userken documentation
userken builds synthetic audiences out of public opinion data: app store reviews, online communities, published surveys and your own uploaded CSVs. You define an audience, userken clusters the real people in it into personas, and you survey or interview those personas. Every answer is grounded in real text, and the accuracy of the method is measured in public.
These pages describe the product as it is built. Numbers such as credit costs and plan limits are read from the running code, so they cannot drift from what you are charged.
Start here
- Sign in with the magic link sent to your email address. No password.
- Pick an audience. Every app and category in the catalogue is already a built-in audience, so you can ask a question in your first minute.
- Ask something. Run a survey for a distribution with intervals, or open a focus group and talk to the personas.
- Connect Claude if you would rather work in a chat than in the dashboard. The same audiences and the same data.
Every page
Getting started
What userken is, how magic link sign-in works, what a workspace holds and how credits are spent.
Connect to Claude
Add userken to claude.ai or Claude Desktop as a custom connector, or use an API key from Claude Code, with the full tool list.
Audiences
Built-in audiences, the filter builder, describing an audience in words, building personas, comparing audiences and audience refs.
Surveys
Survey templates, question types, samples per persona, grounding, how to read the results, follow-ups, exports and what a run costs.
Focus groups
Setting up a room, the discussion guide, moderating, grounding, summaries and the Markdown transcript export.
Sources
App reviews, online communities, published surveys and your own uploads, with the CSV and survey pack formats.
Accuracy
What NDAM and 1-MAE mean, the hold-out method, the baselines, how to read the accuracy page and what pending means.
Plans and credits
The plans, what each action costs in credits, credit packs, and what happens when a workspace runs out.
Workspaces and roles
Members, workspace admins, superusers, personal API keys and connected applications.
Self-hosting
Environment variables, running more than one host, leader election and the system admin area.
Words used here
- An audience ref is the short string that names an audience everywhere in the product:
pub:reutersfor one app,cat:newsfor a category,aud:aud_1a2b3c4d5e6ffor one you built. - A persona is a cluster of real reviewers with a name, a description, a voice, themes and a weight. The weights of an audience's personas sum to 1.
- Grounding means the persona was shown real reviews from its own segment, chosen by similarity to the question, before it answered.
- A credit is the unit of metered work. One standard survey question or one focus group turn costs one credit.