• by Jason Burke , October 4, 2011

In the waning moments of Game 7 of the 2011 Stanley Cup Finals, Brad Marchand, a wing on the Boston Bruins' second line, scored an empty net goal to make the score 4-0. It was a great moment for Marchand, but no analyst or expert was giving him sole credit for the Bruins' victory.

Why then, in online advertising, do we reward a full commission to the media partner whose ad contributed the infamous "last click" before conversion? It's been common practice for years, and it needs to change.

"Conversions" are the instances where a desired marketing goal is met. They are valued greatly because advertisers invested money to reach those goals, and understanding the catalyst for each conversion is necessary for marketers to understand where to allocate budgets and appropriately compensate the publishers that seeded the conversion.

Conversion attribution describes the various factors that should be credited with contributing to the ultimate conversion activity. Just as a mutual friend is often thanked for introducing newlyweds, we should acknowledge what triggers the ultimate goal of the advertiser in the form of conversion payout. Unfortunately, conversion attribution is riddled with challenges:

Challenges in Video Conversion Attribution

Multiple causal metrics: Conversions that are the result of multiple interactions with a brand are common across all types of online advertising, but more prominent in video, where there are multiple opportunities for engaging touch points. When you layer on offline and search exposure, then multivariate analysis is necessary to understand which events actually drove the conversion.

Frequency and decay of exposure: Ad technology platforms recognize that varying the number of and the time between ad exposures has a diverse impact on these signals. Platforms must assign appropriate values using these metrics.

Widespread use of "last ad" attribution methodology: The "last ad" attribution methodology, which credits the last click or final impression prior to the conversion, harms the ad impressions early in the chain that result in future conversions. Viewers come across multiple contributors before the actual conversion takes place, leaving the initial publisher without any credit for the conversion. There is a similar model for "first ad," but both of these overly simple strategies focus on a single advertising event at the bottom of the marketing funnel rather than looking at the entire series of events that lead to the conversion.

Video advertising is focused on upper-funnel metrics: Video is recognized as a strong branding tool for marketers, focused on metrics like awareness and intent that are further away from normal conversion metrics like purchase. As a result, most conversions, as currently measured, are inappropriate for video advertising. This mismatch makes video publishers struggle with conversion rate optimization.

Steps to fix video-based conversion attribution

Different conversion goals for video versus display and search: Display and search are great vehicles for lower-funnel goals like purchase and direct response, but video is built for branding. Branding-focused marketers must think about conversions that raise awareness and shift attitudes about their brand, then track these separate from lower-funnel goal conversions.

Multiple attribution: There is no one-size-fits-all attribution strategy, and it's impossible to uncover all possible conversion contributors. However, recognizing that there are multiple contributing events and learning about the correlations between those causal metrics and the ultimate conversion helps ad technology value and fairly compensate each of the various events in the conversion chain.

How should the commission be distributed for a conversion that follows a user who views a video ad on site A, then watches that same video ad to completion on site B and later engages with the brand's video ad on site C?

Various strategies exist to deal with this problem. Distributed multiple attribution allocates equal amounts of the conversion bounty to each event, but leaves the more valuable events undercompensated. Algorithmic multiple attribution uses statistical analysis to fairly distribute the conversion value across the various activities in the chain. Even for those advanced ad platforms that use educated analysis and econometrics to build an attribution model, it is a non-trivial task and there is certainly not a global silver bullet to shoot down this problem.

Recognizing that video has a wider variation of causal metrics: Google's Floodlight product has a default attribution algorithm that values a click over an impression, as the click is thought to have a stronger impact on conversion. In the search and display world, those two metrics constitute the entire list. Video goes further with engagement, time spent, and other video-specific metrics that contribute to the conversion in varying degrees.

Proper conversion attribution has plagued the advertising world since the birth of the Internet, and with an ever-increasing number of vehicles for digital advertising, the challenges continue to grow. All sides of the advertising equation must understand that there is no single solution for solving this issue, and these challenges are even more pronounced in video advertising. Advertisers must work with their platform vendors to ensure that everyone in the mix gets their due.