Three checks, runnable in month one, that settle whether you are being delivered to.

Why audit at all

The failure you are looking for is silent by design.

Nothing in the system announces that it is under-delivering, because every layer computes correctly from whatever it was handed. A journal missing three fields per trade still produces a weekly summary. A tier ceiling set by habit rather than by drawdown still appears in the panel. A capture-efficiency figure computed from a mistyped fill still renders. The absence of errors is therefore not evidence that the machinery is working, and the only way to establish that it is working is to check a small number of specific outputs against something outside the system. That external reference is what makes these checks worth the minutes they cost — an audit conducted entirely inside the workbook would confirm only that the workbook is internally consistent, which was never in doubt.

Check one

Open an ordinary week and count populated fields on every trade.

The decision trail claims nine fields per trade: instrument, branch, authorised risk, whether the trade was cleared, what happened inside it, whether the plan was followed, whether an override was used, the result, and what the system permitted next. Choose a week that was unremarkable rather than the best one, and confirm all nine are present on every row without exception. Gaps here are by a wide margin the most common defect found, and they are almost always the operator's rather than the system's — which is useful, because it is the one category of defect that is entirely within reach to fix.

Check two

Recompute the tier ceiling from drawdown and compare it with what was used.

Take the current equity and the latest Equity Peak High, derive the drawdown, read the gate row it implies, and note the maximum tier that gate authorises. Then look at the tier actually deployed over the past cycle. The two should agree, or disagree only where an override was logged with a reason. A quiet discrepancy is the highest-value finding available from any of these checks, because it means the constraint that the entire capital-protection claim rests on is not binding in practice — and the failure is invisible in every downstream reading, all of which will describe the deployment that occurred as though it were authorised.

Check three

Reconcile a single closed trade end to end against the broker.

Pick one closed trade and verify the entry, the fill, the stop distance, the exit, the duration and the fees against the broker or platform record, then confirm that capture efficiency, giveback and fee drag follow from those figures. This is the check on capture fidelity, and one trade is sufficient because the errors it detects are systematic rather than random — a stop distance recorded as intended rather than as filled will be recorded that way every time. If the single trade reconciles cleanly, the method of capture is sound; if it does not, the defect is in the procedure and applies to everything. A single reconciliation therefore carries far more information than its size suggests, and repeating it across five trades adds very little.

• Choose an ordinary week — the best week is the least informative sample.
• A silent gap between the gate row and the tier used is the highest-value finding here.
• Capture errors are systematic, so one reconciled trade tests the whole procedure.

Reading the failures

Each check fails in a way that names its own owner.

The three faults are cleanly separable, which is why these particular checks were chosen. Missing fields are a capture problem and belong to the operator. A ceiling that does not follow from drawdown is a configuration or discipline problem — either the gate parameters were altered or the authorisation is being ignored. A trade that will not reconcile is a procedure problem in how evidence reaches the workbook. Only a case where all fields are present, the ceiling traces correctly, and the arithmetic still disagrees points at the system, and that is the rarest outcome by a considerable margin.

The key idea

An audit in month one is worth more than a verdict in month twelve.

The instinct is to wait and judge the system on results, which delays the assessment by a year and then confounds it with market outcomes. These checks assess something narrower and immediately available: whether the machinery is receiving what it needs and producing what it claims. An operator who runs them early and finds three missing fields has recovered eleven months of clean evidence. One who waits has a year of readings and no way to know what they were computed from.