Compare Versions in a Face-Off
Put 2 to 5 versions of your copy in front of the same synthetic audience and let three independent reviewers pick, or abstain when the gap is noise.
A face-off puts 2 to 5 versions of the same content in front of one audience and ranks them with the same three reviewers a single check uses. It is how you settle an internal debate with evidence instead of opinion.
When to run one
Reach for a face-off when the disagreement is about which version to ship, not whether a single version is good enough.
- Two headlines are stuck at a standstill in a thread.
- A writer and a stakeholder each have a preferred subject line.
- You have three openings for the same email and no principled way to choose.
If instead you have one version and want to know where it breaks, run a single check and read the findings.
Start a face-off on the canvas
Add each version as its own content node. Multi-select two or more nodes of the same kind, and a ⚔ Face-off · N pill appears at the top of the canvas, where N is how many versions are selected (up to 5). Click it to run.
Start a face-off in linear mode
Open Linear mode from the top of the project, then switch to the Compare versions tab. Paste each version into its own box, using + Add another version (up to 5) as needed, and click ⚔ Start face-off. The audience is set for you, so there is nothing else to configure.
How it ranks
A face-off uses the same reviewers, the same pairwise-in-both-orders comparison, and the same verdict tiers as a single check. See How content checks work for the mechanics.
The verdict badge tells you how much to trust the ranking: Reviewers unanimous, Reviewer majority, Reviewers split, or Within noise.
When the reviewers abstain
Often the honest answer is that the versions are too close to separate. Studio says so rather than forcing a winner:
"The gap between these versions is within noise, so the reviewers refused to force a call. That is itself useful: stop debating, pick either one, and spend the time on a real test."
An abstain is a result, not a failure. It tells you the choice does not matter enough to keep arguing about, so you can ship either version and put the saved effort into a real test.
Share the verdict
Every face-off produces a shareable results page. Send the link into the thread that started the debate and let the verdict, not seniority, close it. See Read results and verdicts for what the public page shows.
Next steps
- Read results and verdicts: the ranking rows, judge dots, and exits after a face-off.
- Choose an audience: make sure every version faces the audience you actually care about.
- Multi-flow tasks: the platform equivalent for structured A/B testing across audiences with a full scorecard.
How Content Checks Work
How Studio turns your copy into survey probes, pinned findings, fix variants, and a three-reviewer verdict, with honest evidence tiers on every claim.
Choose an Audience for a Check
Pick a calibrated, census-framed, or model-estimated audience for your check, or describe your own, and see how each one is anchored and labeled.