The Complex Journey to Unified Marketing Analytics
3 likes1,018 views
The document discusses the evolving landscape of marketing analytics, highlighting the need for unified metrics and insights amid increasingly complex consumer behaviors and media channels. It emphasizes the importance of integrating Marketing Mix Modeling (MMM) and attribution to enhance accountability and optimize marketing strategies. The exploration of ensemble models is presented as a potential solution to reconcile differences between these methodologies and improve overall performance measurement.
The Complex Journey to Unified Marketing Analytics
- 1. Navigate the Complex Journey Towards Unified Marketing Analytics Joy Joseph Practice Leader IRI Strategic Analytics George Musi SVP, Head of Analytics & Insights Blue 449
- 2. Discussion Agenda The Convergence Imperative for Marketing Metrics Marketing Performance Landscape MMM &. Attribution- what's all the fuss about? Charting a course towards unification Ensemble Models- Vision for the future
- 3. Discussion Agenda The Convergence Imperative for Marketing Metrics Marketing Performance Landscape MMM &. Attribution- what's all the fuss about? Charting a course towards unification Ensemble Models- Vision for the future
- 4. Today’s media landscape is evolving at an unprecedented rate BLACK & WHITE TV FIRST SATELLITE BROADCAST RISE OF CABLE TV MORE ADVERTISER- SUPPORTED NETWORKS DIGITAL AND INTERNET REVOLUTION SMARTPHONES HDTV TABLETS OVER-THE-TOP DISTRIBUTION 1950s 1960s 1970s 1980s 1990s 2000 - 2005 2006 - 2011 2012 AND BEYOND XBOX LIVE CAMERA IN DEVICES/TV- LEARNING THE CONSUMER TARGETED ADS PERSONALIZED RECOMMEND- ACTIONS ENHANCED CONTENT SEARCH CAPABILITIES CLOUD-BASED CONTENT DELIVERY SOCIAL TV DIGITAL RADIO VIA MOBILE AND TABLET APPS … DVR FIRST DOWNLOAD- ABLE CONTENT SOLD MP3 PLAYERS VIDEO STREAMING DIGITALTV CONNECTED WORLD
- 5. Current State of Media: Consumers are overwhelmed and Marketers are Confounded Buy this SALE! This just in Did you know? Click here Like this 50% off Tweet it! National reach Incremental spend ROAS Cross- channel Our data has scale Purchase data
- 6. The complex media landscape and rapidly evolving technology resulted in a complex and often confusing environment Execution Infrastructure Channel Segments Life-cycle & Metrics
- 7. Influential Advertisers have turned up the pressure on accountability on the media industry Heightened media scrutiny will inevitably put pressure on the performance measurement ecosystem for greater consistency & convergence P&G Chief Brand Officer Marc Pritchard doesn't "want to waste time and money on a crappy media supply chain," he said. And he urged others in the industry to follow suit. (AdAge Jan 29, 2017 http://bit.ly/2jmz5kw)
- 8. Discussion Agenda The Convergence Imperative for Marketing Metrics Marketing Performance Landscape MMM &. Attribution- what's all the fuss about? Charting a course towards unification Ensemble Models- Vision for the future
- 9. A rapidly evolving media landscape has resulted in complex & multi-dimensional consumer journeys Smartphone Tablet PC TV 11AM 12AM 1PM 2PM 3PM 4PM 5PM 6PM 7PM 8PM 9PM 10PM 11PM 12PM BED KITCHEN WORK LUNCH WORK HOME RESTAURANT BAR BED 7:30 CNNMobile (5m) 8:45 Instagram: Walking to work (2m) 10:05 Texting (15m) 10:45 Snapchat (1m) 11:15 Texting (5m) 3:35 Daytrotter: Streammusic (1h) 4:30 Yelp: Restaurantsearch (10m) 5:35 Weather: Checkweather (2m) 7:10 Texting (10m) 9:45 Facebook: Watchvideo (15m) 8:35 Snapchat (1m) 8:10 GeneralSurfing (20m) 6:45 Yelp: Restaurant search(15m) 7:00 Open Table:Make reservation(10m) 10:30 Yahoo! News (15m) 11:00 Facebook: Post articles from Yahoo!(5m) 1:00 At work on desktop(4h) 4:45 Gchat withboyfriend (10m) 10:35 Netflix(30m) 8:10 Watched NBC Today Show(20m) 6:45 VEEP: Watch 2 episodes(45min) 7AM 8AM 9AM 10AM
