That makes sense on a surface level, but wouldn't a better proxy for CoL be zip code + gross living area? GLA essentially encompasses your "more stuff -> more expensive" model, and zip code should provide an approximation of what the baseline cost of housing is in the area.

(Goodcover dev here)

It could be, along with say family size. But the UX cost of asking such questions is real, especially on an insurance application form. People get worried why we want to know such things so early on.

Today we're able to provide a pretty good estimate with no personal details, nothing but a zip code in fact. We might work the zip code into temp housing default at some point, but it's not without issues. Changes to the property <> temp housing link built into our rating require solid data and regulatory approvals.

Sounds to me like an A/B test is warranted.

We do plan to do some tests, to the extent that we're able. We can A/B test much of the UI but we can't A/B test the insurance terms that we give people because that would require different pricing and thus different rating, which needs to be the same for everyone legally (also any pricing changes require approvals regardless of A/B testing or not).

Don't take this too literally as I'm just a developer, not a licensed insurance agent. But this kind of thing is harder in a heavily regulated business such as ours.