Designing decision interfaces that reduce friction

Friction is the work you make someone do in order to buy from you

A decision interface is any point where someone has to choose. A pricing page, a form, a comparison table, a checkout, an enquiry that asks them to describe their problem before you have told them what you do.

Most advice about friction is written about checkouts, because that is where friction is easy to see and easy to count. For a considered purchase, that is the wrong place to look. By the time someone reaches your contact form, the expensive friction has already happened. They spent 20 minutes trying to work out whether you were the right sort of business, and the ones who could not tell never arrived at the form to be counted.

My position is that friction in a small business sits mostly in the decision, not the transaction, and it is invisible because nothing measures it. Nobody logs the person who left your services page unable to decide.

What the evidence supports

Two pieces of research are worth knowing here, and they point in slightly different directions.

The first is effort. Matthew Dixon, Karen Freeman and Nicholas Toman published Stop Trying to Delight Your Customers in Harvard Business Review in 2010, based on more than 75,000 customer interactions. Their finding was that exceeding expectations did much less for loyalty than reducing the effort a customer had to spend. Customers do not reward you much for going beyond what they expected. They punish you for making them work. That work produced the customer effort score, which asks one question about how easy an interaction was. The research is service interactions rather than buying decisions, so carry it as a strong signal about effort rather than a rule about sales pages.

The second is abandonment. The Baymard Institute maintains a running average of cart abandonment across 50 studies from 2006 to 2025, currently 70.22%. Their own research on why people abandon during checkout puts extra costs such as shipping, tax and fees at 40%, slow delivery at 20%, payment security concerns at 19%, being forced to create an account at 18%, and a complex checkout at 17%.

Read that list carefully, because it undercuts the usual conclusion. The top reason is not an interface problem. It is a pricing and policy problem that the interface reveals at the worst possible moment. The lesson is not to make the checkout prettier. It is to stop surprising people, which mostly means saying the cost earlier.

That is also the shape of the equivalent problem in services. If you do not publish any pricing signal, every enquiry carries the risk of a wasted conversation, and the buyer knows it.

A friction audit by stage

This is the exercise. Take your own journey, stage by stage, and find the work you are asking a buyer to do. The last column matters most, because friction you cannot detect is friction you will not fix.

Stage	What the buyer is trying to do	Friction that stops them	How you would notice
Arrival	Work out whether you solve their problem at all	An opening screen about you rather than the problem	High exits from the homepage and service pages, short visits
Qualification	Decide whether they are your kind of client	No description of who you work with, or who you do not	Enquiries from people you cannot help, and long unqualified calls
Cost	Judge whether this is affordable before investing more time	No pricing signal of any kind	'Can you give me a ballpark' as the opening question, then silence
Process	Understand what will happen and what is expected of them	No description of how the work runs	Questions about process arriving after the proposal, not before
Risk	Decide whether it is safe to commit	Nothing about what happens if it goes wrong	Long gaps late in the conversation, requests for references
Commitment	Take the next step	A form demanding a detailed brief, with no statement of what follows	Forms started and abandoned, or replaced by a vague email

Fill it in for your own business. Then check the last column against what you already have: your inbox, your call notes and your analytics. Most of the evidence is already there and unread, which is the same point I make about finding user needs versus what users ask for.

Some friction is worth keeping

The instruction to remove all friction is wrong, and following it produces businesses that are easy to contact and full of bad-fit enquiries.

Good friction protects the buyer or improves the decision. Confirmation steps before an irreversible action, cooling-off periods, and a pause before someone cancels something they may need are all deliberate friction that serves the person. Clarasys makes the case for friction that protects the user or builds trust, including examples such as 2-step authentication and Monzo's gambling block with its built-in delay. Protective friction can also read as risk to the person meeting it, when the protection is invisible and the extra step is not. Why safer products can still feel risky works that through with passkeys.

My version for a small business is narrower. Keep the friction that filters, and remove the friction that confuses. A qualifying question on your enquiry form that says who this service is not for costs you enquiries you did not want. A required phone number before you have explained what you do costs you enquiries you did want.

That is the same argument I make about clarity letting people leave with confidence in what government content design taught me. If your interface makes a bad fit obvious sooner, that is a feature.

Choice, handled honestly

The standard claim is that more options paralyse buyers, so you should always show fewer. It is more complicated than that.

A meta-analysis of 99 observations involving 7,202 participants found that choice overload is heavily moderated by other factors, including the complexity of the choice, how difficult the task is, how certain the person is about their preferences and what they are trying to achieve. Fewer options do not automatically convert better.

What holds up better is that people need a basis for choosing. Three service tiers with no explanation of who each suits is a harder decision than 6 tiers with a clear rule for picking one. If you offer options, give the reader the sentence that tells them which one is theirs. That is usually more useful than deleting an option.

The same care applies to anchoring and defaults. They are real effects, and they are also the tools that turn into dark patterns when used against the buyer's interest. A default that saves someone work is good design. A default that opts them into something they would not choose is the kind of practice now covered by the unfair commercial practices provisions I set out in the role of emotions in purchase decisions.

Measuring effort without a research budget

The customer effort score exists and is worth borrowing. Ask one question after an interaction: how easy was it to get what you needed. A 5-point or 7-point scale is enough, and the free-text box after it is where the value is.

Three cheaper signals tell you most of it.

• Count the questions you answer more than twice. Every repeated question is a page that failed, and the repetition is your priority order.
• Time the gap between first contact and the question they were afraid to ask. If price, timescale or risk only comes up on a call, your site is making people pay for that information in social effort.
• Watch one person use your site while trying to decide whether to enquire. Do not explain anything. Note every pause. This takes half an hour and outperforms most analytics.

Start with the repeated questions. Take the last 20 enquiries, tally the questions, and fix the page that should have answered the most common one. That is the highest-value friction you can remove this week, and you already have the data to find it.