Video Consumption (Dwell-Seek™)Consumption                                           • DVR behavior                ...
Ad Strategy        AD                 High ad tolerance              Moderate ad tolerance                 Low ad toleranc...
Upcoming SlideShare
Loading in …5
×

UX Russia 2011 - Sanzhar Kettebekov, Segment Interactive, New Face of Persona: Behavioural Analytics for Design

1,177 views

Published on

Published in: Technology, Business

UX Russia 2011 - Sanzhar Kettebekov, Segment Interactive, New Face of Persona: Behavioural Analytics for Design

  1. 1. New Face of Persona:Behavioral Analytics for Design Sanzhar Kettebekov October, 7 2011
  2. 2. User Experience for a Media Site Monetizing vs. Usability• A fine line between – Lots of ads – no ads – “Pop-up" ads – no flash ads – Registration wall – organic traffic – Catering to all audiences – well defined target audience – Internal community– Social Network connect strategy – One navigation – personalized architecture 2
  3. 3. Elements of UX Design Process1 requirements 2 personas 3 user needs 4 scenarios 5 wire frames 6 3
  4. 4. Key Requirements for Personas• Adequately represents a user population• Captures functional and critical needs• Serves as a communication tool and integrates into the design cycle 4
  5. 5. Problems with Classical PersonasHard to come up for an unstructured audience(vs. enterprise roles)• Lacks quantitative basis• Lacks users psychographical and behavioral representation• Conveys approximate user needs – No easy process to capture needs – Subjective, often to the stakeholders opinion – Biased towards a part of population• Does not provide audience development metrics and BI tools• Limited value for marketing and sales 5
  6. 6. Example: Demographics-Based Segments 6
  7. 7. Why Not Demographics Segmentation?• Less relevant for design – Does not represent what users actually do online – Approximated needs limited to lifestyle data• Not relevant and not accurate – Sample-based approximation• Not real-time and expensive – A field research study takes at least 4-8 weeks 7
  8. 8. Current Media Pain Points• Which audience segments can be monetized and how to find them• How to sell ad space for a premium rate• How to increase monetizable traffic (PVs) 8
  9. 9. Why Behavioral Analytics?• Directly relates to usage and monetization metrics – Enables measurement and tracking of brand loyalty, user engagement and some of the attitudinal metrics – Quantifies user audience segments – Can be correlated with to audience performance on ad conversions• Real-time reporting of website usage by defined segment• Easy to integrate into the product lifecycle 9
  10. 10. 50 000 100 000 150 000 200 000 250 000 300 000 350 000 400 000 0 Sep 27 Sep 29 Oct 1 Oct 3 Oct 5 Oct 7 Oct 9 Oct 11 Oct 13 Oct 15 Oct 17 Oct 19 Oct 21 A Bit Too Late Oct 23 Oct 25 Oct 27 Oct 29 Oct 31 Nov 2 Launch Nov 4 Nov 6 New Design Nov 8 • New design launched Nov 10 Nov 12 Nov 14 Nov 16 Nov 18 Nov 20 Nov 22 Nov 24 Nov 26 Nov 28 Nov 30 Dec 2 Dec 4 Dec 6 Dec 8 Dec 10 Dec 12 – down trend of PVs with respect to UVs Dec 14 Dec 16 Dec 18 Dec 20 Dec 22 Dec 24 Dec 26 Dec 28 Dec 30 Jan 1 Jan 3 Jan 5 Jan 7 Page Views10 Unique Visitors
  11. 11. Site Redesign Example 11
  12. 12. Research Methodology marketing personas Focus Groups Personas• Interest clusters • Behaviors• Demographics • Needs Clusters • Usage Clusters • IA Design• Brand Loyalty • Attitudinal • Problem Areas • Audience Dev • Consumption • Brand Dev Intercept Behavioral Survey Analytics 8 preliminary behaviors 5 fundamental behaviors 3 – primary 2 – intermediate 12
  13. 13. Towards Integrated Product Cycle Business Intelligence Monetization AudienceMarketing & ROI Metrics Targeting Brand Behavioral Targeted Ads Development Analytics Audience Ad Sales Strategy Design for Behavior User Experience Product Development 13
  14. 14. Personas Research: Intercept SurveyBootstrap personas model from consumption behaviorperspective• Goals: – Capture entry points to the homepage. – Evaluate brand perception – Assess interest level in different information topics and services. – Assess user demographics and psychographics.• Evaluation Methodology: – Pop-up survey offered to homepage visitors – 25,781 survey responses. – averaged 4 minutes in length. 14
