The feedback you actually need comes from the person who almost gave up on your product
Had someone tell me they nearly closed the tab in the first ten seconds, the homepage confused them that much. Instead they stuck around and told me exactly what tripped them up, and that one note led to the clearest improvement I've made in weeks. Not from someone who loved it instantly. From someone who almost didn't give it a chance at all.
There's a real bias in how most of us read feedback. The comments that feel good come from people who got it immediately, nice to hear, but they rarely teach you anything, because that person was never confused to begin with. The uncomfortable comments, from people who almost walked away, are the ones actually pointing at the exact gap between what you built and what a stranger experiences walking in cold.
Easy to brush past those in the moment because they don't feel like validation. But they're the closest thing you'll get to someone narrating their real first impression back to you, and that's something nobody who already understands your product can give you.
Has a "I almost gave up on this" comment ever ended up being the most useful thing someone told you? What did it change?
Replies
the thing i would add is that you only ever hear this from the near quitters who stayed. the ones who did close the tab at ten seconds had the same experience and left no note, and there are more of them. so it is the best feedback you can get and it still comes from someone who had a reason not to leave.
which is why i would not wait for it to arrive. in support the equivalent signal is repetition. someone asking the same thing twice, or filling in the same form again, is confusion you can see without anybody volunteering it. that you can go and look at today, and it points at the same ten seconds.
yours is the better version though, because he could say what specifically tripped him up. the repeats tell you where, never why.
when he told you, did you change the thing he named or the thing you thought he meant? i have got that wrong more than once.
@jernej_jan_kocica That last question is the sharper version of the whole point, changing what he named versus what I thought he meant. I fixed what he named, mostly because I didn't trust myself to guess right, and the repetition signal you're describing is probably the check I should be running anyway to confirm the guess would've been wrong or right after the fact.
@terranicgerald fixing what he named first is the right order. guessing is how you end up rewriting something nobody complained about.
one thing the repeats give you that his note does not: they tell you whether the fix worked. if the same second attempt keeps showing up after you changed it, then you fixed the thing he named and not the thing that was actually stopping him. that is the case where one clear piece of feedback sends you confidently in the wrong direction.
yeah, had one of these recently. someone testing our onboarding said flat out they almost bailed because they couldn't tell within the first minute whether the thing was actually doing anything or just spinning. that phrasing stuck with me way more than any generic "looks great" comment would have, because it named the exact moment of doubt instead of the general vibe. ended up adding a visible status step for that reason, something a happy user would never have thought to ask for since they weren't confused in the first place.
the near churn person paid the cost of confusion and chose to help anyway. they earned the right to be trusted more than the person who never hit friction. the harder move underneath your point is getting them to tell you the thing at all. most near churn users leave silently. the ones who write to you or answer a call are either weirdly generous or already have some relationship with you. that relationship is the actual asset. teams that build one channel where a specific human on the team is reachable by name (not 'support@'), and treat every one of those messages as the highest signal in the queue, get the near churn feedback loop the rest of the industry keeps pretending it can extract from surveys.
Had this during onboarding testing. Someone almost dropped off because the import step looked mandatory, there was no visible way to tell you could skip it and come back later. They stayed and said so directly. That one sentence was worth more than a dozen "looks clean" comments, because nobody who already understood the flow would have even noticed the ambiguity was there.