Platform benchmarks - only when they are safe to show

No benchmark is published. That is the honest state today: public benchmarks are switched off. We will not publish a number from a handful of businesses, from test data, or from anyone who has not agreed.

• At least 10 independent, consenting businesses and 200 answered checks in the group.
• No single business supplies more than 30% of the group's checks, so nobody can be picked out.
• Only measured pilot results the owner approved for this use, inside the last 180 days. No internal, demo, fixture or test data.
• Medians, not averages, with an uncertainty range. The same inputs always reproduce the same numbers (methodology benchmark-methodology-2026-09-24.1).
• A person on our team reviews and approves each group. Groups that fall below a threshold later are withdrawn automatically.

It describes a sample of VisibilityAI customers. It is not the market, it does not predict your result, and a difference between groups is an observation, not proof of cause. See the full methodology and the case studies page for what we can and cannot show.