You send the email to a thousand users. Got fifteen minutes to tell us how it’s going? Grab a slot. Forty people book. You run the calls, you take the notes, and a picture starts to form — and it’s a striking one. Half your interviewees are effusive: this changed how they work, they’ve told their whole team, don’t ever remove the thing they love. The other half are done: it’s slow, it broke on them once, they’re already looking at a competitor. Two camps, drawn in bold, with almost nobody in the mild middle.
It feels like a finding. Your users have strong feelings and they’ve split into believers and skeptics; now you know the battle lines. Except you don’t. What you’ve actually discovered is who is willing to give a stranger fifteen minutes — and that turns out to be a very particular kind of person.
The calendar builds your sample, not you
Every user you interview had to clear the same bar first: read the email, decide their opinion was worth an appointment, find a slot, and show up to it. That bar is invisible in your notes — the transcript looks the same whether the person crossed a high fence or a low one to get there — but it is doing more to shape your results than any question you asked. You didn’t choose your sample. The willingness to book chose it for you.
And willingness to book is not evenly distributed. It pools at the extremes. The person whose week you genuinely saved wants to say so; gratitude looks for somewhere to land. The person you let down wants that on the record too; frustration is a strong motivator to show up and tell someone. Both ends have a reason to spend the afternoon. The bar filters for feeling, and it lets through exactly the people whose feelings run hot.
This is the same trap as the product review page — five stars and one star, a hollow trench where the threes should be — and it has the same cause. Nobody with a three-star afternoon fills out the form. The bimodal shape isn’t a fact about the product. It’s a fact about who bothers to respond.
Two camps is a shape recruiting makes
Here’s the part that should worry a research team: the two-camps picture is self-confirming. You go in braced for strong opinions, you recruit the people with strong opinions, and they hand you strong opinions. The method quietly manufactures the result it expected, and the result feels like evidence because it came from real users saying real things. Every quote is true. The sample they were drawn from is a lie.
A distribution with two peaks and an empty middle is not a portrait of your users. Nine times out of ten it’s a portrait of your recruiting funnel — the two feelings strong enough to book a call, and the silence of everyone else.
So the roadmap gets built for the two camps. You ship features to delight the evangelists, who were already going to stay, and you fight fires for the detractors, who were already halfway out the door. Meanwhile the largest group of users you have — the ones for whom the product mostly just works, with a couple of quiet frictions they’ve never mentioned to anyone — go on getting nothing, because they never appeared in the research at all. You can’t build for a group you’ve never heard from, and you never heard from them because you made hearing from them cost an afternoon.
The middle isn’t quiet. It’s just busy.
It’s tempting to read the silence of the middle as indifference — if they had something to say, they’d have booked. But that gets the causation backwards. The middle is silent because the price of speaking is a scheduled call, and a scheduled call is precisely the thing a moderately-satisfied person will never pay for. They’re not apathetic. They’re doing their job. Their product experience is a means to an end, not the end, and “the tool is fine, here’s a small thing that bugs me” is never worth a calendar invite to a busy person who mostly just wants to get back to work.
Which means the middle is holding the most actionable feedback you’ll never see. The evangelist’s praise is pleasant and mostly un-actionable. The detractor’s fury is often about a decision you made on purpose. But the moderate user’s small, specific, un-dramatic friction — the export that takes one click too many, the setting they can never find — is the stuff you could actually fix on a Tuesday, and it lives entirely inside the group with no reason to book.
What changes when answering is free
The whole distortion traces back to a single number: the cost, to the user, of answering a question. Drive that cost toward zero and the filter that built your two camps stops filtering. This is the bet under AI User Interviews — and it’s a different bet than “run more interviews.” More calls faster is a throughput win; it doesn’t fix who is on the calls. What fixes who is on the calls is removing the call.
When an interview is a conversation a user can have on their own time, in their own words, with no slot to book and no stranger waiting — an AI interviewer that asks the real follow-up without leaning on the answer — the moderate majority can finally afford to respond. Not because you convinced them their opinion matters, but because you stopped charging them an afternoon to share it. The two camps don’t disappear; the fans are still fans and the furious are still furious. They just shrink back to their true size, as the tails of a distribution whose real mass was the middle you were never reaching.
That’s the shape you actually wanted all along — not two peaks and a trench, but your users. The loud ends were never wrong. They were just never the whole story, and the rest of the story was sitting one un-booked call away.