What are ghost ads?
How do ghost ads work?
To measure what ads add, you need a control: people who did not get them. The question is what the control group sees instead.
In a PSA test, part of the control group is shown a public service announcement, a non-branded ad, in place of yours. A mobile ad tech vendor's blog points out that you pay to serve those non-branded ads3. Criteo lists the same cost and adds that results may be biased1.
With ghost ads, the system flags each moment the control user would have been served your ad3. Whatever else wins the slot is shown. You compare results for those flagged control users with the matching users who saw your ad. You pay for no stand-in ads.
Because the system must know which slots your ad would have won, it is run by whoever runs the auction. Criteo says ghost ads only work in closed environments where the whole auction is visible1. It is not something you can reproduce from your own sales data.
What is ghost bidding?
A related method described by Criteo, an ad tech company. A bid is placed as usual for the treatment group. For the control group, the system tracks whether the bid would have won the auction, and if so records a "ghost impression"1. The idea is the same: record who would have been shown the ad, without showing it.
Worked example
Easily confused with
Ghost ads vs PSA test
A method in which control users see a public service announcement instead of your ad. A different method, not a synonym.
Ghost ads vs Intent-to-treat design
A design in which everyone assigned to a group is counted, whether or not they actually saw the ad.
Ghost ads vs Conversion lift study
A platform-run experiment that splits users at random into a group that can see the ads and a control group that cannot. Ghost ads are one way to build a control group in a study like this.
Ghost ads vs Geo test
Splits by region, not by person. Ghost ads work at the level of individual users.
Back to the bigger idea: incrementality, the extra sales an activity causes beyond what would have happened anyway.
FAQ
Are ghost ads the same as a PSA test?
No. A PSA test shows the control group a public service announcement. Ghost ads log the missed impression and show a different ad.
Do ghost ads prove my ads work?
They are a way to build a fair comparison. The result is only as good as the test design around it.
Can I run ghost ads myself?
Usually not. Whoever runs the auction has to record the impressions your ad would have won. Ask your agency who runs it.
Sources
Let’s find what’s actually working.
Next step. Before you design any test, check what your current ROAS assumes. Use the Is your ROAS real? calculator.