Why Paid Media Analytics Are Usually Misleading
Digital media has been growing its share of advertising dollars compared to traditional media for decades now largely because 1. increasing digital consumption and 2. ability to cleanly report on results via cookie tracking/ID pairing. Advertisers love being able to point at a click chain compared to something like a billboard where you can’t tell who has even seen it.
At some point, we swung so far into direct response land that we started avoiding non-DR media. With some of the new privacy changes, we cannot rely on that tracking as much anymore, and ironically that has forced advertisers to adopt better analytics and rediscover channels that chronically were undervalued.
Here is the main problem. Cookie tracking.
Some channels drive value that does not show up in channel reports and others grab attribution for conversions that would have happened regardless of seeing an ad. Cookie tracking cannot tell the difference between the two. We create a control group and experiment group to isolate the performance of the ads.
Example 1. Undervalued or Halo Effect
Here is an example of Linkedin ABM where we had poor performance in the channel but when we isolated the total value of the ads we see much more conversions and a better CPL. Users are not clicking but they are sharing and engaging.
We set up a test design to isolate the difference between seeing an ad and not seeing an ad. The green box is the true number of conversions driven. We then compare that to the reported conversions in LinkedIn.
In this example, CPA of LinkedIn from true performance moved from $680 to $120
Example 2. Overvalued/Non-Incremental
The opposite is also true. Here is a Facebook prospecting campaign that had no retargeting exclusions and was optimized toward including 1-day view-throughs. Only 2 percent of the reported conversions occurred because of the ads! This means the algorithm is showing ads to people who would have already converted.
The Main Point
Ok so what’s the takeaway? Some channels are going to generally overreport, some will underreport unless you build a testing structure to isolate ad performance. Just looking at how many clicks turned into conversions will give your program a limited view of what is really happening.
Notes
•Google has a suite of products that ultimately removed the ability to control bids, queries or audiences. Smart shopping, Smart Display and UAC being the most prominent.
•Native automated bidding solutions do not consider incrementally and therefore will push branded search, view through conversions and past site visitors
•Anything with view-through conversions are prone to exaggerate their efficacy.
•YouTube without exclusions is particularly non-incremental as attribution defaults including VTs in conversions
•Branded and Non-branded Search includes any channel with a search function. Google, Bing, Amazon, Apple, etc.
•Facebook Prospecting was particularly overvalued but with iOS removing conversions and Facebook dropping 28 day click attribution it is much more even.
•NB conversions are all very incremental but there is not much of a halo effect..
