how do you actually find out why users drop off
i have been using posthog been staring at dashboard/heatmaps for two days and i think im losing it it tells me 40% of people quit on step 3 of onboarding. cool. thanks. but nothing anywhere tells me why they quit. watched like 3 hours of session replays. you…
i have been using posthog been staring at dashboard/heatmaps for two days and i think im losing it it tells me 40% of people quit on step 3 of onboarding. cool. thanks. but nothing anywhere tells me why they quit. watched like 3 hours of session replays. you see the cursor move around, you see them scroll, you see them close the tab. the recording shows everything except the actual reason. emailed a bunch of users including the paying ones. two replied. one of them said something genuinely useful and it took him 30 seconds to say it. that 30 second reply was worth more than the entire month of dashboards. but you cant do that at scale and most people just dont answer. so how are you actually doing this. do you have a real method or is everyone just guessing and calling it a hypothesis im starting to think its the second one
Collected discussion
In a lot of the products that I worked with, we typically had a user churn workflow or a journey. Here, you would actually solicit feedback from the user on why they have decided to stop using your product. Maybe you can trigger this automatically and have an AI agent on the back of it to take them through a questionnaire. Honestly, I feel doing this with a person is going to give you a lot more benefits, not only in terms of insight but also potentially being able to avoid the churn or try to get the user back to your platform. One easy way to do this is probably just to subcontract to a caller or to a customer success rep through various platforms that exist out there. And of course, this depends on the ticket size of each customer.
The reason that one email worked and the rest didn't is probably just timing. You asked days later, when whatever confused them had already stopped mattering to them. What got me actual answers was asking inside the product at the moment they bail. One box, free text, "what were you trying to do here?" No multiple choice, because with multiple choice you're just handing yourself back your own guesses and then feeling good about the data. Reply rate on that is way higher than email since they still have the thing in their head. Also you're probably watching the wrong replays. Filter to the sessions that died on step 3 with a long idle pause before the tab closed. That pause is usually them reading something that didn't answer their question. One more, and this is the one that took me ages to notice: check whether step 3 asks for something they don't have on hand yet. An API key, or someone else on the team who has to approve it. That looks exactly the same as confusion in a replay, because there's nothing to see, they just go off to get the thing and don't come back. Worth pulling whether any step 3 quitters come back a day or two later. If a chunk of them do, it's not really a UX problem and you can just let them skip it. No idea if that's your situation though.
You're not crazy. Replays are great for showing where people stall and terrible for explaining intent. What worked better for us was pairing the drop-off event with one tiny in-product question right there: "what were you trying to do?" plus a skip-for-now path if that step asks for something they may need to go fetch. A surprising amount of "bad UX" is really "I need one more thing before I can finish".