Which steps are friction and which are thinking

Every step a product removes from a journey is a step where the user used to verify something. Nobody is accounting for what happens when all of them are gone.

Product design has treated step reduction as an unqualified good since the first checkout funnels. Fewer clicks, fewer fields, fewer screens between question and answer. Intercom put it cleanly in 2021: the only purpose of a product is to help someone go from A to B, and the fewer steps, the better the journey.

That rule still stands while the steps you remove are friction. But it breaks when the step you remove is where people are doing the thinking.

Searching used to mean typing keywords, scanning ten links, picking one that looked credible, reading it, deciding whether to believe it, and checking a second source. Growth.design lists that sequence as the problem AI chat solves. Picking, believing, checking. But half that sequence is the user evaluating a claim. Removing those steps transfers a job from the reader to the engine.

Which steps in your flow are overhead, and which ones are someone deciding something? Most teams have never separated the two.

Take any two screens in your product and ask what sits between them. Not what the user clicks. What the user has to work out on their own.

Does the person have to guess something the interface never told them? What are they deciding, and did you give them anything to decide with?

The gaps hide in the seconds after an action, when the product goes quiet. Someone clicks “Start free trial” and lands on a dashboard with no indication of what the trial covers, when it ends, or what to do first. Nothing broke. The flow works. But the person is still guessing.

Samuel Hulick made this the center of his article in Mind the Gap, and it remains one of the fastest audit in product design because it costs nothing. Open your own signup flow and write down every moment you had to know something the screen did not say.

Feedforward is the principle underneath: showing what will happen next before someone acts. A user hovering over “Submit” with no idea whether that clears the form, sends an email, or charges a card is standing in a gap you left. The cues that close it are small enough that nobody thanks you for them.

Someone did not wake up wanting to use your app. They wanted to book a flight or get information about which SUV to buy. Your product intersects a path that started somewhere else and continues after they close the tab.

Map that longer path and two things become visible: where each step could carry less weight, and where you have inserted your own priorities into someone else’s route. The email capture before the content. The plan selector before the person knows what the plans do. The product tour that starts before anyone has used the product.

Timing is the whole argument. A phone number field reads as routine at checkout and as a toll on a landing page. Same request, different moment, opposite outcomes. There is a time and place for the business ask, and it is later than most teams put it.

The peak-end rule is worth understanding here. People judge an experience by its most intense moment and its ending, not by an average of the steps. A journey with one excellent moment and a clean finish beats a smooth one nobody remembers.

Map for the peak and the end, then spend your effort there.

Open most apps and something wants to sell you before you have done anything. Go back to the reason the person came.

Take a screen and go element by element. Why is this person here? What are they trying to finish? Does this element serve that, or does it serve a metric on your dashboard?

The distinction between those last two is the work. A “Refer a friend” banner on an empty-state dashboard serves the growth target. A “Refer a friend” prompt after someone completes their third successful project serves both. Same component, different placement, and only one of them is honest about the moment it sits in.

Hick’s law gives you the cost of getting this wrong. Response time grows with each option added, and a checkout with six payment options, three shipping tiers, and a promo field visible at once stalls people who arrived ready to buy. Cutting choices is not the same as cutting complexity. The goal is fewer decisions at any one moment, not a product that decides for people.

Most redesigns start here. Everything sits at the same visual weight, related things drift apart, and the screen offers no structure to read.

Put related elements close. Push unrelated ones apart. Let the interface breathe. It sounds too basic to write down, and it accounts for more redesign work than any sophisticated technique.

Proximity is doing the work. Place two things near each other and people assume they are related, whether or not you meant them to be. Grouping is meaning. That assumption works for you or against you depending on how carefully you place things, which is why a “Cancel subscription” link sitting eight pixels from “Update payment method” produces support tickets.

Spacing is the other half and gets cut first when a screen fills up. Space is what makes a group readable as a group.

Design without content is coloring.

The design of a page depends on what goes inside it. Start with a menu or a box and you are arranging containers for text nobody has written, which means the text will later be trimmed to fit the box. The box wins. It should not.

Start with the job instead. What is this person trying to get done, and what do they need in front of them to do it? Real content, real length, real edge cases. The longest product name. The empty state. The error.

Jesse James Garrett put content underneath structure in his five planes model. Skeleton and surface sit on top of scope and structure. Designing the surface before the scope exists inverts the model, and every designer who has placed lorem ipsum into a product knows how that ends.

AI chat collapses the search journey. Type a question, get an answer, refine with a chip labeled “best budget” or “fuel efficiency,” decide. Twenty seconds, no source opened. By every measure in this article, that is a better journey. Fewer steps. Less friction. The person got from A to B.

The refinement chips are the part to look at closely. Tapping through them feels like exploring. The person is choosing among options the engine wrote, narrowing inside a frame they did not set. The frame came with the answer.

If people stop exploring, publishers lose the traffic that funds the writing the engine summarizes. The engine’s quality rests on a supply it is defunding. Neither Microsoft nor Google has explained how that resolves.

The design lesson is narrower than the media one. When you remove a step that contained thinking and decision making, you owe the person a way back into it. Authority bias explains why: fluent, confident, well-formatted output reads as verified, and the interface never signals otherwise.

I wrote about the presentation side of this in when tools pretend to be people. This is the structural side. A system can drop every human affectation and still remove your verification steps by collapsing the journey.

Removing steps is usually right. Know which ones.

— Thanks for reading. The newsletter stays free, and I’d like to keep it that way. Your support makes that possible.

Four of the five tactics above run on principles I’ve spent years documenting. Feedforward, Hick’s law, proximity, peak-end. I keep the whole set in one place.

106 principles, each written as a working tool. What it is, how it shows up in real interfaces, where to apply it, and the companies already using it. A searchable Notion database you keep open while you work: filter by cluster, find the principle that fits the screen in front of you, ship something that fits how people are wired.

From the Zeigarnik Effect to the Labor Illusion to Hick’s Law.

Get the 106 principles →