Algorithmic Attribution – If It Was Easy, Everybody Would Be Doing It

Attribution analysis – measuring the value that each marketing impression contributes to a campaign goal – is hard to get right. In fact, it’s so hard that a recent study commissioned by the Interactive Advertising Bureau (IAB) and executed by Forrester Consulting found that 44 percent of interactive marketing executives use no attribution method whatsoever, and an additional 30 percent use the first-or-last click method (in spite of the fact that few marketers, agencies, or publishers stand behind that approach).

There are many reasons that getting attribution analysis right is hard – here are a few important ones:

  1. Missing data. It’s difficult to get the marketing data required for attribution out of the social media, search properties, and digital devices that consume much of our attention, and this creates significant blind spots for attribution analysis.
  2. Complex technology requirements. Large data volumes require the use of modern big data platforms. And the talent required to unlock the marketing value in that data is scarce.
  3. Dynamic campaign success factors. The value of different marketing channels vary by campaign, so every attribution model (whether determined algorithmically or via simple rules) must be unique to each campaign on which it is being applied.

But the fact that attribution analysis is hard is no excuse not to do it, and smart marketers recognize that some measurement is better than no measurement. The benefits are clear – better spending, channel, and strategy decisions lead to dramatic increases in campaign performance. But perhaps most importantly, getting attribution analysis right is the only way to close the direct-response loop on a multi-channel campaign in order to utilize the predictive analysis techniques that truly move the needle on campaign performance for interactive marketers. For example, WPP’s Media Innovation Group reports that better attribution analysis enabled one of its large banking clients to maintain a 25 percent conversion rate with 33 percent lower CPA on its multi-channel campaign. A detailed case study can be found here.

You can achieve similar success by utilizing the following phased approach:

  1. Invest in a big data platform that delivers the scale and performance required to capture and analyze every touch point created across every channel for every campaign you execute. Recent technology breakthroughs in massively parallel processing for data analysis have made this much easier than it used to be.
  2. Walk before you run – keep your initial deliverable as simple as replicating the best of your existing reports (which are most likely last-click based) for a single marketing channel.
  3. Now add data from an additional channel and use a simple rule-based weighting to assign credit across the two channels now tracked in your big data platform. According to the aforementioned study, this milestone puts you ahead of 74 percent of your peers, and this enhancement to last-click attribution will almost certainly improve your campaign performance.
  4. Now add data from each of the remaining channels in your campaigns, adjusting your rule-based attribution weightings appropriately.
  5. Finally, replace your rule-based attribution weightings with an algorithmic approach that takes advantage of the closed feedback loop you have created to evolve your attribution model based on historic results. In other words, leverage machine-learning techniques to optimize your attribution model for each campaign.

According to the IAB study, only a scarce 11 percent of interactive marketers achieve this level of sophistication in leveraging algorithmic attribution models. But these rare marketers create tremendous value for their firms. And in addition to that, marketers with this experience are among the most highly sought after – and highly compensated – marketers in the world. The payoff is clearly worth the effort.

Related reading

Overhead view of a row of four business people interviewing a young male applicant.