A match score is a number out of 100, and on its own that is a number you are asked to trust. The whole point of the engine is that it never arrives on its own.
Nine signals go into it, and each is worth a fixed number of points: the same breed is 35, a related breed 15, the same city 25, the same region 12, a lister whose documents an admin has checked 10, an animal inside the breeding window for its breed 10, vaccinated 8, registered 7, and a listing published in the last fortnight 5. A perfect candidate sums to 100.
Every point produces a reason
The rule the engine is built around is that the points and the reasons are the same list. A point cannot land without producing a reason, and a reason cannot appear without a point behind it. So when a candidate scores 72, you can read back which of the nine got it there.
Two pairs are exclusive. The same breed and a related breed never both count: a Labrador matched with a Labrador scores the 35, not the 35 and the 15. The same city and the same region work the same way. Distance is counted once.
What is never in there
Money. There is no signal for a paid account, a featured listing or a subscription tier, and the vocabulary of reasons has no word for one. Featured placement exists in browse, in positions that say they are featured. It has no route into a compatibility score, because a score that can be bought is not a compatibility score.
What happens when nothing matches
Rather than return an empty page, the engine widens exactly one constraint and tells you which. That is called loosening, and it is reported as breed, location or similar - never silently.
None of this makes the engine right. It makes it checkable, which is the part you can act on: if a reason looks wrong for your animal, you know to disregard the score rather than wonder about it.
