Experiment recipes
These recipes combine FromUsers logic tools into patterns teams use often. Adapt the pieces. Do not enable every feature in one study unless each one has a clear job.
Recipe 1: Classic prototype A/B
Section titled “Recipe 1: Classic prototype A/B”Goal: Each participant uses one of two prototypes on the same tasks.
- Create a randomization group variable with arms A and B.
- Enable compare assets driven by randomization.
- Map A and B to their prototypes, including start/end/success URL.
- Keep task wording and success criteria identical.
- Analyze completion and ratings by group.
Recipe 2: URL-forced variant
Section titled “Recipe 2: URL-forced variant”Goal: Stakeholders or CRM segments open a predetermined arm.
- Create a URL variable such as
variant. - Drive compare assets or display logic with the condition driver.
- Share links that include the parameter for each arm.
- Optionally keep a random fallback for generic links.
Recipe 3: Screener then experiment
Section titled “Recipe 3: Screener then experiment”Goal: Only qualified participants enter the experiment.
- Place screener questions first.
- Use branching to end disqualified sessions.
- Assign randomization groups only on the main path (or ignore DQ rows in analysis).
- Run compare assets / arm-specific content after qualification.
Recipe 4: Shared survey, arm-specific modules
Section titled “Recipe 4: Shared survey, arm-specific modules”Goal: Common intro and demographics, different middle modules by group.
- Assign randomization groups at the start.
- Use section display logic or section branching so each arm sees its module.
- Rejoin to a shared closing section for common outcome metrics.
Recipe 5: Reduce order bias inside an arm
Section titled “Recipe 5: Reduce order bias inside an arm”Goal: Experiment arms plus protection against question order effects.
- Assign groups for the main treatment.
- Inside each arm’s pages, enable block randomization (shuffle all or pick N).
- Optionally shuffle modules with section randomization when every module is required.
Recipe 6: Sampled scenarios
Section titled “Recipe 6: Sampled scenarios”Goal: Many scenarios exist, each person sees a subset.
- Put scenarios in subsections with a uniform structure.
- Use section randomization target subsections, mode pick N or pick one.
- Do not mix loose pages into the same randomized section.
- Analyze scenario-level metrics with missingness in mind.
Recipe 7: Personalized follow-ups after the task
Section titled “Recipe 7: Personalized follow-ups after the task”Goal: Same prototype experiment, different debrief by outcome.
- Run the task under compare assets.
- Use display logic on follow-up questions based on success/failure or ratings.
- Pipe scores or group labels into debrief text with
{{variables}}.
Checklist before launch
Section titled “Checklist before launch”- One clear primary driver variable
- Arm content mapped and named consistently
- Fallback behavior defined for missing driver values
- Preview checked for each arm
- Live test responses confirm assignment and prototype mapping
- Results filters prepared for the group or condition variable
Related reading
Section titled “Related reading”- How rules are evaluated for runtime order
- Randomization groups for balanced assignment
- Compare assets for prototype mapping
Common questions
Which recipe should I start with for prototype A/B tests?
Use randomization groups plus compare assets, keep tasks identical, then segment results by group.
Can I run content and prototype experiments together?
Yes, but keep one primary driver variable and document which pages change by arm to avoid confounded results.
Do these recipes require Pro?
Core logic and randomization tools are available for building experiments. Some distribution and AI features used around experiments may be Pro-gated.