Methodology
How Gate83 research works — and where it stops.
Synthetic research is a fast way to find out where different kinds of visitors might struggle. It is not a replacement for talking to real ones. This page is the one place we spell out the limits, so the rest of the site doesn't have to repeat them.
- 1
Personas are research instruments, not customers
Each of the fifty personas has a stable ID, a designed behavioral profile (persona DNA) and a research history. Their profiles describe intended behavior. Every designed value is a modeling assumption, and the app shows where each attribute came from.
- 2
Not a sample of any population
Fifty synthetic travelers are chosen for contrast, not to mirror a market. Results show where different kinds of visitors may struggle; they are not percentages of real customers and shouldn't be read as such. Validate important decisions with real people.
- 3
Sessions happen in a real browser — or not at all
Journeys are recorded by a browser worker visiting the actual pages: screenshots, clicks, scrolling and timings. Gate83 never fills a gap with simulated sessions. Where nothing has been recorded yet, the page says so.
- 4
Observation and interpretation stay apart
Every point in a report carries one of four plain labels, so you always know whether you are looking at what happened or at what someone thinks it means.
- 5
No score without evidence
Scores are withheld unless every scored session has verified screenshots, traces and timestamps. Missing information is shown as “Not specified”, never guessed.
- 6
AI summarises, it doesn't certify
Consolidated reports are drafted by AI from recorded steps. Any claim that cites a step that wasn't recorded is removed, and confidence is calculated in code from the evidence, not by the model.
The four evidence labels
- ObservedE1
- Directly recorded by the browser: screenshots, clicks, page text and timestamps.
- Persona readingE2
- A synthetic persona's interpretation of what it saw. Useful signal, not consumer evidence.
- InferenceE3
- A hypothesis drawn from observations. Needs validation before it drives a decision.
- CorroboratedE4
- Supported by an external source such as analytics or real customer feedback.
About the examples on this site
Persona portraits, names, families, archetypes and trait values are the real persona data. The quotes on the homepage, and the reactions, stances and outcomes in the sample report, are designed examples that show the format — not results of a real mission.
The sample report is about Harbour & Key, a fictional hotel review site we built for it. Its screenshots are real captures of that page, and every “Observed” line is true of it.
The live pilot on RoomExpert shows only real recorded sessions. Until they are recorded it shows the plan and an empty journey space.