Method · where every number comes from
Every figure in this report is counted, attested or estimated — and says which.
counted·attested or estimate·assumption — and how to firm it up
What came back, by what caused it$2,700 a month
$2,070 seats Aristo woke up · causal$540 hours people counted themselves · attested$90 license spend back at work · counted
The move: Read every source — the return, the causal claim, hours, protection and the covenant.
The covenant — every figure is an aggregate; no page of this report can name a person.
The Value ReportThe illustrative organizationrun 4 · 24 SeptemberSee it live: /v2/value?q=methodSheet 1 of 2
Appendix · where every number comes from
§6 · Methodology
Where every number on these pages comes from.
Plain words, no marketing. A figure that is counted says counted; a figure the people themselves counted says attested; a figure that rests on a modeled price says estimate and names the model; a figure Aristo cannot reach is shown as a dash with the exact thing that would fill it in.
Every number on these five pages, sourced
the return · the causal claim · hours · protection · the covenant
The return — the cover figure is the sum of exactly three things, and nothing else is allowed in: holdout-proven seats × your seat price, attested net hours × your hourly value, and license spend a verified reassignment put back to work. No estimate, no projection and no planning parameter is inside it.
The causal claim — on day 0 a deterministic share of your seats ready to activate becomes a never-nudged control group: stable, not hand-picked, not resettable. After the window, lift = the helped arm's activation rate minus the control arm's, with a 95% Newcombe confidence interval. Proven seats come from the interval's LOWER bound, never the point estimate, so statistical noise cannot manufacture a seat. Below 95% confidence this report says directional and prints no dollar. Negative lift would be reported as negative.
Attested hours — net minutes people recorded themselves, gained minus lost, converted to hours. The ledger is suppressed whole below a floor of five attesters: fewer than five and this report shows no hours figure at all rather than one a small group would make identifiable. Aristo's own activity-based projection of hours saved exists in this product and is deliberately not printed on these pages, because a parameter multiplied by a headcount is not evidence.
The vocabulary — Microsoft's Copilot Analytics Labs publishes agent-assisted hours as its value currency. Where this report presents attested or counted time it uses that same unit, so a Labs read and this page benchmark 1:1. The distinction worth stating out loud: Labs estimates the value; Aristo pairs the number with a control group it never touched.
Counted work — a win is one job somebody finished with Aristo alongside them, counted at the moment it finished; a completed move is a leader-approved strategy move that ran to the end, never an approval; a skill reuse is one download of a shelf skill somebody else built. All three are aggregate event counts from your own tenant, and none of them is a survey answer.
Protected, not returned — steered credits and avoided re-forges are counted events multiplied by the published medium-task envelope. That makes the dollar a modeled envelope, so it is chipped estimate Counted inputs multiplied by a published rate: a labeled estimate, never a measurement., kept on Question 4, and never added to the cover figure. The rerouted license line appears on Question 4 as a cross-reference only; it is counted once, on the cover.
The prices — Both prices are assumed until you set your own in Settings: $30/seat is Microsoft’s published price, used as your Copilot seat price, and $45/hour is the assumed hourly cost of your people. Every dollar derived from an assumed price is labeled assumption An assumed figure, used until your own replaces it. and sharpens the moment your own figures are entered.
What this report will not do — it will not count a seat twice (a seat becomes reassignable only after the activation arm failed on it, so activation and reassignment are mutually exclusive by construction), it will not claim baseline-active Copilot spend as something Aristo produced, and it will not let a good month quietly re-baseline a bad one. The return is allowed to go down.
The covenant — every figure is an aggregate. No page of this report can name a person, the control group is never identified, and the people whose hours are counted are counted in a group of five or more or not at all.
The Value ReportThe illustrative organizationSheet 2 of 2