If you’re a brand advertiser this make total sense, as viewability is a great proxy for quality inventory. However for a performance campaign, total viewability as a goal is neither logical nor effective.
Well-meaning supporters of viewability KPIs might partly base their argument on the correct belief that post-click measurement models don’t show correlation between clicks and conversions (as my previous display click myth article demonstrates).
Similarly, post-impression attribution – which has been suggested as an alternative – has major problems.
It’s not sensible to infer a link between impressions and behaviour regardless of whether or not that ad was in view at any point.
So viewability sounds like the answer. After all, no one should be denying that viewability correlates considerably more strongly with conversions than clicks do.
Should we all be moving away from cost per acquisition (or CPA) as a KPI and focusing on the percentage of ads seen?
The flaw in this thinking
Well, this is where viewability can run into problems. Although higher viewability correlates with a higher conversion rate, it doesn’t necessarily correlate with a lower CPA.
Inventory with high levels of viewability, the kind of inventory a percentage viewability target would lead you to optimise towards, tends to be more expensive than inventory with lower viewability rates, leading CPAs to be higher on the better quality inventory, especially if using a post impression or post click attribution model.
For most advertisers, this simply isn’t an option – viewability is not cost-effective enough as a KPI.
You can see from the example below the impact that optimising to viewability can have on the cost of the inventory:
What’s the solution?
It can feel like we are floating between KPIs, none of which give the client what they are actually looking for.
They’re looking for a reliable way to identify which activity is actually increasing their return on investment (ROI) and which activity is simply ‘cookie-bombing’, piggy-backing off other marketing work.
A good interim measure is the post-viewable-impression CPA.
While it’s not perfect – it doesn’t prove intent and isn’t currently supported by DCM, Google’s adserver and the most popular adserver across the industry – it offers a middle ground between post impression and post click, not attributing conversions to impressions that went unseen, but still avoiding the skew towards poor inventory prompted by a post-click CPA.
The above graph is a good example of how this can affect performance. A client asked us to work on a metric that would value display more accurately than the post-impression and post-click models we had previously used with them.
After identifying the impact of viewable frequency on conversion rates, we suggested moving to a viewability goal for the campaign.
While this did improve our overall conversion rates, it also increased CPMs by 196%, effectively counteracting the impact of the increased conversion rates.
As a result, we suggested moving to a post-viewable-impression CPA model.
This resulted in a drop in overall viewability (almost back to the levels we saw before changing KPI), but a corresponding drop in post-viewable-impression CPA, which had been extremely high before the move in KPI:
This, essentially, is the advantage of a post-viewable CPA: by attaching a cost metric to the KPI, it avoids the inflated CPMs you often see when optimising purely to viewability, allowing advertisers to work to a metric that reflects their commercial reality.
So instead of challenging your media buyer to only buy highly viewable impressions, your goal should be to prioritise a KPI which actually delivers a real return from your performance campaigns, rather than a well-meaning, but perhaps ineffective, KPI that purely values viewability.