Post-view is the least popular metric in online marketing, for two opposite reasons. Some consider it fraud, because a seen ad could explain any later purchase. Others rely on it blindly, because it justifies the display budget. Both are partly right.
The legitimate objection
A wide post-view window explains everything. Look back 30 days and count every impression and you will find a viewable impression for practically every conversion, including the ones that had nothing to do with advertising. That is no longer measurement but a justification with a timestamp.
On top of that: an “impression” does not mean anyone saw anything. A creative can be served without ever appearing in the visible area. Equating delivery with viewability measures the ad server, not the person.
The legitimate counter-objection
Last-click gives all the credit to the last click, as if it alone counted. That falls short where an ad sets things off without a click: if someone buys through search days after seeing a display ad, last-click credits the purchase to the search ad alone. The display ad before it gets nothing.
How to recognise post-view measurement you can check
- Viewability rather than delivery: what counts is what was, by default, at least half in the visible area for one second, measured with our own tag, not the ad server’s word for it.
- A short window: one day by default. A viewable impression is a weak signal; the longer the window, the more conversions receive a share without the effect being any better evidenced.
- Click before view: where a click exists, the click counts. A viewable impression only comes into play when there is no click.
- Reported separately: the post-view share stands as its own figure, not stirred into the total. Anyone who does not believe it can subtract it.
- With coverage: a viewable impression only reaches the purchase if the browser keeps the view identifier. Chrome and Edge do, Safari, Firefox and Brave do not. The linkable share sits next to the post-view share so that nobody mistakes a three per cent share for a three per cent effect.
How to check it
The test is simple: put the same period side by side, once with and once without post-view. If the result moves a lot, your decision rests on an assumption, and that is worth knowing before you shift budget. If it barely moves, your decision does not depend on post-view, and this argument settles itself.
Idea clave
A model is an assumption, not a truth. The difference is not whether you make assumptions, but whether you name them.