際際滷

際際滷Share a Scribd company logo

Cross channel attribution overview feb 2010

Download as PPTX, PDF
xplusone
xplusone

Forrester recognized [x+1]'s strong attribution modeling capabilities. [x+1] can perform various types of attribution analyses to optimize marketing spend, including: 1) Placebo analysis to understand cross-channel interaction and view-through impact beyond last click attribution. 2) Remarketing attribution analysis to assess relative credit of sites driving conversions. 3) Offline conversion integration to enable true business value analysis beyond online metrics. 4) Custom marketing mix modeling using techniques like conversion path analysis to optimize online channel mix.

1 of 21
Downloaded 101 times
[x+1] AttributionFebruary 2010
Forrester recently recognized our attribution modeling capabilities, which ensure we optimize to the right measures[x+1] impressed us with its full service offering which includes some of the richest algorithmic analytics of the group 2
Different questions at different scopes of attribution
Different questions at different scopes of attribution Assign credit to display networks and properties on your media plan in order to evaluate performance, or pay CPA bounties.
Different questions at different scopes of attribution Assign credit to online tactics..search, display, affiliates and social, in order to allocate budget more effectively.
Different questions at different scopes of attribution Optimize marketing investment and mix, by understanding interaction and synergy between online and offline tactics
Attribution Analysis SupportCross channel user interaction / conversion path data enablementPlacebo analysisRemarketing attribution analysisOffline conversion integrationCustom marketing mix modeling
DATA / REPORTING ENABLEMENT: All user level interaction data across online channels available for analysisCross-channel reporting includes:% of unique converters seen in each channel by conversion type
# of click and display events by channel leading to a conversion event
Drill downs by different timelags (1 day, 1 week, etc.)
Optional data transfer for more in-depth sequencing analysis88
PLACEBO ANALYSIS Case Study 1: Cross channel analysis for Financial Services clientBackground:Financial services firm was interested in understanding the true view through impact of brand vs. performance display advertising. Last click/view attribution was not telling the true story.A better approach:A Test/Control campaign was executed, including brand and performance advertising.
Using ad server log file data cross channel overlap reporting helped shed light on the paths that led to conversion.Results:Display ads drove significant view through conversion for as much as 21 days after exposure.
Performance ads drove a significant impact on conversions typically credited to Search.A new way to view the data drove marketing spend changes!
PI Lift for Performance & Brand Ads over PlaceboPerformance ads drive 8x more PI applications than the Placebo ads on the same day as exposure
For days 1-16 after exposure, Performance ads drive about 3x more applications than Placebo and Brand ads drive about 2x more applications than Placebo
Performance and Brand ad performance converges closer in application rate after day 16 PI Lift % = (Display Application Rate % - Placebo Application Rate %) / Placebo Application Rate %
Cross-channel placebo analysisLeap Frog results are estimates
OFFLINE DATA INTEGRATION enables true business value analysis
Display Optimization: Avoid false DarwinismThe default last click/last view approach creates incentives for bad behavior by all on the plan.Properties compete for credit by bombarding the remarketing pooland are dis-incented to BUILD the remarketing pool.Networks and portals have learned to game the system, using a variety of tactics to set the last cookietactics that do not drive sales. Most of the time, clients and agencies get it wrong
Remarketing analysisReport Uses Assess the relative attribution credit of conversions by site.

