Use case
A/B and A/B/n testing
Rank options against each other instead of scoring them one at a time.
The problem
Why this is hard today.
A real experiment needs traffic, instrumentation and weeks. Most decisions do not get one, so options get compared by argument. And a live test tells you which option won without ever telling you why.
How userken does it
Four steps.
- Pick the audience, and decide whether you care about the whole audience or one segment.
- Enter two to six options.
- Run an A/B/n test. Every persona sees the options and is sampled several times, so you get a distribution rather than a single opinion.
- Read the weighted winner, the per segment winner (they often differ), and the reasons.
run_ab_test
Two to six options, weighted winner, per segment split.
ask_audience
Open ended follow up on the option that lost.
A/B/n template
In-app template with verbatims per option.
Tool names are the MCP tools and in-app templates. The same engines power both.
Illustration
What a run looks like.
Illustrative example, not a result we are reporting
Illustration: a browser picks a default new tab layout
- Audience: the browsers category, reviews that mention tabs or the home screen.
- Options: a blank page, a page of shortcuts, and a reading feed.
- What you would read: the overall ranking, and the fact that a power user segment and a casual segment may want opposite defaults.
Related
Next door to this.
Run a a/b and a/b/n testing on your category.
Accounts are provisioned by hand during early access. Tell us the question and the category and we will set you up.