- 10. Consumers move between screens, places, and moments…
- 11. Despite this, fundamental media objectives have not changed CONSUMER MARKETER RIGHT PERSON RIGHT MESSAGE RIGHT TIME RIGHT PLACE RIGHT CONTEXT ENGAGE ASK SHARE RESEARCH TRUST DISCOVER AWARENESS ADVOCATE BUY CONSIDER BROWSE CONSUMER DECISION JOURNEY C A M P A I G N L I F E C Y C L E
- 12. In today's complex atmosphere, marketers need tools that can accurately make sense of myriad and disparate data Marketers need breadth & depth of insights to make the timelier and more informed decisions at the most critical moments Audience (Who) Creative (What) Time (When) Placement (Where) 001010101 001001010 100101000 000111010 001010101 001001010 100101000 000111010 001010101 001001010 100101000 000111010 001010101 001001010 100101000 000111010 001010101 001001010 100101000 000111010 001010101 001001010 100101000 000111010 001010101 001001010 100101000 000111010 001010101 001001010 100101000 000111010 001010101 001001010 100101000 000111010 001010101 001001010 100101000 000111010 001010101 001001010 100101000 000111010 001010101 001001010 100101000 000111010 Transform data into meaningful insights that drive actions
- 13. Reduce time to opportunities Event Data captured Insights delivered Action taken Time Value Action Time Enables us to become more responsive & adaptive to opportunities by reducing the time form data to decisions
- 14. Support a more iterative process of measurement, optimization and simulation PLAN & EXECUTE MEASURE OPTIMIZE TO INTERATIVE & ADAPTIVE PROCESS Marketing Campaign
- 15. Need to be an equal mix of… SCIENCE (STATISTICAL & ANALYTICAL KNOW- HOW) ART (MARKETING & BUSINESS UNDERSTANDING)
- 16. Integrated attribution modeling approach Digital Attribution Modeling Marketing Mix Modeling (Top-Down) (Bottom-Up) Holistic Insights Combining Approaches In The Right Way Micro Decisions Macro Decisions To Improve Marketing Optimization
- 17. Marketing response modeling overview In the marketing analytic space, there are many types of marketing models The generalized goal of these approaches is to develop an understanding of how marketing activities & non- marketing influencers impact business outcomes, like card applications & acquisitions Machine learning (e.g., agent- based models) Bottom-up (e.g., attribution models Econometric top-down (e.g., marketing mix models) Three of the most common approaches to model marketing
- 18. Marketing performance management MMM TV Attribution Attribution FOR BREATH OF UNDERSTANDING AND TO GUIDE OVERALL MARKETING STRATEGY FOR DEPTH OF UNDERSTANDING AND TO GUIDE TACTICAL CAMPAIGN PERFORMANCE Three-stage Modeling Approach: We believe in a comprehensive marketing performance management approach – leveraging MMM, Attribution & TV attribution in concert – to: • Expand the scope & precision of measurement • Better enable both (long-term) strategic & (near-term) tactical decision support • better empower both macro & micro optimizations
- 19. Evolve decision making Decision Support OptimizeUnderstand Foresight Decision Guidance DESCRIPTIVE What Happened? Hindsight DIAGNOSTIC Why Did It Happen? ? ? ? ? ? Insight PREDICTIVE What Will Happen? PRESCRIPTIVE What Should I Do?
