It was about a year ago that a seemingly innocuous picture of a striped dress nearly broke the internet. People practically came to blows over whether a dress originally posted on Tumblr was blue and black—or white and gold.
No matter what color combination people saw, they were confident in their version of reality. Variations in perception and biologic bias produced dramatically different ‘truths’. Without conscious thought, your brain decided what spectra of light was hitting your retina and produced the sensation of some color. You knew the color of the dress, and there was no argument. That system works fine when everyone perceives the same thing. But, when two people look at the same thing and arrive at two contrasting conclusions—issues arise. This same effect is occurring now in digital advertising’s viewability and performance measurement.
As with the dress, you can have two completely correct, yet opposite performance measurements. Campaigns are frequently managed toward last-touch cost-per-action (CPA) goals, but as viewability becomes a common performance metric, you may end up with situations where your best performing partner for viewability is your worst performing partner for CPA and vice versa.
Viewability perception vs. reality
Let’s look at how what we perceive or experience as viewability differs significantly from the mechanical reality of a given page or environment.
As users engage more heavily on their phones and tablets, top publishers have optimized toward mobile devices. One common solution to mobile web surfing is endless scrolls. Content is continuously loaded as the page scrolls, as we see on Facebook and Twitter. Unfortunately, that scrolling we’re all so fond of can scroll us right past ads, pushing them above the screen before they’ve been tracked as viewable. Some websites with high quality inventory (ESPN, Bloomberg) have adopted this page structure and have potentially made themselves look worse from a vieawbility standpoint than websites that just cram ads above the fold on page load.
When iOS 7 was released it had the capability of pre-loading the top hits from search results, Google Chrome similarly has the ability to pre-load predicted links. These features are great for web surfing speed but they also mean all of those ads have zero per cent viewability until someone clicks on the predicted link. The solutions hurt websites with high search results or frequent link clicks (think blogs), because a significant percentage of their ad content is loaded without audience knowledge.
The programmatic viewability situation
For algorithmic programmatic buys (DSPs), viewability and last-touch CPA metrics may oppose one another. If a DSP is optimizing to last-touch CPA and a user visits a website with two ad slots, one above the fold in a highly viewable ad slot and one at the bottom of the page, the optimal choice may be surprising.
Here, two factors are at play to increase the value of the bottom ad slot impression. First, more viewable placements are often highly targeted by vendors specifically to increase viewability rates. That means that above-the-fold ad slot may have a higher predicted fill CPM making it less attractive to an efficiency obsessed algorithm.
Additionally, if both ad units are purchased, the second unit at the bottom of the page will often load after the top-of-the-page unit. That means during last-touch attribution the bottom unit will steal credit from the top-of-the-page unit. Smart algorithms will see this pattern over the thousands or millions of conversions and actively target bottom-of-the-page placements because they have higher conversion rates.
This industry arena has a number of perception issues to counter—everything from those mentioned above to additional issues, even around the concept or elusive promise of “guaranteed viewability.” That in itself could be an entire article.
All of this means the advertising equation is rarely as simple you’d hope. Only paying for viewable impressions sounds like an entirely reasonable request to make of the sell side, and publishers could deliver on ‘guaranteed viewability’, or at least only bill on viewable impressions. However this almost certainly would not improve the effectiveness of their digital media dollars if you’re still measuring a last-touch CPA. Collectively, we have issues to resolve on the viewability proposition. The simplest route is this: just track viewable conversions and optimize toward a singular metric that collapses other problematic metrics.
In many ways, asking vendors to optimize towards both viewability and CPA is as impossible as seeing the dress as both color schemes (I see white and gold). If you combine the metrics, this perception problem disappears and vendors can optimize more easily to a more actionable result.
Owen Ridolfi is director, analytic products at Flashtalking, Inc.