Monthly department chargeback — March 2026 – August 2026

Executive summary

The estate's position across March 2026 – August 2026.

Spend trend estimate

Metered consumption month by month across March 2026 – August 2026, each month carrying its own basis.

Metered consumption rose 18% across 5 completed months, from $229,400 in March 2026 to $269,760 in July 2026, and 0 of the 6 months shown are measured rather than estimated.

The final month is a month-to-date figure standing beside completed months. It is excluded from the movement asserted above.

The current month is drawn open because it is a month-to-date figure standing beside completed months. It is excluded from the movement asserted above — comparing a part-month to a whole one is the single easiest way to read a fall in spending that has not happened.

Cost per user counted

Counted Copilot licence per head, by team, across every department.

Product carries $17 of Copilot licence per head, 1.5× what Manufacturing carries at $12 — not because its seats cost more, but because 260 of its 455 people hold one against 210 of 537.

This is a TEAM total divided by a TEAM headcount, at five people or more. It is not a per-person figure, there is no per-person figure in this product, and no filter combination on this page produces one.

The consumption allocation is deliberately not divided by headcount here. That apportionment multiplies a flat per-user rate by a department size, so dividing it back out returns the same rate for every team — a column that looks like a finding and can never contain one.

1 row is not ranked here: the privacy-merged bucket has no headcount to divide by, and a team with no counted seat join has no licence bill. They are named in the department table below with an em dash rather than a guess.

Department efficiency counted

Every team ranked on the share of its counted seats that are working, across every department.

IT runs the tightest estate at 82% of its 250 seats in use, while Sales leaves $6,300 a month on 210 seats nobody has touched.

The two money columns are on different bases. Licence cost is counted seats at your seat price; the allocation is an apportionment of metered funding, labelled an estimate everywhere it appears. Adding them together would state a total this organisation does not owe.

Ranked on the share of a team's counted seats that are genuinely active — the one column here with no estimate and no price inside it. The two money columns are on different bases and are never added: licence cost is counted seats at your seat price, the allocation is an apportionment of metered funding.

1 row carry no rank: a privacy-merged bucket or a team below the utilisation join's five-person floor has no counted use to rank. They are shown with an em dash rather than dropped, so the table still adds up to the estate.

There is no “change since last month” column and no per-team trend line, and that is a missing fact rather than a missing feature: Aristo stores the estate month by month, but every per-department record it keeps — the utilisation join, the directory headcount, the seat grid — is a single current snapshot overwritten on each refresh. So these teams can be ranked against each other today and cannot yet be ranked against themselves last month. A per-department monthly record is the one upstream change that fills both columns.

Usage vs cost counted

Seat use against licence cost, one mark per team, sized by counted seats.

Sales, Operations, Marketing sit in the low-use, high-cost corner: together $38,400 a month of licence against seat use at or below the estate's median of 66%.

Every mark is a team of five people or more. The bubble is drawn as an AREA in counted seats, so a team with four times the seats gets four times the ink — scaling the radius instead would quadruple the visual weight of a doubled quantity.

The usage axis runs the full nought to a hundred whatever the data does. Auto-scaling a percentage axis to its own spread makes a five-point difference between teams read as a chasm.

1 row is not plotted: the privacy-merged bucket and any team without a counted utilisation join carry no coordinates, and a point cannot be invented for them.

Product utilisation counted

Every paid Microsoft product, in counted seats — bought, assigned, and never handed to anyone.

Microsoft 365 Copilot carries the largest unassigned block: 160 of its 4,200 paid seats have never been handed to anyone, which is $4,800 a month at your seat price.

Seats are counted for every paid product. Money appears only where this product holds a real price for the product — Copilot today — because an invented price makes a total look precise and be wrong.

5 of the 6 products shown are reported in counted seats and deliberately never dollarised. Their headroom is real; its value is not yet knowable.

Variance analysis projection

What moved across March 2026 – August 2026, against the prior month, the prior run, your cap and the labelled projection.

The largest movement across March 2026 – August 2026 is that metered consumption went from $246,780 in May 2026 to $258,420 in June 2026, up $11,640 (4.7%). Alongside it, cold seats fell from 1,283 at the previous run on 2026-07-30 to 1,280 at this one, down 3 seats — $90 a month of licence.

A movement between two months on DIFFERENT bases is not purely a movement in spending — one endpoint counts Aristo-mediated consumption and the other counts everything your billing export sees. Every row states the basis of its weaker endpoint, and says so in words where they differ.

Seat and cohort movements are measured between RUNS, not between calendar months, and are labelled that way. A run is when Aristo counted; the calendar is not what moved.

Estimated months are included in the comparison and marked, rather than dropped. Dropping them would leave a tidy series with a hole in it and no sign that the hole was ever there.

Recommendations

Selected from the positions ranked above. Every figure in them is one of the figures already in this document.