What it does, where it stops, and how your data is handled.
Two rules on different platforms whose conditions can both be satisfied by the same user. A US enterprise customer on mobile matching a LaunchDarkly flag and an Optimizely experiment at the same time is a collision.
LaunchDarkly flags, and Optimizely audiences and experiments, via JSON export. Other platforms are not supported today.
Statsig is the most requested. Nothing is committed — the order follows what teams actually ask for.
No. It means two treatments could have reached the same people, which is worth knowing before you interpret a result. Whether it changed the outcome is a judgement your team makes with production data.
One rule used a construct the parser cannot yet normalize. It is reported as needing manual inspection and never silently counted as clean. See the interpretation guide.
You can, but the value is limited. The engine exists to compare across platforms; within one platform your own tool already has that view.
The diagnostic is free ($0, no account). A free registered account ($0) provides saved projects, baselines, and history. The Founding Plan is $29/month for full entitlements including 10 projects, change tracking, exports, and 24-month retention. See pricing.
No. Analysis runs on offline JSON you paste or upload. There are no tokens to issue and no permissions to grant.
For anonymous diagnostic audits, it is parsed and compared in memory, then discarded. It is not written to a database and its values are not written to application logs. Registered users who explicitly save projects, baselines, or audit runs have that data durably stored under their account until deleted or pruned according to retention policy. See the data path.
No. FlagConflict is strictly read-only and has no write path back to LaunchDarkly or Optimizely.