Attribution Demystified: Digital World Expo 2012

897 views

Published on

This presentation on Attribution and Media Measurement was made at the 2012 Digital World Expo. For an electronic copy please email info@encoremetrics.com

Published in: Business, Technology
0 Comments
1 Like
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total views
897
On SlideShare
0
From Embeds
0
Number of Embeds
10
Actions
Shares
0
Downloads
0
Comments
0
Likes
1
Embeds 0
No embeds

No notes for slide

Attribution Demystified: Digital World Expo 2012

  1. 1. Online Attribution !Media Measurement Demystified!September 2012! Steve Latham, CEO Encore Media Metrics @stevelatham
  2. 2. Background and Objectives My Background q  Founded, grew and sold a digital media agency (2002-2010) §  Managed campaigns for FedEx, Continental Air, 76 Gas, The Scooter Store §  Intense focus on accountability and results q  Launched Encore Media Metrics in 2011 §  Media measurement, attribution and reporting as managed service My Objectives for Today q  Define the problem q  Explain Attribution q  Show where we are (and where we’re going) q  Keep you from falling asleep!@stevelatham Information is confidential and may not be shared. © Encore Media Metrics 2012 2!
  3. 3. Now let’s have some fun with digital media measurement! …Said No One. Ever@stevelatham Information is confidential and may not be shared. © Encore Media Metrics 2012 3!
  4. 4. ContextSince early days, we’ve measured performance via: q  Clicks and Click-through-rates (CTRs) q  Direct Conversions and Cost Per Action (CPA) q  Credit allocated to Last-Click or Last-ViewWe now know these are no longer sufficient q  Customers utilize numerous channels q  Keywords often play an Assist role q  Display Ads create awareness §  Success is NOT correlated to clicks §  Only small % of post-view conversions ! involve display clicksThe Solution: Multi-Touch Attribution q  Identify all interactions that precede conversions q  Attribute value to each interaction that plays a supporting role @stevelatham Information is confidential and may not be shared. © Encore Media Metrics 2012 4!
  5. 5. Distinguishing AttributionTwo Approaches q  Statistical – looking for relationships between data sets §  Paid clicks vs. Conversions §  Impressions vs. Conversions §  Impact of Display on Search §  A/B Testing q  Operational – analysis of conversion paths for each visitor §  Lower-funnel (click-based) §  Full-funnel (clicks + impressions) §  Types of Models •  Equal weighting •  Subjective weighting •  Algorithmically weighted @stevelatham Information is confidential and may not be shared. © Encore Media Metrics 2012 5!
  6. 6. Operational Approach (Steps 1 and 2) 1 Aggregate Visitor Data Based
 2 Synch Ad Server Impression 
 on Source:! Data to See Engagement Paths! ►  Display ads ! ►  Record all interactions that ►  Search (paid and natural)! precede visit or conversion! ►  Email! ►  Note: requires data cleansing! ►  Affiliate, CSE! ! ! Site Display Display Vist 
 Imp: #1! #1! Imp: #4! SEM! Display! Email! Affiliate! Site Display Vist 
 Display Imp: #6! #2! Imp: #5! Master DB! Site Vist Display Conversion! #3! Imp: #7!@stevelatham Information is confidential and may not be shared. © Encore Media Metrics 2012 6!
