Windtunnel method
Shortlist rate
Shortlist rate is the share of valid AI-recommendation answers that place the brand among the five ranked recommendations, computed per model and weighted across scenarios.
Five ranked recommendations, parsed exactly.
In an AI shortlist test, the candidate message is supplied as untrusted context to brand-neutral shopping situations, and the model is asked for exactly five ranked recommendations. The reply is parsed deterministically: exact brand and alias matching, no fuzzy matching, and a malformed answer stays invalid rather than being retried until it looks right.
The rate is computed separately per model, because each model is its own population, and weighted across scenario clusters so no single scenario dominates. As a proportion it carries a Wilson 95% confidence interval.
Where the shortlist family lives.
Shortlist rate is one of the two rates reported by the AI recommendation half of a Message Lift test. The Message Lift section on the landing page shows how the test is built; the term first appears in results as shortlist rate, then as shortlist lift when two messages are compared.
Three things a shortlist rate is not.
- Not a purchase prediction. It says where a model placed a brand in a simulated recommendation, not what buyers will do.
- Not search visibility. Nothing here measures a search engine; the shortlist is the model’s own ranked output under disclosed prompts.
- Not an absolute verdict. Results are SIMULATED and labeled, and they compare contexts inside one test, not brands across tests.
Labels travel with numbers: figures traced to stored answers from real model routes are MEASURED; Message Lift results are SIMULATED and comparative only; anything below the n≥30 gate or drawn from a single prompt cell is DIRECTIONAL. No figure on this site ships without its label, sample size, and date.