What is the 3W Loop?

Most teams I walk into have plenty of data and plenty of user feedback, and still ship features nobody uses.

That is the part that confuses people. The dashboard exists. The research repo exists. Someone talks to customers most weeks. And the last four things the team shipped are sitting at single-digit adoption.

The problem is rarely the tools. It is the order in which the questions get asked.

The 3W Loop puts them in order. Where it hurts, Why it happens, Whether the fix works. Three questions, three different instruments, and the answer to the last one starts the next loop.

The three questions, and the instrument for each

The reason teams stall is almost never that they cannot answer a question. It is that they answer one of the three well and quietly guess the other two.

Where: product analytics

What is broken, for whom, and how badly?

This is a funnel step, a retention cliff, or a two-number comparison between segments. Something specific enough to point at. Not "activation is weak" but "62% of new accounts never reach the second session, and it is concentrated in self-serve signups".

Skip it and you get research with no target: interviews about everything, insight about nothing. The artefact you leave behind is a funnel or a segment comparison. If you have never had to build one before you have numbers worth trusting, that is its own problem.

Why: story-based interviews

What is actually causing it?

Not "would you use a feature that..." but "tell me about the last time you...", asked of the people in that segment rather than whoever answers first. The distinction matters more than any other technique in discovery, and five to eight conversations gets you most of the pattern.

Skip it and you fix the obvious thing for people who were never struggling with it. The artefact is an interview snapshot: what you heard, from whom, and what it changes.

Whether: a scoped experiment

Does the fix move the number?

A test card, written before you run it: we believe that, to verify we will, and measure, we are right if. Then a learning card after. The mechanics of getting this right are covered in designing a product experiment.

Skip it and you ship correlations as features and never learn what worked.

The four rules

The order is the whole method, so the rules are mostly about the order.

1. Ask in order. A Why without a Where is unfocused. A Whether without a Why is a lottery ticket.

2. You may skip Why, but only when the fix is cheap, reversible, and has one mechanism everyone in the room can name. That is a decision you make out loud, not an omission you notice afterwards. Working out which situation you are in is the argument every team has, and it has a decision rule.

3. You may never skip Whether. A test you cannot fail is a demo.

4. Close the loop. The learning card's number is the next Where.

One loop, end to end

The clearest example I have is elyps, a French-Belgian neobank where I was the first product hire.

Where. Roughly eight in ten new sign-ups were abandoning onboarding, and analytics showed it was not spread across the funnel. It was concentrated almost entirely on one screen: the step right before identity verification started.

Why. The obvious read, which market reports agreed with, was that the ID check was too much friction. Story-based interviews with churned users found something else. People were opening the app on the subway and at work, and did not want to pull out their ID in public. The problem was the moment, not the step.

Whether. We added a "Do it later" option on that screen, a notification consent screen that explained the reminder, and a push at 8pm with a deep link back. Success criterion written first: overall onboarding churn down at least 10%. Result: drop-off on that screen fell from roughly 80% to 60%. The full case study has the test card and the learning card.

Notice what the Why bought. Without it, the honest next move would have been redesigning a legally mandated identity check, which is expensive, slow, and in this case would not have worked.

When you should skip the middle question

Sometimes the mechanism really is obvious and the fix really is a toggle.

At UX Pilot, users who tried one particular generation mode in their first week converted to paid at roughly nine times the rate of those who did not. The fix was turning it on by default. Cheap, reversible, one mechanism nobody argued about. We went straight from Where to Whether and ran it as a proper A/B test anyway. Free-to-paid conversion rose 44.67%.

That is rule 2 working as intended. The skill is knowing which question is actually open, rather than defaulting to the instrument you personally prefer.

What each step leaves behind

One artefact per question, which is what makes the loop auditable three months later:

• Where leaves a funnel or a segment comparison.
• Why leaves an interview snapshot.
• Whether leaves a test card, and then a learning card.

That last number is not the end of anything. It is the next Where, which is why it is a loop rather than a checklist. If your experiments currently end in a slide instead of a number, that is the thing to fix first.

The framework is free

The 3W Loop is published under CC BY-SA 4.0. The full description, the four rules, and the downloadable kit are here: the graphic in PNG, SVG and print PDF, light and dark, plus an A4 canvas for running the loop on paper with your team. Put it in your decks, teach from it, adapt it. Credit it and share-alike anything you adapt.

Reading it is the easy half. Making it the way a team actually decides things takes running it on a real number a few times, which is what the Product Discovery Workshop is built around.