Friction

The fact that conversion was low has always been visible in the numbers. Five thousand people reached the signup screen and 1,200 created an account — anyone could read that. What nobody could read was the next part. There was no word for what the other 3,800 ran into.

So improvement meetings turned into arguments about taste. One person said the screen looked cluttered, another suggested changing the button color, and a month later the same number came back and the same conversation happened again. Without names for the individual reasons people left, no list of things to fix ever got written.

Friction is the word borrowed to fill that gap. In physics it describes the resistance that slows a moving object, and treating a user as something in motion makes the accounting work. A product contains forces pushing people forward and forces holding them back, and however strong the push is, people stop wherever the holding force is stronger. Once teams had the word, meeting notes stopped saying "improve the signup screen" and started saying "remove the company-name field, three-minute wait for the verification code, card number requested before anyone has tried the product."

Friction accumulates as the price someone has to pay to get through one step, and that price comes in three kinds.

1. Time. Two minutes to fill in nine fields, three minutes waiting for a verification text, three seconds for a screen to load. This kind is easy to measure, so it's the kind teams find first.
2. Thought. A field label nobody understands, a set of options with no obvious right answer, a screen that doesn't show what to click next. Anywhere a person has to stop and work something out, a charge is added.
3. Unease. A card number requested before the product has been used at all, a company-name field with no visible purpose, a checkout step that says nothing about how to cancel later. Unlike the first two, this kind doesn't shrink when you make the screen faster.

Finding it follows an order. Start with the pass rate at each step, pick the step that sits well below its neighbors, then take that one step and watch ten session recordings of it. That's where you see where people stop and what they type and delete. Then put the same screen in front of five people who have never seen it, and the reasons that recordings can't show get said out loud.

Removing it follows an order too. Delete what can be deleted, defer what can't be deleted, and simplify what can't be deferred. The company-name field can usually be deleted outright. Information you genuinely need eventually, like team size, comes out of the signup screen and gets collected in pieces while people use the product — that's progressive profiling. Creating and remembering a password can't be deleted or deferred, so it gets replaced with social login, which lets people in through an account they already have. When Pagedeck cut its signup screen from nine fields to three, the pass rate went from 24 percent to 38 percent.

It gets addressed first at the narrow points every user has to pass through: the signup flow, which is the sequence of screens between arriving and having an account; the checkout steps; and onboarding, the stretch where a new user settles into the product. A single percentage point at these points raises the headcount of every step behind them.

The effect of removing friction is read through the conversion rate — the share of visitors who take the target action in a period. That number also moves with seasons and ad spending, so isolating the effect of one change calls for an A/B test, which splits visitors at random between the old screen and the new one and compares the pass rates of the two groups.

Removing all of it isn't the goal. Some friction is kept on purpose: the confirmation before a document is deleted for good, the screen that restates the amount before an annual charge, the check that a signup is a person rather than a script. Take the friction at those points to zero and mistakes and abusers get through at the same speed as everyone else. One rule sorts what stays from what goes. Hold people before actions that can't be undone, and let them through before actions that can.

"There's too much friction on the signup screen. Two of those nine fields are things we need right now."

A proposal to delete the other seven or ask for them later.

"That friction is intentional. Permanent deletion should ask twice."

At this point preventing mistakes outranks convenience, so leave it alone.

"I don't think the copy is why conversion is flat. I think it's the wait for the verification text."

The cause isn't on the screen, it's in the waiting.

Last quarter Ravi brought the step-by-step signup numbers to Pagedeck. Of the people who reached the signup screen, 24 percent were finishing an account. Noa watched ten session recordings of that screen and pulled out three places people stopped: the cursor sat in the company-name field for a long time, half of them closed the window at phone verification, and some left while reading the password rules.

Those three went straight onto a list and Theo worked through them. The company-name field came out, phone verification moved to checkout, and passwords were replaced with Google sign-in. Mia kept one thing as it was — the confirmation before a document is deleted for good. One document deleted by mistake costs more than the three seconds that dialog takes.

The most common mistake is reading friction as something to eliminate on principle. After a few rounds where every removal raises conversion, that's what the lesson hardens into. But friction is the force that holds people back, which means it holds back bad actions too. Remove the confirmation step so an annual charge goes through on one click and that month's transaction count rises, while the next month brings refund requests — and once support time and cancelled contracts are subtracted, nothing is left.

The second mistake is treating friction as a visual design problem. It gets discovered on screens, so that's where it seems to live. In reality it accumulates out of waiting time, information requirements, company policy, and legal steps as well. Miss that and a team can spend half a year running experiments on button colors and wording. The three-second load and the card number requested up front are still there the whole time, and every experiment ends in a difference too small to mean anything.