  15. 15. Results: Poor Content FitQ3: What are you looking for today? (yes/no answers)Q8: What are you interested in? (4-point scale) 15
  16. 16. Deriving Taxonomy 16
  17. 17. Personas Research: Focus GroupsDefine generalizable and extensible online consumptionbehavioral models applicable for other markets.• Goals: – Understand user needs, both met and unmet. Evaluate brand perception. – Understand psychographic and attitudinal characteristics of target audience segments. – Tie user interests, needs, and attitudes with their online consumption behavior. 17
  18. 18. Personas Research: Focus Groups Methods• 8 focus groups of 8 people each. – 2 focus groups in each location.• 2x4 groups design – Main independent variable: users and non-users Location Age Kids in latimes.com HH User 1 Pasadena 35-60 Mixed Yes 2 Pasadena 35-60 Mixed No 3 West LA 25-40 Mixed Mixed 4 West LA 35-60 Mixed Mixed 5 Downtown LA/ Silverlake 25-40 Mixed Yes 6 Downtown LA/ Silverlake 25-40 Mixed No 7 Irvine 25-60 Yes Mixed 8 Irvine 25-60 No Mixed 18
  19. 19. Focus Groups: Consumption Behaviors • Three primary behaviors. Individuals may exhibit all three behaviors. – Confirmed retroactively using qualitative and quantitative approaches – Comprehensive coverage of behavioral, motivational, and attitudinal factors • Individuals are classified into behavior profiles based on their most frequent behavior patterns. Grind Scan OpportunisticFocus Source Topic One factSource loyalty Strong Middling None – brand agnosticTime spent, Long, 1-2x /day Short, many times per day Very short, sporadicfrequencyOpinion Trusted unique Plural AnyMindset Fear Greed One fact Comprehensive (in subject area) Incremental knowledgeMotivation Habit – must finish (time or Habit by interest, hobby Concrete goal set by subject) external stimuliKey phrase “I go through a list of bookmarks” “Many sources” “My kids want to know …” “I start with this site every day” “I Google a lot” 19
  20. 20. Actionable Behavioral Analytics web analytics data Behavioral Classification Custom Research Actionable Metrics : Design • Content fit IA for each • Chanel reliance behavioral segment • Engagement • Brand loyalty • Provide relevant and actionable analytics and optimal strategy for monetizable audience growth • Maximize traffic by optimizing information architecture and content merchandising to fit site’s audience 20
  21. 21. Fundamental Behaviors (grind-scan TM )Engagement & • Primary behavior – Grind Source Reliance • Secondary – Scan • Highest engagement • Primary behavior – Scan • Frequent visits • Secondary – Grind • Typical for Sports or Maximum Focus: Entertainment fans Track Page Views Grind • Search traffic Subject • Focus on the topic Devotee • Loyal users • Seek plurality of • Focus on “trusted” sources opinion • “Grind” through a number of pages • Short but frequent Easily Monetizable • Fewer but longer sessions sessions Segments • Couple of visits per week Scan Focus: Unique Visitors Maximum Opportu nistic • Brand agnostic • External stimuli • Time constraint • Short sporadic visits Brand Adoption 21
  22. 22. Behavior ShiftsUVs, Homepage and News Sections800,000 6.3 % Earthquake day700,000600,000 54.4% 6.4%500,000 Opportunists400,000 4.8 % Scan 56.9%300,000 36.8% 3.9% 4.8% News Track200,000 33.8% 5.9% 6.0% 10.7% 35.4% Grind100,000 50.7% 30.8% 52.4% 0 Homepage, All News, Mon Homepage, Tue All News, Tue Mon • Relatively low grind % - not a primary source of information – Observed organic growth between a major news event and a Monday without major shift in behavior suggests problems with information architecture not allowing grind-like conversions (x3-4 PVs) 22
  23. 23. Local Vs. National Audience 100% 7,5 10,2 90% % UVs 80% 22,6 Scanner -like 70% 59,4 Opport 15,6 60% Scan 50% Subj Devotee 10,9 Track 40% Grind 3 30% 40,3 20% 30,1 10% 0% Local Brand National Brand• Brand loyalty, reliance, and engagement is better for the local brand – Local audience is better from advertisers’ perspective 23
