Four surfaces is the floor. Below that, it records rather than governs.

The distinction being tested

Governing means an output constrains the next action.

A system records when it stores what happened, and governs when it produces something the operator is obliged to act on before the next trade. Almost every narrow adoption of the architecture records superbly and governs not at all, which is not a criticism — an execution laboratory is supposed to explain rather than constrain. But it does mean the modes are not points on a single scale of completeness. They divide into those that close a control loop and those that do not, and the dividing line falls at a specific and identifiable set of four surfaces.

Why not three

Each of the four is load-bearing for a different reason.

Remove the throttle and the gate produces a capital state nobody converts into a deployment figure, which leaves an accurate description and no instruction. Remove the gate and the throttle has no capital state to read, so authorisation collapses back to whatever the operator felt like risking. Remove weekly aggregation and the gate is classifying against a peak derived from unstructured data. Remove the journal and none of the three above it has an input. The four are not a recommended package; they are a chain in which every link is the only thing supplying the next one. No fifth surface anywhere in the architecture can substitute for a missing member, because none of the others produces a capital state or an authorisation.

What the floor actually costs

The minimum loop is cheap in time and expensive in appeal.

In practice the four surfaces are a weekly obligation measured in tens of minutes rather than hours, because three of them are largely mechanical once the journal is current. That makes the floor unusually accessible — and it is still the mode operators are least likely to choose, for a reason worth naming. It produces almost nothing pleasurable to look at. There are no efficiency charts, no regime labels, no percentile fans; there is a tier, a pool, and a directive. It is the part of the system that works and the part nobody buys a system for, and the two facts are related — a constraint has no presentation layer because its entire content is a number the operator is not allowed to exceed.

The common inversion

Most partial adoptions start at the top of the stack rather than the bottom.

The predictable pattern is to begin with the analytics — the expectancy monitor, the efficiency laboratory, the benchmark comparison — because those answer the questions that felt most pressing at purchase and because their outputs are immediately interesting. The result is an operator with a detailed account of their expectancy, their capture efficiency, and their position against the envelope, deploying capital by the same instinct as before. The diagnosis improves considerably and the behaviour does not, which is the specific failure mode that makes people conclude a measurement system did not work for them. It worked exactly as configured; what was never installed was the part that says no.

• Journal, aggregation, gate, throttle — remove one and the loop stops closing.
• The floor is inexpensive in time and offers almost nothing to look at.
• Starting at the analytics improves diagnosis while leaving deployment untouched.

Building upward from the floor

Everything above the four makes the same decision better rather than adding decisions.

Once the minimum loop is closing, additional surfaces do not introduce new obligations — they raise the quality of the one authorisation already being issued. The Cycle Command Console sharpens remaining capacity from an estimate into a figure. Volatility tooling establishes whether the intended management model suits current conditions. The MAE/MFE Lab explains why expectancy moved. The SDE and benchmark separate variance from decay before the operator responds to either. This is a coherent order of adoption precisely because each addition improves an existing decision rather than creating a new one to maintain.

The key idea

The floor is a genuine adoption, not a trial version.

Running four surfaces is not a partial implementation waiting to become a real one. It is the smallest complete governance loop, it addresses the failures that most reliably destroy workable methods, and an operator who runs only that for a year will have a materially better-governed account than one running eight surfaces without a throttle. The stack is worth building out, and the order matters more than the pace: close the loop first, then spend the remaining effort making its single output more accurate.