  7. 7. Operational Approach (Steps 3 and 4) 3 4 Produce Actionable Reports! Attribution Modeling! ►  Data Cleansing! ►  Engagement paths ! ►  Statistical modeling ! ►  Attributed Actions & CPA! ►  Weigh impressions vs. clicks! ►  Optimal Frequency! Ac#onsByChannel 4,000 Last Click Attributable 3,526& 3,500 3,000 2,904& 2,679& 2,504& 2,500 2,207& 2,062& 2,002& 2,000 1,500 1,000 256& 307& 436& 423& 504& 500 0 Direct Nav Org Search Referrals Email Paid Search Display@stevelatham Information is confidential and may not be shared. © Encore Media Metrics 2012 7!
  8. 8. Sample Reports Reports should be… q  Insightful q  Actionable q  User-friendly 10 Impressions*Required*by*Vendor* 9.6$ Attributed)Cost)Per)Action 9 Visit$ Conversion$ 9.0$ Vendor CPA 8 7.8$ Publisher*8 $********** 32.43 7 7.0$ 7.0$ 6.6$ Ad*Net*1 $********** 42.53 6.2$ 6 5.4$ Publisher*9 $********** 44.88 5 4.7$ 4.4$ Publisher*5 $********** 64.46 4 3.5$ 3.2$ Publisher*7 $********* 105.53 3.0$ 3 2.4$ 2.4$ Publisher*2 $********* 108.83 2 1.7$ 1.8$ 1.8$ 1.3$ 1.6$ 1.3$ Publisher*1 $********* 126.56 1.1$ 1 Publisher*3 $********* 142.38 0 Publisher*6 $********* 269.87 Ad*Net*2 $********* 374.89 8 1 9 7 2 1 3 2 5 6 4 r et r r r r r et r r r he he he he he he he he he N N is is is is is is is is is d d bl bl bl bl bl bl bl bl bl A A Publisher*4 $********* 919.04 Pu Pu Pu Pu Pu Pu Pu Pu Pu@stevelatham Information is confidential and may not be shared. © Encore Media Metrics 2012 8!
  9. 9. Conversion Path Analysis Quantifies impact of each channel Conversion)Path)Analysis Engagement)Overview Sources)of)Visits)within)Conversion)Paths)========> IMPs Visits Display Direct0Nav Nat.0Search Paid0Search Referring All Converters (100%) 5.2 3.0 0.5 0.4 0.8 0.5 0.7 Relative Contribution 54.2% 45.8% 18.0% 14.0% 26.7% 18.0% 23.0% Cluster 4 (45%) 1.5 2.8 0.8 0.0 0.1 0.0 1.8 Cluster 6 (21%) 4.7 3.8 0.2 0.6 0.8 1.5 0.6 Cluster 3 (15%) 6.1 4.0 1.5 0.2 1.8 0.3 0.2 Modeled Conversion4.8 Cluster 8 (13%) 3.9 Path data shows: 0.0 1.2 2.0 0.1 1.3 q  Converters were exposed to 5.2 display ads q  Converters visited 3.0 times before taking action q  Search comprised 45% of visits in conversion process@stevelatham Information is confidential and may not be shared. © Encore Media Metrics 2012 9!
  10. 10. Channel Level Attribution Attributing credit across all touch-points  deeper insight into performance of each channel Ac#onsByChannel Cost%Per%Ac+on%By%Channel%4,000 $80 Last Click Attributable $71.29 3,526& CPA Last CPA Attrib.3,500 $703,000 2,904& 2,679& $60 2,504&2,500 2,207& $50 2,062& 2,002&2,000 $401,500 $30 $19.53 $17.95 $20 $16.281,000 256& 307& 436& 423& 504& $8.95 $9.23 500 $10 0 $0 Direct Nav Org Search Referrals Email Paid Search Display Email Paid Search Display q  Display is usually the laggard when q  Display CPA is overstated by “last viewing “last click” results click” metrics q  Attribution reveals which channels are q  Often see 50-300% lift in actions driving customer actions and 33-66% drop in CPA @stevelatham Information is confidential and may not be shared. © Encore Media Metrics 2012 10!
  11. 11. Vendor Level Attribution View the true performance for each vendor (last-click vs. attributable) Vendor Level Report Last-Click vs. Attributed CPA Last Click Attributable Cost%Per%Ac+on%by%Vendor% Display Vendor Rank CPA CPA $1,200 Ad Network 1 $ 52.16 $ 25.81 Winner Last Click Attributed $1,000 DSP $ 72.49 $ 35.98 Winner Ad Network 2 $ 130.51 $ 38.46 Winner $800 Publisher 2 $ 138.12 $ 48.46 Winner Ad Network 3 $ 207.53 $ 54.24 Challenger $600 Publisher 1 $ 155.23 $ 57.33 Challenger Publisher 3 $ 210.67 $ 149.87 Laggard $400 Publisher 2 $ 292.97 $ 154.01 Laggard $200 Totals $ 128.57 $ 50.93 $0 8 1 9 5 7 2 1 3 6 2 4 r et r r r r r r r et r he he he he he he he he he N N is is is is is is is is is d d bl bl bl bl bl bl bl bl bl A A Pu Pu Pu Pu Pu Pu Pu Pu Pu q  As shown above, many vendors are short- q  Wide discrepancies can exist between changed by last-click analysis “Last-Click” and Attributable CPA (Contributors vs. Closers)@stevelatham Information is confidential and may not be shared. © Encore Media Metrics 2012 11!
