Scores
Composite, defined
A composite is a total built from parts by a rule; the rule is the part that matters and the part you are least likely to be shown.
Guides on Scores: How much weight a rating tool result deserves, Every part of a rating result, and what each one is for, What can and cannot be compared, and how
A composite score is a total built out of several parts by some rule: proportion, symmetry and presentation go in, one number out of ten comes out. The parts are the axes, and the rule that turns them into one number is, almost without exception, the part a results page never shows you.
Why the rule is the whole game
Two tools can score the identical submission on the identical three axes and land on different totals, purely because they combine them differently. An equal average weighs every axis the same. A fixed weighting might put half the total on one axis alone. A non-linear rule can let one low axis drag the total down further than an average would, or let one high axis lift it more than its share. All three are "composite scores." All three describe a different submission as best. The effect is large where it has been measured: building a composite index for 117 English acute hospitals, Jacobs, Goddard and Smith (2005, Medical Care) found composites "sensitive to changes made to the weighting system and to the aggregation rules", with some hospitals jumping "almost half of the league table" after subtle changes.
That is the practical consequence worth taking from this: a composite is not neutral just because it is built from real components. The neutrality, if any, lives in the rule, and the rule is exactly the thing that stays behind the curtain. How that weighting actually gets chosen is worth reading once you understand that the choice exists at all - most readers never get past assuming a total is a plain average.
A quick check
A composite that is a plain average is verifiable from the outside: add the shown axes, divide, compare to the total. When the arithmetic does not match, the gap tells you the total is carrying hidden weights you were never told about. It takes thirty seconds on results you already have, which makes it one of the few things about a rating tool a reader can check without any special access.
A composite score is not the same idea as a measured total in centimetres - that is a sum of a different kind, owned by measurement rather than rating - and it is not what a human reviewer produces either, since a person's overall impression isn't assembled from stated weights the way a rubric is. Composite is strictly a rating-tool term for how several judged parts became one number, whatever engine is doing the judging on the model layer beneath it. Rate Cock states its combination rule as a documented average rather than leaving readers to reverse-engineer it, which is the exception this whole piece is describing the absence of.