Sample report
A full report, run on our own site
We cannot show you a customer's report, so we ordered studies of this website through our own API, the way a customer would, and paid for them from our own credits. The sample is the full Friction study: 360 simulated visits, two audiences (people in product roles at software companies, and developers who use AI coding assistants daily) and a control group, each run twice.
It is a report about us, and not a flattering one. The finding that held in every group was that the previous sample report's grey labels and small controls were too small to read, especially on a phone — 50 of 120 in each audience said so, and roughly a third of those left there. We raised every text size in the report and the site the same day. An earlier Quick read had found that the sample link itself led to a raw error page; that was fixed first, and a Fix check on the same people confirmed visitors now reach the report. Most participants in every group decided the product was not for them. We publish that too.
What to look at
A report is long on purpose: every number in it can be opened to the people behind it. If you have five minutes, read these in order.
1. The summary
A few paragraphs written the way a research lead would brief you, with a point of view. Every number in them is a count from the tables further down; the summary cannot say anything the data does not.
2. Who took part
The participants table says, once and plainly, that they are simulated. It lists each group as resolved from a plain-English description, how many personas in the pool matched, and the control group with no use for the product.
3. The findings
Each is rated blocker, major, minor or cosmetic by a fixed rule from the sessions, not by opinion. Under each: the screen it happened on and the control acted on just before, how many raised it and how many left there, the rate in each group against the control, whether it held across both runs, three quotes, a suggested change, and how to check the change worked.
4. Three narrated visits
Step-by-step accounts built from the session records: someone who left at the top finding's screen, the participant who tried hardest, and one person from the control group.
5. What to change first
A table of changes with severity, screen, estimated effort and how to verify, in priority order.
6. Limits and instrument health
What this study could not see, at the same weight as what it found, and how clean the run was: sessions that errored, clicks that had to be forced, sessions cut off by the step budget. A study graded suspect is not delivered.
7. The appendix
How it was made, generated from the study's own record: the persona pool, what each participant saw, the models, the step budget, how things were counted, and every participant with their own words.
What not to read into it
Percentages in the report describe simulated participants. They are for comparing groups and versions on the same site, not for forecasting how many of your real visitors will do something. Why.
A Quick read is the short form of this: one run of 30 per group, themes and quotes, and no percentages.