- 20. Use Assortment Of Analytical Techniques Analyzing The Present Reporting On The Past Predicting The Future To get the most benefit from analytics, we’ll need to have an analytic assortment that is balanced – one that combines techniques for:
- 21. provides increased… Brush Strokes? Color Mixture? Composition? Full Picture? Clarity Confidence
- 22. ULTIMATE GOAL BETTER & CONNECTED DATA BROADER & DEEPER INSIGHT TIMELY & SMARTER ACTION
- 23. Discussion Agenda The Convergence Imperative for Marketing Metrics Marketing Performance Landscape MMM &. Attribution- what's all the fuss about? Charting a course towards unification Ensemble Models- Vision for the future
- 24. Why is there friction between MMM & Attribution? 1. Wrongfully viewed as substitutes rather than complements 2. They speak different languages in terms of metrics and interpretation 3. The same activity measured in MMM vs. Attribution can have different ROI 1 2 3
- 25. Understand that MMM addresses different business needs from Attribution… 1
- 26. The advent pf programmatic has further exacerbated the great divide 1 MMM Attribution
- 27. MMM today does address tactical questions for offline media 1
- 28. Digital Measurement in MMM is typically at total platform level 1
- 29. Attribution operates at much more granular levels required to optimize digital campaigns 1
- 30. MMM & Attribution speak different languages – “contributions” = "lifts” 2 MMMs measure marketing contributions vs. campaign “lifts” in Attribution- o Contribution = % of annual sales that was incrementally driven by marketing o Campaign lift is the incremental sales generated by consumers exposed to the campaign during the campaign period compared to those not exposed. Lifts are typically higher than contributions because of a smaller denominator -campaigns run over weeks or months not a full year. For CPG and other faster moving goods, a well-run campaign typically results in substantially higher trial rates, above normal consumption levels, often pulling purchases forward ahead of normal purchase cycles during campaigns (purchase acceleration)
- 31. MMM & Attribution speak different languages – “ROI” = “ROAS” 2 ROI measured in MMM is typically $ profits generated per $ spent of a marketing activity and is more appropriate when making cross-platform annual budget allocations where EDIT impact is an important consideration ROAS or Return on Ad Spend measured in Attributions are usually $ Sales generated per $ spent on a marketing activity- does not take brand profitability into account
- 32. MMM & Attribution can still yield divergent results- they are after all inherently different methodologies It is however important to align the two, primarily because MMM is still used to manage the annual strategic budget allocation process, including digital platforms Reconciliation starts with recognizing limitations 3
- 33. Recognize MMM limitations… Digital is challenging for MMM where geo data (down to markets or trade areas) is infeasible o Recent advances here- e.g. geo-fenced social campaigns matched to store-level mix data sets. Geo-targeted campaigns do not have the scale to register accurately in MMM o Reading local Campaigns with less than 20 Million impressions in MMM is a coin toss Traditional MMM “fits one size to all”, while marketing is increasingly targeted Synergies are important in cross-platform campaigns and are difficult to read without either sufficient variation in execution or single source data (down to individual audience level) Finally, the broad and encompassing demand measurement MMM undertakes comes with long cycle times- model updates can often take upwards of 8 weeks from data gathering to insights/execution o Some advances here as well with API driven data integration and “self-updating” models 3
- 34. And understand Attribution pros and cons… Pros o “Bottom-up” approach, measuring total impact up from consumer transaction level o Ability to measure impact and optimize spend down to publishers, content and creative o Rapid cycle-times (2-3 weeks) o Supports digital-specific requirements: o Pay-for-performance o Addressable platforms o Niche- targeted campaigns and Impression level optimization o Exclusion of Offline Platforms o Single source Data integration challenges- data fusion techniques widely used introduce biases- o Stratified Sampling due to limitations in consumption data o Lookalike matching where “walled gardens” prevent direct matching across platforms due to PII restrictions o Models run on short windows are inherently tactical, lacking any longer-term brand impact o Does not quantify non-marketing drivers’ impact 3 Cons
- 35. Discussion Agenda The Convergence Imperative for Marketing Metrics Marketing Performance Landscape MMM &. Attribution- what's all the fuss about? Charting a course towards unification Ensemble Models- Vision for the future
- 36. The Unification journey starts with a process of course correction between the two approaches
- 37. Course correction ensures Ensuring “Single Version of Truth” 4th of July Father’s Day Mother’s Day Attribution Average Without MMM input MMM Average Without Attribution input Back To School DigitalDisplayROAS
- 38. Single Version of Truth ensures harmony across organizational planning milestones
- 39. Discussion Agenda The Convergence Imperative for Marketing Metrics Marketing Performance Landscape MMM &. Attribution- what's all the fuss about? Charting a course towards unification Ensemble Models- Vision for the future
- 40. Meta models or “Ensemble Models” can resolve Conflicts between models by establishing rules and relationships to combine models “There are a number of functional components for such a system sitting in proprietary assets, we just need the right motivation to transcend our “walled gardens” to allow them to work together”
- 41. Platform Lift Campaign Lift Brand Equity Lift MMM Attribution Brand Health Metrics Copy Tests Engagement Brand Affinity Creative Affinity Importance of an input= Its contribution to overall system performance and accuracy Marketing Analytics System- Bayesian Belief Network Latent MetricsInputs Output KPIs
- 42. A more immediate possibility are hierarchical models that drive both micro and macro decisions Source: IRI-PepsiCo presentation at IRI Growth Summit 2016
- 43. Immediate future will see a more cogent integration of MMM and Attribution A greater focus on targeting will push measurement down to at least Household-level, if not individual consumer level Now that we have stopped being awestruck by the sheer amount of data that is creating “insights”, it is time to address issues like data bias Focus will shift to predictive accuracy vs. backward looking campaign scorecard Advanced probability based models will be explored to compensate for data biases and for understanding non-addressable offline media Closing thoughts….
- 44. Thank You!!