  12. 12. Keyword Performance How Most View Keyword Reports (%-51<"&1 .+/%0-1 .+/%0- (<.1234-51 !"#$%& ()*+,- :;"0 ()*+,1 234-51()*+,6 2.7&*896 ()*+,6 2(<(61 !"#$%&() *+),- -. ./0 1-2/ 103./ 1..3/. !"#$%&(.. *+4*2 -. 2, 1/42 103*- 1./32/ !"#$%&(*2 *+,)) -2 2. 1)*0 1035, 1.)3,2 !"#$%&(. /+.., -- .0* 142. 103.- 1.)340 !"#$%&(5. .+.*4 *, .0, 122* 103/4 150344 !"#$%&(*4 2+.-) -0 .// 1.+*-0 103.2 15.300 !"#$%&(.2 .+424 ,0 .2- 1.+2./ 103). 15-35* !"#$%&(.* *+-4, 50 .-4 1.+*4) 103,* 1-*3)2 !"#$%&(.) 5+4-2 5* ..2 1.+-), 1035) 1-/32* !"#$%&(*/ .+5)0 -/ ..0 1*+*5* 1.3/. 1-43,* !"#$%&(- )+0// -, .-0 1*+*.4 103*- 1-)3*) !"#$%&(/ -+50, .4 )- 1)/2 103** 1,532* !"#$%&(./ 5+*.- -* ... 1*+*)- 1032. 1,-3/* !"#$%&(5- ..+/40 ,0 .-5 1*+4-4 103*- 1,/3)/ !"#$%&(*. 2+./- *4 .5, 1.+4.) 103*, 1/-3)/ !"#$%&(.- -+5,2 .) .-- 1.+*-2 103*) 1/,3/5 !"#$%&(.4 ,+0-0 50 ). 1*+042 103-. 1/)3,2 !"#$%&(5* .+*4) .5 2, 1)/0 1032- 12534, !"#$%&(*, -+,24 .* .0/ 1)4/ 103** 14*3.2 !"#$%&(4 )+20, */ .-4 1*+552 103*- 14)344 !"#$%&(50 5+4)- *4 .-. 15+.55 10340 1...34)@stevelatham Information is confidential and may not be shared. © Encore Media Metrics 2012 12!
  13. 13. Keyword Attribution How We Should View Keyword Performance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stevelatham Information is confidential and may not be shared. © Encore Media Metrics 2012 13!
  14. 14. ROI Modeling Use data to forecast value of optimizing spend % of Proforma Proforma Proforma Vendors Spend ROS Revenue Impact Budget % Budget ROS Revenue Winners $21,781 21% $19.18 $417,792 65% $19.18 $1,297,843 $880,051 Challengers $35,745 35% $6.46 $230,910 35% $8.40 $300,183 $69,273 Laggards $45,880 44% $1.21 $55,428 0% n/a $0 ($55,428) Total* $103,407 100% $6.81 $704,130 100% $15.45 $1,598,027 $893,896 *excludes buys not tracked by Encore Relative Improvement 127% Incremental Revenue $893,896 q  Insights from Attribution forecast 127% improvement opportunity@stevelatham Information is confidential and may not be shared. © Encore Media Metrics 2012 14!
  15. 15. Additional Considerations Accounting for Excess Impressions q  Many assist impressions are not necessary q  Need to remove them to have a clean data set Modeled&#&of&Transitions&using&Mean NumberofTransitionsReqd&for&a&VISIT& &&&&&&&&&&&& 3.9 Impression*Distribu6on* 60,000" NumberofTransitionsReqd&for&a&CONVERSION &&&&&&&&&&&& 5.6 DSP"Only" Other"Vendors" 50,000" Modeled&#&of&Transitions&using&Median Impressions*Count** NumberofTransitionsReqd&for&a&VISIT& &&&&&&&&&&&& 2.5 40,000" NumberofTransitionsReqd&for&a&CONVERSION &&&&&&&&&&&& 3.5 30,000" Actual#ofImpressionsServedperConverter &&&&&&&&&& 19.3 EfficiencyOpportunity(%tobere?allocted) 70.8% 20,000" 10,000" 0" 1" 2" 3" 4" 10 o"9" 20 "19" 30 "29" 40 "39" 50 "49" 0"t 9" 0"t 9" 0"t 9" 0"t 9" 10 0"to 9" 20 "to"1 9" 30 "to"2 9" 40 "to"3 9" "to 9" 50 9" +" 00 10 "to"9 20 o"19 30 o"29 40 o"39 50 o"49 00 "99 00 99 00 99 00 99 9 "49 "to "to "to "to 5"t Impressions*Served*Before*Visit*@stevelatham Information is confidential and may not be shared. © Encore Media Metrics 2012 15!
