Stop Asking Users What They Want

Live in the Discomfort

Stop Asking Users What They Want

When I think back on the most valuable customer discovery I’ve ever done, none of it happened behind a desk.

At Zendesk, when we were working on Zenly, we used to take our prototypes straight to the street. No script, no formal testing setup. Just us, an iPhone build on TestFlight, and a stranger who fit our target demo (usually someone between sixteen and twenty-five). We’d hand them the phone and say, “Try it. Tell us what you feel.”

Then we’d film everything.

It sounds chaotic, but those unfiltered reactions (the confusion, the hesitation, the small flashes of delight) taught us more than any polished usability lab ever could. Watching someone struggle through your onboarding is deeply uncomfortable. That’s exactly why it works.

The Discomfort is the Point

Most founders want validation. What they need is friction.

Because friction tells you where the product actually breaks, not where your hypothesis says it should.

So much of discovery is about getting comfortable being uncomfortable. You’re not there to guide the user to success. You’re there to watch what happens when you don’t.

Film them. Capture both their screen and their face. You’ll see the truth in their expressions before you ever hear it in their words.

The One Thing AI Can’t Replace

Here’s what I remember most from those Zenly sessions:

A sixteen-year-old girl in Paris picked up the phone, opened the app, and her face went completely blank. Not confused. Not frustrated. Just… nothing. She tapped around for maybe ten seconds, then locked the screen and handed it back.

“I don’t get it,” she said.

That moment (that specific emptiness in her expression) changed our entire onboarding flow. We’d been so focused on explaining *features* that we forgot to explain *why you’d care*. An AI could’ve told us our retention numbers were bad. It couldn’t have shown us that look.

AI is incredible for synthesis. You can drop transcripts into an LLM, spot patterns instantly, generate summaries in real time. But if you skip sitting next to users and feeling their confusion, you lose the thing that actually matters: the gut punch of watching someone not understand what you’ve built.

When someone struggles through your flow in person, it rewires something in your brain. You can’t unsee it. You can’t rationalize it away. And you sure as hell can’t ignore it in your next sprint.

That’s not data. That’s empathy. And you can’t automate it.

Iterate in Hours, Not Weeks

At Zenly, we’d run three or four user sessions in a single day.

Build in the morning, test at lunch, rebuild in the afternoon. The moment feedback started to converge, we knew we were close. There’s no reason to wait weeks for a “research readout” anymore. The tools exist to synthesize in real time. But the feedback still needs to come from real humans, not survey panels.

Here’s a trick I still use: If a session goes perfectly and the user never looks confused, it’s probably a waste.

If they frown, hesitate, or ask “Wait, what am I supposed to do here?” That’s gold. Those moments are what you timestamp and dig into later. Because that’s usually where the product is broken, or occasionally, where it’s brilliant.

And the nice thing is that with AI and the ability to rapidly prototype, you can do this even more quickly than we did it just 4-5 years ago. The methodology still matters, but the speed of iteration can be even faster now with the latest tools - this is what we’re doing at Troon to test and rapidly ship.

Context Changes Everything

One of my favorite examples of discovery in action came from Showmax, our streaming video product in Africa.

We assumed we were competing with Netflix. Turns out, we were competing with *the movie guy*. A local who torrented shows, burned them to DVDs, and sold them on the street. He was the entire recommendation engine for Nairobi.

That one discovery changed our roadmap completely: bandwidth optimization, offline downloads, local caching. None of which would’ve come up in a U.S. customer interview.

We only found that out because we went there. We sat in people’s homes. We asked how they actually watched movies, not how they *wished* they watched movies. And we paid attention when the answer didn’t match our assumptions.

You can’t phone that in. You can’t simulate it in a survey. You have to go.

The Real Feedback Loop

Customer discovery isn’t about the perfect script or the right questions.

It’s about being present enough (and humble enough) to watch people use what you’ve built, fail with it, and tell you the truth without words. That blank look. That hesitation. That moment where they almost get it, but don’t.

Most founders are terrified of that feedback. I get it. It’s brutal to watch someone not understand the thing you’ve spent months building.

But that discomfort is the entire point.

Because on the other side of it is a product people actually want to use. Not the one you wanted to build. The one they need.

And the only way to find that gap is to sit in it.