A single number is easy to read and easy to misread. When people see a score on a listing, they want to know one thing quickly: can I rely on this? But the moment the number surprises them, they want to know why — and a bare number has nothing to say.
This note sets out the principle we are designing TrustScore around, and the trade-offs that come with it.
Reasons travel with the number
Every score we show is meant to carry its reasons with it. Not a full audit trail, but the few things that moved it most, written in plain language. A score that is lower because the location claims disagree should say so. A score that is higher because details have been confirmed should say that too.
This changes what a score is for. It stops being a verdict and becomes the start of a decision the person still makes for themselves.
What we have given up
Showing reasons constrains which signals we can use. A signal that cannot be explained in a sentence is hard to justify, even if it would make the number slightly more accurate. We have chosen legibility over that last bit of precision, deliberately.
It also means the score has to be stable enough to explain. A number that jumps around for reasons nobody can name erodes exactly the trust it is supposed to describe.
Scores should move when the world does
The opposite failure is a score that never changes. A home that was accurately described once may not be accurately described now. We are working out how and when scores should be revisited, and how to show that a reason is recent rather than inherited from an old listing.
The specific signals and update rules are still being settled, and we will describe them here when they are.
