Why being excellent at one thing does not raise the Architecture Score

The claim

A firm cannot earn a high Architecture Score by being strong on one engine while weak on another. Strength across all three is required.

That is the whole property, and it is the thing that makes the composite worth having. It is also not how most people assume a composite works.

Why not an average

The instinctive way to combine three scores is to average them, and an average has a specific defect for this purpose: it lets strength in one place buy off weakness in another. A firm with excellent portfolio efficiency and serious concentration problems would land in the same place as a firm that is merely adequate at both.

Those two firms are not in the same situation. The second has no acute problem. The first has one, and averaging has hidden it behind a number that looks reassuring.

The composition used here does the opposite. A weak component pulls the composite down harder than a strong component lifts it, so the score tracks the weakest dimension more closely than the strongest. Improving something you are already good at moves the number very little. Fixing the thing you are worst at moves it a lot.

The boundary that makes it concrete

The clearest illustration is the edge case. A genuine zero on any one of the three components takes the Architecture Score to zero, regardless of how strong the other two are.

In practice a real zero is rare, because it requires a configuration that fails comprehensively on one dimension. The reason to state it is that it shows what kind of composite this is: one where the three inputs are treated as necessary conditions rather than as contributions to a pot.

What this means for reading your own score

Three practical consequences.

The score points at your weakest dimension. If you want to know why the number is what it is, look at the lowest of the three engine scores first. It is doing most of the work.

Improvement is not linear. Moving your weakest component up ten points will change the composite more than moving your strongest up ten. This is not a scoring quirk to be worked around; it is the model saying that balance is the thing being measured.

A middling composite can hide a wide spread. Two firms at the same Architecture Score can have very different profiles, one consistently moderate and one excellent-and-poor. The composite alone cannot distinguish them, which is exactly why the drill-down exists and why the engine scores are shown alongside rather than underneath.

What the composition is not doing

It is not asserting that the three dimensions are equally important to a firm's returns. Nothing in the model claims that, and the evidence would not support it.

It is asserting something narrower: that a firm claiming to be well-architected should be able to withstand examination on all three, and that a composite which allowed one to be traded against another would not be measuring architecture. It would be measuring the best argument a firm can make about itself.