User journey study
You give a site address and a goal — for example “find out how much delivery to a parcel locker costs” or “find out how to renew a passport”. Up to twenty simulated people with different traits try to reach it in a real browser. In about an hour you see how many got there, which way they went, where their paths split and where someone would have given up.
Why different people, not one agent
The same site works well for someone who skims headings and clicks boldly, and badly for someone who reads everything in order, is afraid of making a mistake or is on a phone. One run shows one path. A population of profiles shows who finds the site hard — a question teams rarely get a quick answer to.
Each profile is a set of ten traits, each at one of three levels:
- attention and motivation
- device (phone, computer, computer at 200% zoom)
- domain knowledge
- self-efficacy
- learning style
- information processing
- reading style (skims / reads)
- attitude to risk
- satisfaction threshold
- web proficiency
The population is chosen to cover as many pairs of trait levels as possible: five people are three archetypes and the two profiles most different from them, ten are the archetypes and seven different ones, around twenty give full pairwise coverage. There is always one baseline run without a profile as a reference point.
What you get
- How many reached the goal — separating those who got there straight away from those who would have given up on the way, while the study kept going to see the rest of the path.
- A map of paths — which pages each person went through and where the paths split.
- Every step with a screenshot — what the person saw, what they expected, whether the site delivered and whether the step brought them closer to the goal.
- Site problems — gathered across runs, with the traits that were over-represented among those affected (e.g. “more often for people on a phone who skim”).
- Effect of traits — how success changes with the level of each trait, marked when a difference is within chance.
When it makes sense
- After every release of key journeys — purchase, sign-up, contact, finding information. After a fix you rerun the same study on the same profiles and compare.
- Before research with people — so sessions are not spent on obvious obstacles and tasks are better chosen.
- At the start of conversion optimisation — a cheap first review and a list of hypotheses to test.
What it does not replace
It is a simulation using language models. It is good at finding likely obstacles and showing who they affect, but it does not know the motivation, emotions or context of real people. Treat the result as a quick review and a source of hypotheses, and confirm important decisions with research involving users. More about the method: how it works.