Positive Friction: Why Slowing Users Down Drives Adoption

Most product teams treat friction as pure loss. Every extra field, every confirmation, every question between the user and the result is something to delete. For AI products the pressure is even higher: type a prompt, get a finished app, feel the magic. But the fastest path to a result is not always the path that keeps the user. Positive friction, a deliberate and well-placed slowdown that makes the output better or the relationship stronger, is often what separates a product people try from a product people stay in.

I recently sat down with a VP of Product who leads product at a fast-growing AI app-builder, and her team reached the same conclusion from their own data. Users don't just want the fastest result. They want one that works well enough to serve their goal, and they stay with the product that helps them get there.

What is positive friction in product design?

Positive friction is a slowdown you add on purpose because it improves the result the user gets or the trust they place in you, not because your system requires it. It is the opposite of the accidental friction teams spend their sprints deleting.

There is a well-documented reason a little effort pays off. Researchers Michael Norton, Daniel Mochon, and Dan Ariely found that people were willing to pay 63% more for furniture they assembled themselves than for identical pre-built units, and the effect held only when the work ended in a finished product. When a user invests some effort shaping their own result, they value it more and are more motivated to keep going. A product that does everything for them in one shot leaves them nothing to own.

When should you add friction instead of removing it?

Add friction only for the users who need it, and only where it lifts the outcome. The same clarifying question that rescues a vague request annoys the user who already knows exactly what they want.

When someone types "I want a shoe store," the product can race to generate something beautiful, or it can ask two or three targeted questions first. Her team runs the prompt through an analysis step and asks only the questions that specific request is missing. A shoe store and a barbershop booking app need different questions, and a user who mentions a "vibe" gets asked about visual direction while one who doesn't, isn't. The discipline is knowing who to skip. A user who arrives with a detailed, well-formed prompt goes straight through, because asking them questions just to seem thorough is friction with no payoff.

The same logic drives a quiet planning step. Power users of coding tools deliberately plan before they build; most people have never heard of a plan mode and would never ask for one. So the product takes a four-word prompt into a short planning pass on its own, because the result is better and the user never had to know they needed it. For a product leader the lesson is that friction is a targeting problem before it is a design problem. Instrument who actually needs the extra step, then spend it only on them. And targeting needs a target: name your activation moment, the one action that predicts whether a user stays, and treat every friction decision as protecting the path to it.

How do you time a paywall so users see the wall before they hit it?

Show users the wall before they hit it. Offer the upgrade slightly before their limit runs out, on a signal that combines how much they have used and how long they have been working, never in the frustrated moment after they have already crashed into it.

Time it too early and the prompt reads as a shakedown: the user has barely started and you are already asking for money. Time it too late and you get the worst version, a user who was deep in flow fixing something, who slams into a credit limit with no warning, and who now blames you twice, once for the wall and once for not saying it was coming. The moment that works sits just before the limit. You are about to run out, you will need more to keep going, upgrade now and stay in your project. The user sees the wall approaching and steps over it, and because you have been useful up to that point, they trust the ask.

Getting that moment right is a strategy question I work through with product leaders in advisory: where to place friction and when to ask for money without spending the trust you have built.

What kind of friction backfires?

The wrong friction treats a broken experience as a pricing problem, or a decision problem, when it is neither: consolation credits read as a bribe, and a full pricing menu at checkout is overload. Two experiments her team ran are worth stealing.

The first: when they spotted a frustrated user, easy to detect because unhappy users say so right in the chat, the instinct was to comp them free credits, the way a restaurant sends over a dessert when a dish disappoints. It did not work. A user who asked for something and did not get it does not want a consolation prize, they want the thing they asked for. The credits read as a bribe, and people see through a bribe. What worked was fixing the underlying result behind the scenes and re-earning the trust directly.

The second: at the payment moment, handing the user the full menu of plans to choose from felt respectful, and it backfired. Full choice at a money-sensitive moment is cognitive overload, and it quietly signals that you will not commit to a recommendation. The fix was to be more opinionated: recommend one specific plan before the user ever reaches the picker, the way a "most popular" tag does the deciding for you. Both point to one rule. Never add friction when a user is already frustrated or already deciding about money.

How much should you decide for the user?

Decide the most for the user at the start and at the moment of payment, and the least in the middle. The right amount of "we know better" changes across the funnel.

One rule holds the table together: the options you offer have to be ones the user is actually capable of taking. Offer a novice the advanced connector they have never heard of and you have built a wall. Offer them the next thing they can reach, and they trust you a little more each time. This is why friction is a risk you manage, applied only where it earns its place. How you guide an enterprise buyer is not how you guide someone's parent trying the tool for the first time, and the broader your user base, the more precise you have to be.

Even "finished" is a friction decision. Users feel done the moment they publish, but the product keeps growing, so the team brings them back with segment-specific nudges tied to their business. A shoe store hears about one new capability and a to-do-list app about another, framed as a way to grow rather than a reminder to use more tools. Positive friction does not stop at activation. It is how you keep earning the next session.

Key takeaways

• Add friction only where it earns its place. Target the users who need a clarifying question or a planning pass, and send the ready ones straight through.
• Give friction a target. Name the activation moment that predicts retention, and treat every slowdown as protecting the path to it.
• Show the wall before users hit it. Prompt the upgrade just before the limit, on usage plus time, never after the crash.
• Be more opinionated at the money moment. Recommend one plan; a full pricing menu at checkout is overload, and consolation credits read as a bribe.
• Decide most at the edges, least in the middle. Closed at the blank page and at payment, broad and guided while they build.

If your activation numbers are strong but retention leaks, the problem may not be too much friction. It may be friction in the wrong places. If you want a second set of eyes on where your product should slow users down and where it should get out of their way, book a call through product advisory. More field notes like this live on the blog.