How to Run a Five Person Test on Your Pricing Page

How do you find out why your pricing page is not converting?

Watch five people try to use it and say nothing while they do. Analytics tells you where people leave. It cannot tell you what they misunderstood. Thirty minutes of watching someone read your plans aloud will explain a drop off that a month of dashboards did not.

I have run this on a lot of sites across six years and seventy plus projects, and it is still the single cheapest piece of research available to a founder. No tooling budget, no panel, no statistical training. A call, a shared screen, and the discipline to keep your mouth shut.

What follows is the exact sequence I use. It takes an afternoon to set up, a day to run, and an hour to turn into a change list.

Why five people instead of a proper study?

Because you are hunting for blockers, not measuring preferences. Blockers are loud. If your plan names confuse people, the first two will trip on them and you will not need a third to be sure. The point of five is to see the same problem happen more than once.

A larger sample buys precision you cannot use yet. You are not trying to learn that thirty one percent of visitors prefer annual billing. You are trying to learn that nobody can tell what happens when they exceed the included seats, which is the kind of finding that fixes itself in an afternoon.

The trap is treating small qualitative work as though it produced numbers. It does not, and if you present it that way in a room someone will correctly dismiss it. Report what you saw, how many people it happened to, and stop there.

Who should you actually recruit?

People who could plausibly buy. Not friends, not designers, not anyone who has seen the page before. The closer they are to the role of your real buyer, the more their confusion means. Five people from the wrong role will send you in the wrong direction with great confidence.

If you sell to operations managers at mid sized companies, find operations managers. Your existing customers are acceptable for some tasks and useless for others, because they already know what your plan names mean and cannot un know it. Recent customers who bought in the last month are the useful middle ground.

Recruiting is the part people give up on. Ask in the communities where your buyers already are, offer a modest thank you, and keep the ask to thirty minutes. I have never found a round of five to take more than a couple of days to fill when the ask was specific.

What tasks should you set them?

Three, in this order. Work out what this product costs for a team of your size. Decide which plan you would pick and say why. Then find out what happens if you outgrow it. Tasks, not questions, because you want to watch behaviour rather than collect opinions.

That third task is the one that exposes the most. Overage, seat limits, upgrade paths and what happens at the end of a trial are where pricing pages are vaguest, and where a buyer's real anxiety lives. A page that cannot answer it sends people to your sales inbox or away.

Give the tasks one at a time, not as a list, and do not tell them what the page is meant to do. If you explain the pricing model before they start, you have destroyed the only thing you were trying to measure.

The same principle applies to anything you are asking a stranger to complete. Ask for the minimum and watch what happens, which is the argument I made in designing a form that asks for less.

What should you say, and what should you never say?

Say almost nothing. Ask them to think aloud, then let silence do the work. When they ask you a question, hand it back: what would you do if I were not here? That single sentence is most of the skill in moderating, and it is hard because the instinct to help is strong.

Never explain, never defend, never say it is actually there at the bottom. The moment you correct a misunderstanding you have lost the data. Write it down instead and let them struggle, because a real visitor will have nobody to correct them either.

Avoid leading questions. Was that confusing invites agreement. What were you expecting to happen invites a description. The second gets you a sentence you can act on, and often the exact wording that should have been on the page.

What are you watching for while they work?

Hesitation, re reading, scrolling back up, and any moment they say a number or a word that is not on the page. Those four signals cover almost everything worth finding, and none of them require you to interpret anything in the moment.

Re reading is the strongest one. When somebody reads the same line twice, the line is doing badly, regardless of how elegant it looked in the design review. Note the exact line, not your theory about it.

The second strongest is substituted vocabulary. If your page says workspace and everyone says account, your page is using an internal word. That is the cheapest fix in this entire exercise and it is usually sitting in the plan names. It is also a good signal about positioning generally, which is worth reading alongside how do you know your positioning is wrong.

Write timestamps as you go. Five sessions blur together by the next morning, and a timestamp lets you go back to the moment rather than trusting your memory of it.

How do you turn five sessions into a change list?

Put every observation on its own line, then group the lines that describe the same misunderstanding. Count how many of the five hit each group. Anything three or more people hit is a fix. Anything one person hit is a note. Anything two is a judgement call.

Then sort the fixes by effort. Most of what this test surfaces is copy: a plan name, a missing sentence about overage, a comparison row that assumes knowledge nobody has. Those ship the same week, which is why this method is worth running before any redesign rather than after.

Write the change list as specific edits, not themes. Clarify pricing is not a task. Add a sentence under the team plan stating what happens when you exceed ten seats is a task, and somebody can do it tomorrow.

A recurring finding is that the page hides the thing buyers most want. If your prices are not public, the test will tell you very quickly how much that costs you in confusion, which is the problem I worked through in what your pricing page should say when you hide prices.

What does this test not tell you?

Anything about magnitude. It will not tell you which price is right, how many people would buy at a different number, or what your conversion rate will become. Five people cannot answer those questions and pretending otherwise is how good research gets a bad reputation.

It also will not tell you about people who never reached the page, or who left before your session started. Those are traffic and positioning problems, and they need different instruments entirely.

What it does tell you is which parts of your page fail a motivated reader who is trying to give you money. That is a narrow finding and an extremely actionable one. Keep the claim that size when you present it and it will survive any scrutiny it meets.

What should you do next?

Book five thirty minute calls this week. Use the three tasks exactly as written. Say nothing, write everything down with timestamps, then group the observations and count. You will have a change list by Friday and most of it will be copy you can ship yourself.

Run it again after the changes, with five different people and the same three tasks. That second round is what turns an anecdote into evidence, and it is the part almost nobody does.

If you would rather have someone else moderate, or you want the findings turned into the actual page rather than a document, reach out. I work on fixed fees and I am happy to tell you whether this is even your biggest problem. Let's chat.