How to count building stories from satellite imagery

A story count discrepancy is one of the most common reasons a risk gets re-rated after binding. The application says two storeys. The roof photo from the last renewal shows something that looks like three. Nobody caught it because nobody looked closely enough, and by the time a claim forces the question, the premium's already locked in at the wrong number.

Counting storeys from overhead imagery isn't the same exercise as reading a floor plan. You read a roof from above, maybe a facade at an angle, and work backward to a number that has to hold up in an underwriting file.

What you're actually measuring

A straight-down (nadir) image gives you the building footprint, the roofline, and whatever shadow the structure casts, the raw material for an indirect storey estimate. Shadow length, combined with the sun angle at capture time and the image's ground sample distance, gives you an estimate of building height. Divide that height by a typical floor-to-floor measurement for the construction type (residential wood frame runs differently than masonry commercial), and the result is a storey estimate, still short of a verified count.

Oblique imagery, captured at an angle rather than straight down, is more useful here because it shows you the facade directly. Window rows, cornice lines, and parapet breaks on an oblique shot let you count visible floors the way you'd count them walking past the building, provided the angle and resolution are good enough to separate one row of windows from the next.

Roof geometry matters too, and it's easy to misread. A single-storey structure with a steep gable or a mansard roof can throw off a height estimate badly if you're using height alone to infer storeys. A flat-roofed two-storey commercial building and a pitched-roof single-storey house with a tall attic can cast shadows of similar length. That's the gap between "building height" and "storey count," and it's where a lot of manual reviews go wrong when someone's working from a shadow calculation alone instead of corroborating it against roof shape and material.

Where this breaks down without a verification step

Application forms are filled out by applicants, agents, or whoever happened to know the property best at binding time, and "storeys" gets interpreted loosely. A walkout basement with a kitchen gets called a storey by one applicant and a basement by another. A loft addition over a garage gets counted by some applicants and left off by others, depending on who filled out the form. None of that is fraud. It's just inconsistent terminology colliding with a number that drives your rating tables.

The imagery doesn't care how the applicant defined the space. It shows what's built. Combined with roof material and footprint data pulled from the same pass, a storey estimate derived from overhead and oblique imagery gives underwriting something to check the form against before the policy is bound, rather than something to argue about after a claim.

This is the gap Property Characteristics is built to close: a structured record per address built from very high resolution imagery, covering roof geometry, material, storey count, and visible site features, captured on an annual cadence so the record doesn't go stale between renewals. It works alongside an inspector's driveway visit, giving underwriting a check between what the form says and what the roof actually looks like, before the binder goes out.

A quick gut-check for reviewers

If you're eyeballing imagery without a structured pipeline behind it, a few questions catch most of the obvious misses: does the shadow length roughly match the declared height for that construction type, does the roofline show a pitch that could be hiding an extra floor, and does the facade (if you have an oblique angle) show window rows that line up with the stated storey count. Any one of those flagged on its own isn't proof of anything. Two or three flagged together on the same address is worth a second look before the policy binds.

Ready to see what a per-address roof and storey record looks like for your book? Property Characteristics builds that record from imagery before you bind, not after a claim forces the question.