  24. 24. Use of Navigation and Hot Topics Use of Navigation Use of Hot TopicsGrind Heavy for most: they grind Light. through Homepage and preferred section fronts. Some will not use nav. Target user of nav.Track Heavy. Use nav to scan. Medium – use during scan periods. Target user of Hot Topics.Scan Very light. Heaviest. Target user of Hot Topics.Opportunistic Some search, some use nav. Sporadic, mainly light.Subject Matter Devotee Some are heavy users, others Light. Use during scan have key sections periods. bookmarked. 24
  25. 25. Return Behavior Return Visit Behavior Inducements to ReturnGrind Regular. May be daily, Grinders return to our site because or Grinder may be on they are loyal: they value our a multi-day cycle. reputation and credibility. In addition, familiarity, usability, and appealingness are important. Past Days tab could induce Grinders to return more frequently, especially those on a multi-day cycle.Track Regular. May be daily, Inducements to return might include: or Grinder may be on • Timelines, A-Z pages, Past Days a multi-day cycle. Tab, and other ways to give them easy access to more details about the news. 25
  26. 26. Return Behavior Return Visit Behavior Inducements to ReturnScan Are most likely to land on the site as a result of Scanners are most likely to come back if we SEO – if their search term returns an have: LATimes.com result. • Good search – to allow them to find topics they’re interested in May return if they consider the LATimes.com an • Hot Topics – to promote topics they may be authoritative source on a topic they are interested in interested in. • Past Days Tab – to follow news from previous news days on topics they are interested in • Good contextual linking – to get them interested in other topics related to what they were originally looking forSubject Are most likely to land on the site as a result of Subject Matter Devotees will return if theyDevotee SEO – if their search term returns an consider us to be a comprehensive and definitive LATimes.com result. source for topics they are interested in. • A-Z pages would help them follow their chosen May return if they consider the LATimes.com an topics. authoritative source on a topic they are • Hot Topics bar giving a comprehensive interested in. overview. • Ways to track back a topic over time, including a Timeline or Past Days Tab. 26
  27. 27. Traffic Sources ( a local media example) In-Market In-Market (ext) Out-of-Market Grind 44.00% 42.80% 37.40% Track 15.30% 13.80% 7.50% Subj Devotee 17.30% 17.20% 14.00% Scan 15.10% 17.50% 28.40% Opportunist 7.70% 8.40% 12.10%  Key monetizable audience segments have non-local origin  Surprisingly, both Grinders and Subject Devotees show consistent distribution of Out-of-Market traffic  Scanner Out-of-Market trend is typical as it is driven by search Definitions: In-Market –county; In-Market Extended – region; Out-of-Market – everywhere else 27
  28. 28. Channel Reliance (a local media example) Average Visits Visited Once Channel Reliance Grind 25.4 47.7% Fair Track 88.5 0* High Subj Devotee 17.3 49.5% Fair Scan 4.85 75.1% Low Opportunist 4.19 82.4% Low • Key monetizable audience segment does not perceive the site as a primary source of information • Grinders and Subject Devotees have moderately low reliance • Scanners and Opportunists as expected have a high incidence of drive-by visitation * Trackers, by definition, are visitors that have visited the site more than once 28
  29. 29. Engagement Analysis (a local media example) Pages per Visit Day Engagement Grind 5.99 Moderate Track 15.32 Good Subj Devotee 6.11 Good Scan 1 Low Opportunist 4.19 Moderate• Key monetizable audience segments are reasonably engaged but could be improved through better content fit• Moderate engagement of Grinders may be indicative of Out-of- Market traffic 29
  30. 30. Content Fit Analysis (a local media example) SITE YOUR HOME PAGE NEWS SPORTS BUSINESS OPINION TOTAL NEWS [%]Grinder 43.1 62.5 43.9 33.3 40.0 27.9 28.8Tracker 13.2 22.8 18.5 9.2 16.9 19.7 12.0Subj Devotee 7.7 7.2 6.6 8.1 28.7 16.5 10.1Scanner 25.8 3.4 28.1 42.5 11.0 32.6 18.1Opportunist 10.2 4.1 2.9 6.8 3.4 3.3 31.0 • Overall content architecture requires re-focusing – Monetizable audience is underrepresented for News, Business, and Opinion – Sports section needs SEO investment – Huge potential with Opinion but failing to convert to repeated visitors 30
