PostHog's marketing team put a pineapple poll on a job listing and voted 60% yes. I agree with them, which left nothing to argue about and one obvious job undone. If pineapple belongs, where do you actually go for a good one? I went through nine highly rated Toronto pizzerias, then built the answer as a live PostHog demo.
PostHog put a pineapple poll on a job listing, and the marketing team voted 60% yes. That tells you something about how they hire, so I took it seriously and went to look at the data.
They are right, so there is no argument to have. What is missing is the follow-up. Knowing where you stand on pineapple is useless if you do not know where to go for one.
Nine highly rated Toronto pizzerias, weighted by review volume so a five-star place with forty reviews did not outrank one with seven thousand. I used AI to scan their menus and review text for mentions of pineapple, then reviewed the results myself.
This was a search rather than a menu-by-menu audit, and these are not the definitive top nine. Two returned a clear positive result. For the rest, the absence of evidence reflects the limits of the method rather than a result.
Two of the nine came back positive. The interesting part is not that they serve it, but what they use instead of ham. Both landed on the same answer, separately.
The full ranking is on its own page, built to look like a dashboard. Every click on it is a real event in a real PostHog project.
Live page The Pineapple Question Nine pizzerias, two clear matches, and a poll you can vote in →The pizza is the excuse. Three PostHog features are running live on the demo, because using a product is a better argument than writing about it.
Nine restaurants is a small sample, and review text is a rough proxy for a menu. This is a product demonstration, not food research. The point is the instrumentation and the decision you can make from it.
When a question is already settled, running the poll again with more people does not help. Asking the next question does.
Work through the remaining menus properly, since that is the weakest part of this and the cheapest to fix. Run the headline test long enough to reach significance instead of calling it early, which is the mistake I see most often in A/B tests. And add a follow-up question for anyone who picks the third option, since they are the only group whose mind is still open.