  16. 16. Additional Considerations Defining the right attribution model q  What is the value of an impression vs. a click? q  How far back should you look? è Use machine learning and statistical modeling Allocating credit for Offline conversions to Online engagement q  Quick and easy way: use post-purchase visits §  Dedicated pages for customers (not indexed) §  Connect post-purchase visits to previous Impressions and Visits Excluding Out-of-View Impressions q  Requires ad verification tools (e.g. AdSafe, Adogoo, Double Verify) §  Identify out-of-view Impressions §  Exclude from Attribution analysis@stevelatham Information is confidential and may not be shared. © Encore Media Metrics 2012 16!
  17. 17. Where Are We Today? Using Geoffrey Moore’s model from Crossing the Chasm… q  2008-2010 Innovators q  2011-2012 Early Adopters q  2013-2014 Early Majority q  2014-2015 Late Majority@stevelatham Information is confidential and may not be shared. © Encore Media Metrics 2012 17!
  18. 18. Takeaways 1. Media measurement is changing q  Moving from last-click to holistic views 2. Don t Fear Attribution. Embrace it! q  Be a thought leader, not a laggard 3. Choose approach that fits your strategy q  Lower-funnel (SEM, SEO, Facebook ads, email) q  Full-funnel (Display, Rich media, video, etc. 4. Take a pragmatic approach q  Walk before you run q  Don’t wait for perfect solution (it doesn’t exist) For more info… -  Visit our blog: http://Attribution101.com @stevelatham Information is confidential and may not be shared. © Encore Media Metrics 2012 18!
  19. 19. Q&A Encore Media Metrics Steve@EncoreMetrics.com Steve Latham, Founder and CEO 1633 Broadway, 5th Floor @SteveLatham New York, NY 10019 @EncoreMetrics 646.820.1006 http://EncoreMetrics.com http://Attribution101.com @stevelatham Information is confidential and may not be shared. © Encore Media Metrics 2012 19!
  20. 20. About the Presenter Steve is the founder and CEO of Encore Media Metrics (http://encoremetrics.com), which helps marketers optimize spend through better analytics. Serving leading brands and agencies around the globe, Encore s on-demand solution enables clients to achieve deeper insights into campaign performance while lowering the cost and complexity of online measurement. Prior to launching Encore, Steve founded, grew and sold Spur Interactive, a digital marketing agency that specialized in strategic planning and execution of integrated media campaigns. In this role, Steve planned and executed successful campaigns for leading brands, including FedEx Office, Continental Airlines, ConocoPhillips and The Scooter Store. Steve is an accomplished speaker and thought leader, speaking frequently at industry events including Ad-Tech, OMMA Global, Search Engine Strategies, IAB Mixx, eMetrics Summit and others. Steve s articles have been published by Adotas, MediaPost, Online Media Daily, iMedia Connection, Marketing News and Bizjournals.com. Steve has been quoted in B2B Magazine, PR Week, Fortune and CNN.com. Steve received an MBA from Harvard Business School and a BBA from the University of Oklahoma. You can read Steve s blog at http://Attribution101.com. http://facebook.com/slatham http://twitter.com/stevelatham www.linkedin.com/in/stevelatham @stevelatham Information is confidential and may not be shared. © Encore Media Metrics 2012 20!

×