  31. 31. Behavior-Centric IA 1 Easy access to past dates 1 2 designed for: Trackers, Grinders 3 2 Horizontal Navigation Grinders, News Trackers 4 3 “Hot News” topics for easy access to the prominent news topics Subject Devotees, News Trackers 4 Visual preview of topics on the 5 Homepage Scanners 5 Local, aggregated breaking and feature news from many sources Scanners, Trackers 7 6 6 Personalized news Scanners, News Trackers, Opportunists 7 Aggregated “best of” content from around the Web Scanners, Trackers 8 8 Mouseover headline preview News Trackers, Scanners 31
  32. 32. Behavior and Interests (a national media example ) Grinders Scanners World & National News (47.9%) World & National News (63.8%) Sci & Env (34%) Environment & Business (21.4%) Business (29%) Crime (9.8%) Entertainment & Living (50%) Entertainment & Living (26.3%) Leisure & Shopping (12.6%) Local News (80.6%) Sports (20.4%) Sports (28.2%) SoCal News (59.8%)Classifieds (2.4%) • Interests of Grinder segments are used to derive primary and secondary navigation to improve reliance and engagement • Scanners interests are used to derived article-level related content to maximize PVs 32
  33. 33. Behavior-Centric IA: Article Page • Different for grinders and scanners – Drives more PVs 33
  34. 34. Behavior-Centric IA: TaxonomyLocal News U.S. & World News Entertainment Sports Business & Technology Travel Health Living More• Several sections consolidated under one header• Drives users to high-CPM, high- traffic areas 34
  35. 35. Extending Personas demographics grinder>55, affluent 55-40, affluent 55-40, mid-income 25-40 affluent, usage grinder grinder grinder grinder marketing personas 55-40 25-40 grinder grinder • Behavior is primary • Not all permutations make sense 35
  36. 36. Audience Segments Revised Primary: Grinders Secondary: News Hounds Primary: Scanners Subject Matter Secondary: Devotees News Trackers Primary: Subject Subject Devotees Devotees Sports Grinders Primary: Grinders Primary: Scanners Secondary: Secondary: Opportunists Sports Grinders News Hounds News Hounds Subject Devotees 36
  37. 37. Audience Strategy (example) Homepage Section Fronts Story-level Page VideoRetain Grind (>40) – maintain Grind and SDs/News Scan (all Zones)– <35 – maintain brand source reliance Trackers– maintain maintain UVs consideration brand consideration Focus • Aligning with the interests Ease of access SEO Content merchandising • Show scopeGrowth Trackers SDs (all zones) & Grind Grind (all zones) – SDs / Trackers(<35) and (all zones) – improving and Trackers(<35) – improve stickiness . Scanners – improve Stickiness and source improve brand Scan (<35) – improve brand consideration reliance consideration and content syndication source reliance Focus • News aggregation (SoCal • Aggregated section • Content syndication • Source agnostic and Breaking) fronts • Contextual integration • Improve web curation • Content partnerships aggregation • Better SoCal coverageAcquisition • New Grinders, News New Subject Devotees Convert Scan behavior Grinders (>40)- improve Trackers(<35), Scanners - – expand brand into News Hounds/ acceptance expand brand consideration. SDs – build brand consideration. consideration • Opportunists - utility focus Focus • Improve programming Improve the • Marketing campaigns • Usability • headline syndication to programming to • Content • Local programming major aggregators capture niche interests merchandizing 37
