Is the team using Claude, how often, and how deeply?
By month. Active means 5 or more active days in the month.
What does usage cost per person, and what is spent on licences nobody uses?
USD. Licence is the fixed monthly fee; usage is token consumption.
Are Claude Code suggestions accepted or rejected? Covers Claude Code only; chat has no quality signal. Acceptance shows behaviour, not correctness.
Share of decisions a person made that were accepted. Automatic approvals are excluded. Roles with few decisions move more from month to month.
Cells with fewer than 30 decisions are not rated.
One row per user per month in the selected range.
FR-9| Metric | Definition | Calculation |
|---|---|---|
| Licensed users | People who held a Claude licence on at least one day of the month. | Distinct users with a licence in the month |
| Active day | A day on which a person had at least one interaction with Claude. | Any scored activity on that day |
| Active users | A monthly measure: people with 5 or more active days in the month. There is no weekly or daily active-user figure. | Distinct users with active days ≥ 5 |
| Adoption rate | The share of licensed users who were active in the month. No target is set. | Active users ÷ licensed users |
| Engagement score | Weighted points for what a person did with Claude, using the scoring rules below. Higher weights go to deeper use. | Sum of engagement points |
| Engagement benchmark | A typical day's score for each role: the median daily score of active user-days in the previous three complete months, recalculated each quarter. Q1 2026 uses its own months because there is no earlier data. | Daily benchmark; weekly = daily × 5; monthly = daily × 20 |
| % of benchmark | How a person's month compares with a typical month for their role. | Monthly score ÷ role's monthly benchmark |
| Meeting benchmark | The share of active users whose monthly score reached their role's monthly benchmark. | Active users at 100% or more ÷ active users |
| User segment | Idle: no activity in the month. Light: under 50% of benchmark. Regular: 50% to 150%. Power: over 150%. | Based on % of benchmark for the month |
| Licence cost | The fixed fee for each licence, spread evenly across the days it was held. | USD 20 per user per month (assumed) |
| Usage cost | The cost of tokens consumed, billed on top of the licence. | Sum of usage cost |
| Total cost | Licence cost plus usage cost. | Licence + usage |
| Cost per active user | What the month cost for each active user. It includes the licences of people who were not active. | Total cost ÷ active users |
| Idle licence | A licence held at the end of the range by someone with no activity in the last two months of the range. | No active day in the last two months |
| Days idle | Days since the person's last active day, or since the licence was assigned if they never used it. | Range end − last active day |
| User decision | An accept or reject of a Claude Code edit made by a person at a prompt. Edits approved automatically by settings or hooks are excluded. | Decision source is a user choice |
| Acceptance rate | The share of user decisions that accepted Claude's suggested edit. | Accepted ÷ (accepted + rejected) |
| Auto-approved share | The share of all Claude Code edits approved without a person reviewing them. | Automatic decisions ÷ all decisions |
Each activity is turned into units, multiplied by its weight, and capped per person per day. Caps stop one high-volume activity, such as chat messages, from dominating the score.
| Activity | Category | Weight | Counted as | Daily cap (points) |
|---|---|---|---|---|
| Chat messages | Basic prompt | 1 | 1 unit per 2 messages | 10 |
| Files uploaded | Summarise / extract | 2 | Each file | None |
| Web searches | Summarise / extract | 2 | Each search | 10 |
| Artifacts created | Generate | 3 | Each artifact | None |
| Office add-in messages | Generate | 3 | 1 unit per 2 messages | 9 |
| Projects used | Advanced / agentic | 4 | Each distinct project | None |
| Connector calls | Advanced / agentic | 4 | Each call | 12 |
| Claude Code sessions | Advanced / agentic | 4 | Each session | None |
| Cowork sessions | Advanced / agentic | 4 | Each session | None |
Example: 12 chat messages and one Claude Code session score 6 × 1 + 1 × 4 = 10 points.
Daily benchmark in points. Each quarter uses the previous quarter's data; Q1 has no earlier data, so it uses its own. Weekly is daily × 5 and monthly is daily × 20. Q4 applies from the October refresh.
| Visual | Requirement | What it shows |
|---|---|---|
| Headline figures | FR-1 | Six figures for the last month of the selected range, each compared with the month before. A green arrow is an improvement and a red arrow a decline. For the two cost figures, red means spend rose and green means it fell. |
| Active users against licensed users | FR-2 | How many licensed users were active each month. The gap between the two lines is licences not in regular use. |
| User segments by role | FR-3 | How each role splits into Idle, Light, Regular and Power users in the last month of the range. Selecting a role filters the page. |
| Monthly cost, licence and usage | FR-4 | Each month's cost split into the fixed licence fee and usage. The line below shows cost per active user, so a rising cost can be read against how many people use Claude. |
| Idle licences | FR-5 | Everyone holding a licence who has not used Claude in the last two months of the range, longest idle first. These are the first candidates to reassign or drop. |
| Acceptance rate over time, by role | FR-6 | The share of Claude Code suggestions people accepted, by role and month. Roles with few decisions swing more from month to month. |
| Acceptance rate by language and edit tool | FR-7 | Where Claude Code suggestions are accepted most and least, by file language and edit tool. Cells with fewer than 30 decisions are left unrated. |
| User detail | FR-9 | One row per person per month with segment, activity, cost and acceptance. Use it to check the people behind any figure on the page. |