More Related Content

Cross channel attribution overview feb 2010

  • 2. Forrester recently recognized our attribution modeling capabilities, which ensure we optimize to the right measures[x+1] impressed us with its full service offering which includes some of the richest algorithmic analytics of the group 2
  • 3. Different questions at different scopes of attribution
  • 4. Different questions at different scopes of attribution Assign credit to display networks and properties on your media plan in order to evaluate performance, or pay CPA bounties.
  • 5. Different questions at different scopes of attribution Assign credit to online tactics..search, display, affiliates and social, in order to allocate budget more effectively.
  • 6. Different questions at different scopes of attribution Optimize marketing investment and mix, by understanding interaction and synergy between online and offline tactics
  • 7. Attribution Analysis SupportCross channel user interaction / conversion path data enablementPlacebo analysisRemarketing attribution analysisOffline conversion integrationCustom marketing mix modeling
  • 8. DATA / REPORTING ENABLEMENT: All user level interaction data across online channels available for analysisCross-channel reporting includes:% of unique converters seen in each channel by conversion type
  • 9. # of click and display events by channel leading to a conversion event
  • 10. Drill downs by different timelags (1 day, 1 week, etc.)
  • 11. Optional data transfer for more in-depth sequencing analysis88
  • 12. PLACEBO ANALYSIS Case Study 1: Cross channel analysis for Financial Services clientBackground:Financial services firm was interested in understanding the true view through impact of brand vs. performance display advertising. Last click/view attribution was not telling the true story.A better approach:A Test/Control campaign was executed, including brand and performance advertising.
  • 13. Using ad server log file data cross channel overlap reporting helped shed light on the paths that led to conversion.Results:Display ads drove significant view through conversion for as much as 21 days after exposure.
  • 14. Performance ads drove a significant impact on conversions typically credited to Search.A new way to view the data drove marketing spend changes!
  • 15. PI Lift for Performance & Brand Ads over PlaceboPerformance ads drive 8x more PI applications than the Placebo ads on the same day as exposure
  • 16. For days 1-16 after exposure, Performance ads drive about 3x more applications than Placebo and Brand ads drive about 2x more applications than Placebo
  • 17. Performance and Brand ad performance converges closer in application rate after day 16 PI Lift % = (Display Application Rate % - Placebo Application Rate %) / Placebo Application Rate %
  • 18. Cross-channel placebo analysisLeap Frog results are estimates
  • 19. OFFLINE DATA INTEGRATION enables true business value analysis
  • 20. Display Optimization: Avoid false DarwinismThe default last click/last view approach creates incentives for bad behavior by all on the plan.Properties compete for credit by bombarding the remarketing pooland are dis-incented to BUILD the remarketing pool.Networks and portals have learned to game the system, using a variety of tactics to set the last cookietactics that do not drive sales. Most of the time, clients and agencies get it wrong
  • 21. Remarketing analysisReport Uses Assess the relative attribution credit of conversions by site.
  • 22. Assess the effectiveness of the site to generate conversions on a weighted basis. Converter OverlapReport Uses Assess the relative attribution credit of conversions by site.
  • 23. Assess the effectiveness of the site to generate conversions on a weighted basis. Case Study 2 Attribution for multistage conversion processBackground:Entertainment company, sales funnel starts with a free trial and ends with a paid subscription.
  • 24. As company tried to scale campaign, cost per sale was increasing, with little subscription growth.Dynamics:To scale the campaign, client added CPA deals to plan.
  • 25. CPA providers bombarded the remarketing pool, gaining credit for last view. Non CPA providers lost share of sales and had their budgets cut as their credited CPAs went up.
  • 26. Remarketing pool shrank, sales flattened at a higher overall cost. A better approach:For each provider, measured unique contribution to reach overall and to remarketing reach.
  • 27. Rewarded trial drivers and reach providers and eliminated remarketing for all but one partner. New approach led to renewed subscriber growth!
  • 28. Proven online marketing mix modeling techniques are applied to drive full online channel optimization Three primary approaches:Conversion interaction analysis Goal impact of immediate response (click and immediate viewthrough) behavior on overall channel responses (view and organic conversions)Online Cross channel analysisGoal - decompose drivers of online conversion using more detailed display drivers (offers, promotions, etc) and additional online channels (search, affiliate)Online Conversion analysis tiered approachGoal - understand the detailed drivers of online conversions across marketing elements, both online and offline, and the interactions between them.17
  • 29. Conversion interaction analysis Immediate response multiplierWhat you get: A basic understanding of relationship between click-based, view-based and organic conversions to help in forecasting and planning identifying the multiplierRequirements: Differentiation in levels of display media execution Approach: Use regression to decompose display impact on organic and view-based conversion, controlling for high level market changesExample regression equationExample approachView through + Organic Conversions = constant + 硫1 * (PC conv.) + 硫2 * (lagged PC conv.) + 硫3 *(1 hr. PI) + 硫4 * (media = 1) + 硫5 * (week, month) + 硫6 * (site change) + 硫7 * ()18
  • 30. Online Cross channel analysisWhat you get: A basic understanding of drivers of online conversion using more detailed display data (offers, promotions, etc) and additional online channels (search, affiliate)Requirements: Cross channel tagging or log filesApproach: Use regression to decompose display impact on organic and view-based conversion, and other channels. Example regression equationApproachView through + Organic Conversions = constant + 硫1 * (PC conv.) + 硫2 * (lagged PC conv.) + 硫3 * (site change) + 硫4 * (media = 1) + 硫5 * (week, month) + 硫6 * ()Search or affiliate click conversion = constant + 硫1 *(PC conversions) + 硫2 * (lagged PC conversions) + 硫3 * (media = 1) + 硫4 * (week, month) + 硫5 * ()19
  • 31. Online Conversion analysis tiered approachWhat you get: A detailed understanding of drivers of online conversion and the interactions between them, including online and offline effortsChallenges: Collinearity , too strong of a read, no read, model breakdownApproach: Basic approach is similar to classic marketing mix analysis, time series regression. However, tactic interactions are explored in more detail, through interaction variables, synergy terms, or additional models (i.e. search or display as a dependent). Expanding further, multiple models are run exploring different parts of the funnel.Classic marketing mixExample regression equationsConversions = constant + 硫1 * (price) + 硫2 * (special offer) + 硫3 * (economic factors) + 硫4 * (TV GRPs) + 硫5 * (Print TRPs) + 硫6 * (Search) + 硫7 * ()Search traffic = constant + 硫1 * (economic factors) + 硫2 * (TV GRPs) + 硫3 * (Print TRPs) + 硫4 * ()20
  • 32. Attribution Analysis Support OverviewCross channel user interaction / conversion path data enablementPlacebo analysisRemarketing attribution analysisOffline conversion integrationCustom marketing mix modelingAttribution analyses can lead directly to optimization of online marketing programs through POE