Governance runs on six clocks, not one.

Why more than one

Every signal has a frequency, and a review can only catch its own band.

A review cadence is a filter. A daily loop sees execution and misses drift entirely, because drift moves too slowly to register between two consecutive sessions. An annual review sees structural change clearly and cannot detect a sizing error that occurred on Tuesday. Neither is more rigorous than the other; they are tuned to different frequencies, and a system running a single cadence is deaf to everything outside that band. The six horizons exist because the failures MARS is designed to catch do not share a timescale. The same logic explains why adding a faster loop never compensates for a missing slow one: increasing sampling frequency cannot recover a signal whose period is longer than the entire review window.

What each one owns

Six horizons, six distinct questions.

Each horizon answers something the others structurally cannot, and the assignments are not interchangeable:

• Daily — was the process followed, and is exposure reconciled?
• Weekly — is branch expectancy healthy, and did the week earn its risk?
• Monthly — what regime was this, and does attribution still hold?
• Quarterly — are parameters and strategy allocation still correct?
• Annual — is the capital plan and risk architecture still fit?
• Structural — has the machine itself changed shape, on any timescale?

The collapse failure

One cadence always becomes the weekly one, and drift survives it.

Left to preference, review cadence collapses toward whatever interval feels responsive — in practice, weekly. The weekly loop is genuinely valuable and it has a specific blind spot: a metric declining two percent per window produces no alarming weekly reading at any point, and a trader running weekly-only review can watch six months of steady deterioration one acceptable week at a time. Drift is invisible at the frequency most traders sample. That single fact is most of the argument for keeping the slower horizons alive even when nothing appears to be happening.

The opposite failure

Running every horizon at once produces reactive noise.

Over-sampling is the mirror error and it is common among diligent traders. Reading structural diagnostics daily converts long-horizon signal into short-horizon noise — the readings move, the operator responds, and the response contaminates the very series being measured. Each horizon has a natural sampling rate below which its output is not yet information, and the discipline of not looking is as load-bearing as the discipline of looking. A quarterly parameter review consulted weekly stops being a quarterly review and becomes a weekly opportunity to fiddle.

Authority by horizon

Faster loops observe. Slower loops are allowed to change things.

The horizons are not equal in what they may authorise, and the ordering is deliberate. Daily and weekly loops are permitted to report, reconcile, and flag — they do not change rules. Parameter changes belong to the quarterly horizon; architectural changes belong to the annual one. This is the same logic as the gate ladder applied to the review process itself: the loop closest to the emotion of recent results has the least power to act on them, and the loop furthest from it carries the authority to rewrite the system. A useful test of whether the horizons are genuinely separate is to ask what the last rule change was and which review produced it — if the answer is 'a bad week', the authority ordering has already collapsed.

The practical shape

Six horizons is less work than it sounds, because most of it is already recorded.

The cadence reads as heavy and is mostly assembly rather than analysis. Because the workbooks capture evidence continuously, the weekly review consumes what the daily loop already reconciled, and the monthly review consumes the weeks. Nothing is re-derived. The genuine time cost sits at the quarterly and annual horizons, which are the two most traders have never run at all — so the marginal effort of adopting the full cadence is smaller than the list implies, and it is concentrated in exactly the reviews that were previously missing.