How three threads become one answer — and what the answer throws away.

The construction

Three readings, weighted, summed once.

Each thread is scored independently before any combination happens. Compliance is measured against the plan's own rules; compression is measured from the drawdown and gate record; gate improvement compares the capital state at the month's end against its start. Each becomes a normalised reading, each is multiplied by its weight, and the three are summed into a composite. Nothing in the chain is a judgement, which means the composite is reproducible — the same month scored twice produces the same number, and the same month scored by a different operator produces it too. That reproducibility is the whole reason a verdict is worth stating as a number rather than a sentence.

What the number loses

The same composite can arrive four different ways.

This is the cost of compression and it should be understood rather than worked around. A mid-range score can describe a month with excellent compliance and severe compression; a month with moderate everything; a month with poor compliance rescued by a benign capital path; or a month whose components are all near the middle for unrelated reasons. Those are four different systems requiring four different responses, and the composite collapses them into one figure. An operator who reads the score alone will treat all four identically, and in three of the four cases that response will be wrong. The number is a summary, and a summary read without its components is an opinion with a decimal point.

• Always read compliance separately — at sixty percent it can single-handedly explain a score.
• A mid-range composite is the least informative reading the score produces and the one most in need of decomposition.
• Two identical scores in consecutive months can conceal a complete reversal in what is driving them.

Why compress anyway

A verdict nobody reads is worth less than a verdict that discards something.

Given the losses, the case for the composite has to be made rather than assumed. It rests on the same argument the regime engine makes: three correct numbers requiring interpretation every month is not a stable instrument, because the interpretation will be generous after a good month and harsh after a bad one. One number produced by a fixed function is the same function every time, which makes a series of them comparable and a trend in them meaningful. The composite is not better than its components — it is more *durable*, in the specific sense that it resists the operator's mood in a way three raw readings do not.

The reading discipline

Composite first, then straight to the components.

The practical protocol is short. Read the composite to establish whether the month warrants attention at all. If it is high or low enough to be unambiguous, note it and move on. If it is mid-range, or if it has moved materially from the prior month, decompose immediately — check compliance first because it dominates, then compression, then gate improvement. The composite's job is to decide whether the decomposition is worth doing, not to substitute for it. Treating the number as the end of the enquiry rather than its trigger is the single most common way a well-built composite becomes useless.

The key idea

Build the summary, keep the parts, and know which one you are reading.

The value of a composite is that it makes months comparable. The danger is that it makes different months look identical. Both are properties of the same compression, and neither can be removed — which means the only defence is a reading habit that treats the number as a pointer rather than an answer.