How to leverage logs to analyse ChatGPT Ads, IAB Tech Lab's AAMP 3.0 and "performance CTV". Here's the four-question test to run before the money moves.
The TV platform underreporting its own performance is the example that stuck with me. Everyone assumes vendor dashboards flatter the vendor, so a number that runs the other way probably gets ignored before anyone checks the raw data.
I was thinking the exact same, Raksham. My assumption is always that platforms are set to serve the most favourable possible stats. This shows the value of log audits as how many marketers would have turned the TV campaign off?!
Raw logs over dashboards, always. With ChatGPT ads I'd also watch the answer sitting above the ad. That probably moves results too, and no report will show it.
Good point, and it's a new version of an old problem. With display we argued for years about whether the page around the ad helped or hurt, and at least the logs told you which page it was. With ads under an AI answer, the answer is the context, it changes every time, and as far as I can tell nobody logs it. If the answer has just recommended a competitor, your ad underneath is working uphill and the report will never show it. Thanks Kedar.
Working uphill is the right way to see it. With display you could at least pull the placement report. Here the context is gone once the answer changes. Thanks Mike.
The TV platform underreporting its own performance is the example that stuck with me. Everyone assumes vendor dashboards flatter the vendor, so a number that runs the other way probably gets ignored before anyone checks the raw data.
I was thinking the exact same, Raksham. My assumption is always that platforms are set to serve the most favourable possible stats. This shows the value of log audits as how many marketers would have turned the TV campaign off?!
Raw logs over dashboards, always. With ChatGPT ads I'd also watch the answer sitting above the ad. That probably moves results too, and no report will show it.
Good point, and it's a new version of an old problem. With display we argued for years about whether the page around the ad helped or hurt, and at least the logs told you which page it was. With ads under an AI answer, the answer is the context, it changes every time, and as far as I can tell nobody logs it. If the answer has just recommended a competitor, your ad underneath is working uphill and the report will never show it. Thanks Kedar.
Working uphill is the right way to see it. With display you could at least pull the placement report. Here the context is gone once the answer changes. Thanks Mike.