The Ratio Was Never the Number

Revenue per Steward is the revenue a single Steward's overseen business generates, attributed to that Steward alone and reported as an absolute figure — not as a multiple of anything. It is the number that sits underneath the Revenue to Headcount Advantage, Arco's published 10:1 target ratio, and it is a number the ratio does not, by itself, tell you.

The ratio was never the number

Arco has stated the 10:1 target publicly since March. Memo #01: The Difference Between an Automated Business and an Autonomous One establishes it first, as Arco's primary performance benchmark; Memo #03: Overhead Is a Design Choice develops it as the practical expression of eliminating the Coordination Tax; Memo #10: The Stewardship Model restates it as the Stewardship Model's target ratio. It is correctly framed in all three places as a target Arco is building toward, not a result already achieved and audited. That framing is right, and this memo does not change it.

But a ratio, even an honestly labelled target, hides a question it cannot answer: ten times what? A business generating ten times the revenue-per-employee of an incumbent earning modest per-employee revenue is a very different business from one generating ten times the revenue-per-employee of an incumbent already operating at scale. The multiple is the same. The absolute number a Steward is responsible for is not.

Stated schematically, not as a result: if an incumbent in a given vertical generates $X in revenue per employee, the 10:1 target implies a Revenue per Steward figure of $10X at equivalent output. That arithmetic only means something once three things are specified — which is the rest of this memo. $X and $10X are illustrative variables here, not a claim about any Arco business's actual figures.

What counts as revenue, what counts as a Steward, and against what

Three things have to be drawn tightly for the ratio to mean anything: what counts as a Steward, what counts as revenue, and what the comparator is. A loosely defined metric is not measuring the thing it claims to measure — the same discipline the companion Cost per Completed Outcome memo applies to defining "completed."

Steward means what the Stewardship Model already defines: a single competent operator who oversees an agentic stack as architect and exception handler, not executor. A person still performing T2 or T3 execution work themselves, rather than governing the system that performs it, is not yet operating as a Steward in the sense this metric requires — they are closer to the operational centre of an automated business, where the Coordination Tax persists because human decision-making has not left the centre of operations. Counting that person's oversight revenue as Steward revenue overstates the metric before any calculation is done. The Steward is the accountability node for the figure, not its source — the label attaches to the governance role Arco built, not to a hero-operator narrative the model was designed to route around.

Revenue means recognised revenue attributable to the specific business under that Steward's governance, over a stated trailing-twelve-month or annualised run-rate period declared alongside the figure. It excludes pass-through amounts, intercompany or studio management fees, and deferred or uncollected amounts not yet recognised. Where one Steward governs more than one agentic stack, revenue is divided proportionally to the oversight time and accountability each business actually receives — not counted in full against every business the Steward touches. A Steward nominally responsible for three ventures who spends nearly all their attention on one is not generating three ventures' worth of Revenue per Steward; they are generating close to one venture's worth, plus whatever the other two produce on their own architecture with minimal governance. Shared studio services — legal, infrastructure, editorial — are studio overhead, not venture revenue, and belong outside this calculation entirely.

The comparator is the incumbent's revenue-per-employee within the same vertical and at a comparable output unit — the same class of transaction, resolved the same way, in the same market — not an industry-wide average pulled from a different kind of business. Where more than one incumbent benchmark exists, the comparator is the median of the class, named explicitly, not the most favourable single data point available. "Ten times an unnamed incumbent" is not a comparator. It is a multiple with nothing underneath it.

What the number means at each stage of the ramp

Revenue per Steward is not a figure withheld until a business has already proven itself. MTTI is not reported only once a business clears 72 hours — it is tracked from day one to show whether a business is climbing toward that target, and Revenue per Steward should be read the same way. The number is meaningful throughout the ramp; what changes is what it's evidence of.

A pre-revenue business — Simulation status, where a quantitative model exists but there is no live operational data — has no Revenue per Steward figure at all. Not zero: not applicable. Reporting zero implies a failed metric where none yet exists to measure. Once a business reaches Validated or Live status and begins generating revenue, the figure should be reported alongside the current 80 Percent Threshold and MTTI figures at that same moment — not held back until both hit target. A business at 60% autonomous handoff and 12-hour MTTI, generating a modest figure, is not a failure case. It is exactly the case the metric exists to make visible: an operation still absorbing Coordination Tax at the centre, with the gate figures showing how much runway remains before the ratio actually applies.

The distinction that matters is reading the figure correctly at each stage, not gating its publication. A small, hardworking team running an automated business can post an impressive revenue-per-employee figure without ever approaching the Revenue to Headcount Advantage, because the Coordination Tax that ratio is designed to eliminate is still fully present — compressed onto fewer people, who are absorbing it rather than escaping it. The 80 Percent Threshold and MTTI figures reported alongside Revenue per Steward let a reader tell the two cases apart: a high figure with low autonomous handoff and short MTTI is compressed Coordination Tax wearing Stewardship language; a rising figure tracking upward alongside rising handoff rate and extending MTTI is the ramp actually working.

The gaming risk is the same risk that shows up whenever a favourable metric is reported without its definitions made explicit, and the fix is the same discipline Arco applies elsewhere: report the figure only for individuals who genuinely meet the Stewardship Model's definition, define revenue and the comparator precisely, attribute revenue proportionally rather than in full across every venture a name is attached to, and pair the figure with the current 80 Percent Threshold and MTTI numbers. A Revenue per Steward figure reported without those companion numbers is not verifiable, and an unverifiable favourable number is not evidence. It is a number chosen to look good.

The Operator's Verdict

The 10:1 target is the headline. Revenue per Steward is the arithmetic underneath it, and arithmetic does not care about headlines. A business can quote the ratio honestly and still be building toward a number that, stated in absolute terms, would not survive the quoting.

State the absolute figure, by business and by period, alongside its comparator and the current 80 Percent Threshold and MTTI figures — not held back until either hits target, and not attributed to one Steward when it was generated across three. If the number holds up stated that plainly, the ratio built on top of it means something.

Technology changes what is possible. Revenue per Steward determines what the ratio is worth.