A founder described his consumer app's onboarding this week: ten open-ended questions instead of the usual pick-three-interests. Five minutes in a market that optimizes for thirty seconds. He expected brutal drop-off. Instead completion came in higher than expected and the answers arrived as paragraphs, not fragments. Then he asked the question that matters: am I being principled or just stubborn?
He built a filter and is calling it a flaw. That confusion is everywhere, and it costs founders real information.
Does removing friction always help?
It helps the number you are staring at. More people complete a shorter signup, so signups go up and the dashboard looks healthier. But a signup that costs nothing measures nothing. When everyone gets through, your user base fills with people who were curious for eleven seconds, and every downstream metric inherits that noise. Activation, retention, conversion, all of them are now computed over a population that was never going to stay.
The slow onboarding does the opposite. It shrinks the top of the funnel and concentrates it. The people who answer ten real questions have told you something true about themselves before using the product once.
What does each kind of friction measure?
A free click measures curiosity. A step that costs effort, a form with real questions, a bank connection, an import, measures intent. A step that costs money measures demand. These are three different instruments, and most funnels only instrument the first one, then present the result as validation.
A hiring marketplace posted its funnel this week: searching candidates is free, talking to one costs $99. One user went from signup on Monday to a signed contract on Thursday. The fee did not scare demand away. It located it. Whoever pays $99 to run an interview has a real opening and a real deadline, and the marketplace knows it within minutes instead of guessing from page views.
When is friction just bad design?
The filter argument is not a license for clunky products. Friction only measures intent when the cost is one that the right user is willing to pay and the wrong user is not. Ten thoughtful questions filter for people who care about the answer. A confusing interface filters for people with high pain tolerance, which is a different population and usually not your market. If the people giving up are exactly the ones you built this for, that is not a filter, that is a leak.
How do you read a filtered funnel?
Smaller, but truer. A thousand frictionless signups tell you your ad worked. Eighty people who finished a five minute onboarding tell you eighty people want this. When you compare cohorts, the completers against the skippers, you find out whether your filter selects for the behavior that predicts staying. If completers retain at three times the rate, the friction is doing research for you every day, for free.
The founder with the ten questions does not have an onboarding problem. He has an instrument most of his competitors deleted in the name of growth.
Key takeaways
- Every funnel step is also a measurement. Zero friction means zero information about intent.
- Effort filters for intent, money filters for demand, free clicks only filter for curiosity.
- Good friction costs something only the right user will pay. Friction that filters on patience is just bad design.
- Judge a filtered funnel by cohort behavior, not by the size of the top.
Growth counts the people who got in. Evidence is what they gave up to get there.