  38. 38. Audience Risks Grinders Trackers Scanners Subject Devotees (primary – grind; (primary- scan; secondary secondary -scan) grind)Brand Easy to maintain, Less predictable than Mostly agnostic to Can be loyal to aconsideration loyal audience (over Grinders – follow the brands (majority of particular section.(UVs per month) 40). Hard to recruit news (especially <35). To improve PVs More UVs –improve new (>35) – need >35). LAT is a – SEO and content the programming acceptance as a secondary source. To syndication strategy (e.g., sports, health, trusted source. improve UV – focus auto). on Scan acquisitionSource reliance Usually low. Hard to Medium. To improve Centered around Could be higher than(visits per day) increase – interferes (<35 + core) more particular topic or Scanners, if the with the habit breaking and relevant news. To improve – source accepted as (SoCal) news focus on Breaking primary news (<35)Engagement High. PPV easy to Can be as high as for Usually low. To Higher than Scan. To increase with related Grind. PPV increase – increase PPV – focus increase PPV –(PPV) info aggregated news on related info improve depth of aggregation coverage Low Risk High Risk 38
  39. 39. Results new IA 2008 2009 2010• Nearly doubled UVs• Nearly doubled PPVs 39
  40. 40. Limitations• Need existing site to generate data – A developed audience should exist• Mechanics of behavioral segmentation is not trivial• Investment is needed to take advantage – Integration with CMS and other BI or CRM platforms – Retagging may be required for web analytics• May require organizational and practices change 40
  41. 41. Where Does It Fit?• This in no way means you don’t need other UX research methods – In fact, it is a complimentary method• This is a good bootstrapping methodology as any interface has its peculiarities 41
  42. 42. Behavior Analytics Today (Ad Industry) current click history interest target past extrapolated approximation “behavior” from external data suppliers lifestyle segments• Modeling interest rather than behavior• Assumes user’s interest remain the same – Inaccurate: 80-90% of ad impressions are wasted 42
  43. 43. Next Level of Behavioral Analytics measurable behavior computed intent behavioral targeting likely to consume any offered content & media Engagement Habits likely to consume only relevant content & mediaAdoption Reliance Attitudes Motivations not likely to consume any extra media• behavior-analytic engines provide translation from observable behavioral metrics to intent-level cognitive modeling followed by proprietary mapping into actionable behavioral profiles. 43
  44. 44. Display Ad Perception/ImpactEngagement & Source Reliance Open to consume Track suggested content Grind Subject Devotee High Scan Ad Recall Moderate Opport Ad Recall Use of Interactive Ads unistic Low Ad Recall Brand Consideration 44
  45. 45. Audience Engineering• Powerful behavior-centric analytics – Maximum online audience understanding and tracking of motivations, habits, behavior, and more classify• Relevant to what users actually do online – Achieve maximum cohesion of User Experience and business goals through behavior-centric design architectures• Complete cycle of audience engineering – Maximizing revenue and enabling proactive audience and brand development optimize 45
  46. 46. Вопросы?Sanzhar Kettebekov, Ph. D.sanzhar@segmentinteractive.com
  47. 47. Social Networks Behavior (Dwell-Scan™) • Extravert behaviorConsumption & • Not as good consumptionContribution as Dwellers • May use multiple Maximum platforms blogs, tweeter Dwell • Introvert behavior • Loyal users • Interest/topic driven Broadcast • Hooked on the source • May not have • Emotionally invested distinguished info • Use it as a primary source • Focus – address channels for getting info book utility • Demand driven • Brand agnostic visits • External stimuli • Time constraint Scan • Short sporadic visits Maximum Communicate Opportunistic Channel Adoption 47
  48. 48. Online Video Consumption (Dwell-Seek™)Consumption • DVR behavior Subject • Focus – shows of Dwell interest Devotee • Mixed genre • Mixed media usage including TV • Library access • Loyal source • Seeker behavior behavior devotees • Interest in the particular • Genre(s) driven • Developed visiting topic that instance interest habits and source • Ease of access • Not as good source reliance Track reliance as Dwellers • Use it as a primary discrimination between the sources • May use multiple source of online sources entertainment • Brand agnostic • External stimuli Seek • Time constraint • Infrequent visits Opportunistic Source Reliance & Adoption 48
  49. 49. Ad Strategy AD High ad tolerance Moderate ad tolerance Low ad tolerance 49

×