# Evidence grades

The tactic library reports facts and cites evidence; it does not hold opinions. Every tactic carries one of three grades, and grades are computed from the evidence, never assigned by a person. A recompute job derives them from the evidence; nobody edits a grade by hand.

| Grade | Meaning | How it is earned | What we say | What we never say |
|---|---|---|---|---|
| **Observed** | Top SaaS sites in the corpus use it. This counts adoption; it does not prove an effect. | Seen on corpus sites by two independent annotation passes. | How many corpus sites use it, where on the page, with examples you can look at. | That it will change your numbers. Popular is not the same as proven. |
| **Tested** | Outside evidence supports it: a published experiment, or tracked sites that added it and kept it. | At least one experiment-grade evidence link, or a change kept on 3 or more tracked sites. | That evidence exists, with links to the sources so your agent can read them. | A number for your site. Experiments ran elsewhere, on other traffic. |
| **Measured** | The effect was measured on connected sites after the change shipped. | Outcome rows on at least 2 sites, each with 500 or more sessions per period. | The measured change, with the metric, the number of sites and the periods behind it. | That the same change will repeat on your site. It is a measurement, not a promise. |

Grades only go up when evidence arrives, and a tactic whose annotation quality drops is demoted to draft automatically (see [sell-readiness](/docs/sell-readiness)).

## The rule we enforce

- No page, doc or tool response states a numeric result for a tactic unless its grade is **measured**.
- The grade is shown next to every claim, on the website and in every tool response.
- A measured result is reported with its metric and the number of sites behind it. It describes what happened on those sites; it is not a forecast for yours.

The rule is enforced in code: public pages render claims through a single helper that refuses to print a number for anything below measured, and a test fails the build if a number appears next to a claim anywhere else.

## When a tactic is published

A tactic stays in draft, and is never shown publicly or returned to agents, until:

- it has evidence from at least 2 independent sources **and** is observed on at least 5 corpus sites, **or**
- it has at least 1 experiment-grade source **and** is observed on at least 3 corpus sites.

Only published tactics appear on [the tactics pages](/tactics), in [llms.txt](/llms.txt) and in tool responses.