Compare spend per person across teams

# Spend per employee

## Purpose

Compare teams on their spend per person, and spot those that stand well away from the average.

## The limit that shapes this skill

**The MCP does not know headcount per team.** `get_users` returns each user's name and email, but not their cost center, team or role. So it is impossible to count the people in a cost center.

Two ways to proceed, to state to the user:

- **A. With your headcount** — the user supplies the number of people per cost center. It is the only way to get a true spend per head.
- **B. Per person who spent** — without headcount, spend is divided by the number of employees with at least one expense. That is not a headcount: a team of 30 where 4 people spend will be divided by 4.

Ask for headcount before choosing B.

## Step 1 — Spend by cost center and by employee

State the exact period.

```json
{
  "tool": "spendesk_analyze_spend",
  "params": { "fromDate": "YYYY-MM-DD", "toDate": "YYYY-MM-DD", "groupBy": ["costCenter", "employee"], "limit": 50 }
}
```

- Each cost center row gives its amount; its `breakdown` lists the employees who spent there.
- `limit: 50`: by default only 10 rows come back per level, the rest goes into "Others". If an "Others" row remains at employee level, the number of people who spent is understated: say so.
- The `employee` dimension ties an expense to the person who incurred it. Supplier invoices do not always have an employee: they can appear with no name.
- Credit notes included, netted off.

## Step 2 — Calculate

- Method A: the center's spend ÷ supplied headcount.
- Method B: the center's spend ÷ number of rows in the `breakdown` (excluding "Others").

Then the overall average and, for each center, its ratio to that average:

- 🔴 more than twice the average
- 🟡 1.5 to 2 times
- 🟢 0.5 to 1.5 times
- ⚪ less than 0.5 times

## Step 3 — Explain a gap

For a center that stands out, break down by supplier:

```json
{
  "tool": "spendesk_analyze_spend",
  "params": {
    "fromDate": "YYYY-MM-DD", "toDate": "YYYY-MM-DD",
    "groupBy": ["costCenter", "supplier"], "limit": 50,
    "filters": { "costCenterIds": ["<center id>"] }
  }
}
```

With this filter, keep `costCenter` first and read the `breakdown` of the center's row: `summary.totalAmount` counts the matching expenses in full, including their lines on other centers.

## Report

```
Spend per employee — from <date> to <date>
Method: supplied headcount / people who spent

| Cost center | Spend | People | Per person | vs average |

Gaps to look at: center, main supplier, amount
```

## Limits to state

- With method B, the denominator is not a headcount: do not present the result as "spend per employee on payroll".
- Some teams spend more by nature (field sales, events).
- Amounts in functional currency, including VAT, on the document date.

## Observed result (demo account, 5 October 2026)

`get_users` returned each user's ID, name and email, with no cost center or team: the per-team headcount that the previous version of this skill computed is not possible.

## Example prompts
- "Here is our headcount by team: what is the spend per person this quarter?"
- "Which teams spend the most per person who spent?"
- "What explains the Marketing team's level of spend?"