The load-bearing sentence in Anthropic's documentation is not a metric definition but a piece of advice: owners are told to monitor individual usage patterns to identify power users who can share best practices, or to spot team members who might benefit from additional support [10]. As enablement copy that is unremarkable. As instruction to a director defending a seat budget in a flat headcount year, it describes a ranked list with names on it, and nothing further needs to be built for the second reading to win.
The contribution beta improves the measurement and extends its reach at the same time. Accepted lines count what a person took from a suggestion box, which is a thin proxy for work that shipped; the beta closes part of that distance by correlating Claude Code session activity with GitHub commits and pull requests [8]. What comes out is a per-person view of merged pull requests and committed lines, reported both with and without assistance [7], which is more honest about output and is also the precise shape of evidence a manager would use to argue that a named engineer is underusing a paid tool [16]. Fidelity to shipped work and usefulness as a performance instrument rise together, and that tradeoff is worth stating plainly.
The dashboard sits behind Owner and Primary Owner accounts [4], one step removed from where week-to-week performance conversations happen. The export closes that gap anyway. The dashboard's own figures clear when the calendar month turns [9]; a downloaded CSV of email addresses and line totals does not [3]. Twelve exports across a year yield twelve months of named history out of a system that displays one [15], kept under whatever access rules apply to the folder it lands in.
The Value tab is franker than most vendor ROI screens, showing estimated productivity lift, cost per commit and estimated annual value with each formula inline and each input adjustable [5]. That visibility also exposes what the estimate rests on. The board-deck version reads as a clean upward story: acceptance climbing, cost per commit falling, annual value looking substantial. It is incomplete because the inputs are chosen by the sponsor of the purchase, and the counter underneath rewards the volume of accepted suggestions rather than whether they were right [2].
What the record does not settle is worth saying rather than glossing. The help page describes owner-side access and setup, and sets out no employee-facing view, no notice and no opt-out for user-level metrics [17]; it also confirms the feature is unavailable on individual Pro and Max plans [12]. So the governance position defaults to whatever the first month of use produces. The choice available this quarter is narrow and concrete: who holds Owner, and whether user-level lines accepted is ruled out of scope for review before anyone has seen the list. The consequence arriving next quarter is the first calibration meeting where somebody has the CSV open.
