Cross-Fund Concentration
When concentration is a strategy rather than a problem
Concentration that accumulated while nobody was watching and concentration a firm chose on purpose and disclosed to its LPs can look identical in the numbers. Which of the two you are looking at decides whether it is a finding or a strategy, and it is why two of these four distributions are scored and two are only described.
The decision
Sector concentration and single-company exposure carry flags and affect the Cross-Fund Concentration Score. Stage and geography are shown and not scored.
This is a deliberate methodology choice rather than a limitation waiting to be lifted, and the reasoning is worth understanding because it is the principle the whole engine rests on.
Emergent patterns and declared ones
The distinction
The engine scores concentration that accumulates without anyone deciding it. It describes concentration that a firm chose on purpose and told its LPs about.
Sector concentration is emergent. No firm writes down that 60 percent of its capital will end up in one sector; it arrives through a sequence of individually reasonable investments. Single-company exposure is the same, assembled from follow-ons each of which was the right call at the time. Both are worth flagging precisely because nobody chose them and nobody is watching them.
Stage concentration is different. A seed fund concentrated at seed has not drifted into anything. That is the fund. Geography is the same: a firm that raised on a thesis about one region and deployed there has executed its strategy, and a model that flagged it for concentration would be scoring a firm down for keeping its word.
Why scoring them would produce noise
Suppose the engine did flag stage concentration. A single-stage specialist would be flagged permanently, from its first investment to its last, with no available action. The flag would appear on every report and mean nothing, and its presence would teach a General Partner to ignore the flag list.
That is the real cost. A flag list is only useful if every item on it is worth reading, and the fastest way to make one useless is to fill it with conditions the reader has already decided about. Restraint about what to flag is what makes the flags that do appear worth attention.
What descriptive actually gives you
Not scoring is not the same as not showing. Both distributions are rendered across the firm's deployed capital, and they are frequently the most interesting thing on the page, because a General Partner can bring context the model cannot.
A firm that intended a stage-diversified book and sees its capital bunched at one stage has learned something. A firm that describes itself as continental and finds most of its capital within a few hours of its own office has learned something. Neither is a finding the model could have made, because in both cases the distribution is only meaningful against an intention the model does not hold.
The point of a descriptive view is that it supports the reader's pattern recognition instead of substituting a threshold for it.
Where the line could move
The emergent-versus-declared distinction is a judgment rather than a law, and there are real cases that sit near the edge. A firm that declared multi-stage and deployed almost entirely at one stage is a strategy-execution gap of exactly the kind the engine scores elsewhere.
The Portfolio Efficiency engine catches part of that at the fund level by reading declared stage against the rest of the fund's configuration. Whether the firm-level version deserves a flag is a methodology question rather than an arithmetic one, and the honest answer today is that it is described rather than scored.