# Cohort Benchmarks · Docs · Sales as Code [← All feature docs](https://salesascode.com/docs/library) Feature docs · /cohort-benchmarks # Cohort Benchmarks Place your metric within the anonymized cross-tenant cohort, privacy-safely, suppressed entirely below five peers. [Open Cohort Benchmarks →](https://salesascode.com/cohort-benchmarks) [All docs](https://salesascode.com/docs/library) ## Overview [Cohort Benchmarks](https://salesascode.com/cohort-benchmarks) answers "how do I compare?" without ever exposing another customer. It has two halves: your standing in the cohort we compute for you, and a compare box for a metric we do not compute. Both place a value the same way, with your percentile or band, and the cohort's own quartile shape. They reuse the same privacy floor as the platform's benchmark engine, so the guarantee is consistent everywhere. ## Your standing, with nothing to paste The page loads your standing when you open it. Once a week we compute a set of metrics across every opted-in tenant, closed-won deals, win rate, meetings, drafts, runs, actions, average days to close, and the notes kept per account, and the table shows each one with your own value, your band, the cohort's p25 / p50 / p75 / p90, and how many peers are behind it. A metric we cannot compute for you yet reads "not computable yet" instead of a zero. A metric is only published once at least five opted-in tenants have a value for it, so a thin cohort shows nothing rather than a distribution a small group could be read out of. ## Opting out The same card carries the switch. Opt out and your numbers leave every cohort, and no placement is computed for you; rejoin and both come back on the next weekly pass. Your value is never shown to anyone either way: peers only ever see aggregate quartiles. ## Comparing a metric we do not compute For anything outside that set, the compare box takes your value plus the anonymized peer values you bring, and places you the same way. It is the manual path, and it is subject to the same five-peer floor. ## Privacy-safe by construction The comparison is suppressed entirely when the cohort has fewer than five peers, and a suppressed result carries no distribution data at all, not even the quartiles, so a small group can never be reverse-engineered into an individual. Above the floor, only the aggregate distribution is ever shown, the quartiles and your own placement, never any peer's value. Peers are opt-in, and no cross-tenant identity is ever read. ## What you get Above the floor, the compare box reports your percentile (the share of peers at or below your value), your band (bottom-quartile, below-median, above-median, or top-quartile), how far above or below the cohort median you sit, and the cohort's p25 / p50 / p75. Your standing reports the equivalent from the computed cohort. Either way it is enough to know where you stand and how much headroom there is, and nothing more. ## Deterministic The comparison is deterministic: the same value against the same cohort always produces the same placement, so the number is stable and defensible. It reuses the platform benchmark engine's percentile math and its minimum-cohort-size rule, so the privacy contract is identical to the one the rest of the product honors. --- More for agents: [llms.txt](https://salesascode.com/llms.txt), [full reference](https://salesascode.com/llms-full.txt), [OpenAPI](https://salesascode.com/openapi.json), [sitemap](https://salesascode.com/sitemap.xml).