Proceed With Caution: Attribution's Insights Not Always Straight Line

I was recently on  vacation in a particularly mountainous region in Asia and needed to drive from one resort town to another.  Taking the wheel of the tiny rental car I had to squeeze myself into, I headed up the steep, narrow roads -- totally devoid of guard rails -- that were fraught with blind curves irregularly marked with signs that read “Proceed With Caution.” As I carefully weaved my way around each mountain-hugging curve, I couldn’t help but notice that as each new piece of road revealed itself to me, some were exactly as I expected them to appear (from the other side of the curve) and others were totally surprising and unexpected.

What You See Is Not Always What You Get

In much the same way, the insights that are produced by your cross-channel attribution management solution, and the resulting optimization actions that you take, may at first glance appear to be very straightforward in terms the results they produce. But often, these insights reveal results that are totally surprising and unexpected. 

For example, after attributing all the appropriate credit for conversions and return on advertising spend from all your marketing tactics across channels, you find that particular combination of publisher, display ad, creative and size produces a particularly attractive conversion rate, cost per action (CPA) and return on advertising spending (ROAS) results.  So you decide to double your investment over the next month in that particular combination, anticipating that you double the resulting revenue at the same CPA and the same ROAS.  Perhaps you will, but you probably won’t.

Other Factors at Work

 Just as there was undoubtedly a logical reason behind why the mountainous road I was navigating was engineered to take the twists and turns it did, there are valid reasons for this lack of accurate predictability when trying to reproduce previous results that were attained at different spend levels:

  • The publisher has a limited reach that might be exceeded by the expanded spend level.
  • The spend level might mean the ad is shown at a different frequency.
  • The spend level might exceed the publisher’s available inventory.
  • There may be a minimum spend level to start seeing any results.

Each of these factors could cause the ad to be shown to the target audience you hoped to reach with your expanded spend a different number of times, which would have a direct impact on the results that spend produces.

Optimize With Caution

That said, what’s a marketer to do?  Isn’t attribution supposed to provide actionable insights that enable marketers to predict the results of those actions?

 The answer to that question is still yes.  But the degree of accuracy of the output of your attribution-based recommendation engine is directly tied to the data collection process and optimization factors used by your attribution management solution.  If properly done, the predictive analytics functionality that drives those recommendations will be able to take into account the factors above and accurately predict the results you’ll attain at new spend levels.  If not properly done, the solution will simply assume the same CPA, the same ROAS and will double the predicted revenue -- and will likely disappoint you when the actual results are revealed. You should discuss the methodology used by attribution management vendor and how their solution takes these factors into account.

Knowing what’s coming around the next curve allows you to loosen your grip on the wheel, relax a little and enjoy your journey more.

Tags: metrics
Recommend (15) Print RSS