Grow
Grow finds documentation gaps from your own observation data — search queries by impressions, high-exit pages, recorded task failures — and proposes bounded, evidenced work. It does not promise traffic or conversion.
How a proposal is made
- Connect an observation source (your search console or product analytics) with a stated window.
- Signals below a minimum threshold are ignored; each proposal cites the counts and the window it came from.
- An owner authorizes an experiment; the system assigns visitors deterministically with a declared holdout.
- Evaluation reports observed differences with confidence intervals and a stopping rule. Overlapping intervals are reported as inconclusive.
- Stopping an experiment restores the canonical content and records the result.
What we will not claim
- Search ranking gains: those depend on engines we do not control.
- Conversion improvements from a single noisy sample.
- Growth without a measurement window you can inspect.
Actions
Actions
- View experiment example
Opens the workspace experiments surface, which needs a running control plane and shows an honest state when none answers. No invented results appear on this public page.
- Create campaign
Campaigns are created inside a workspace after an observation source is consented to; the preview flow is the documented starting point and costs nothing to begin.