measurement
What an Aerial Measurement Cannot See
Remote takeoff is accurate about the things it can observe. Knowing which parts of a roof it cannot observe is what makes the number safe to quote from.
A remote measurement is a measurement of an image. Everything it reports well is something visible from above on the day the image was captured, and everything it reports badly falls into one of a small number of categories.
Knowing the categories is the difference between a tool you can quote from and a tool you have to verify every time.
Obstruction
The most common limitation and the easiest to check for.
Tree cover. A mature canopy over an eave hides the roof edge. The measurement either stops at the visible edge, which underestimates, or extrapolates the line, which is an estimate presented as a measurement. Either behaviour is defensible. Only one of them is defensible silently.
Shadow. A north-facing plane photographed in low winter sun can be dark enough that edges are not recoverable. Shadow also distorts automated pitch inference, because many methods use shadow geometry as a cue.
Snow. Snow cover hides detail, changes apparent edges where it has drifted, and can produce a roof that looks simpler than it is. Imagery captured in winter in a northern market should be treated as suspect until the capture date is checked.
Adjacent structures. On a dense terrace or a zero-lot-line property, where one roof ends and the next begins is a judgment, and automated methods make it from colour and texture rather than from ownership.
Vertical detail
Aerial measurement is good at plan geometry and weaker at anything whose extent is vertical.
Overhangs. The eave projects past the wall. From above, the roof edge is visible and the wall line is not, so the overhang is either measured from the shadow, inferred from a default, or ignored. A report that gives a wall area for siding without saying how it handled the overhang has made an assumption you cannot see.
Fascia and soffit lengths depend on the overhang, so they inherit the same uncertainty.
Wall heights need elevation data or an inference from shadow and known object sizes. A single nadir image carries no height information at all.
Material and condition
This is the category where the mistake is most expensive, because the number looks the same.
An aerial image can often identify a material plausibly, and identifying material from texture and colour is pattern matching, not inspection. More importantly, it says nothing about what is underneath: deck condition, existing layers, rot at the valleys, previous repairs.
A measurement is a quantity. It is not a scope, and a quote built from a quantity alone has priced the material and guessed the labour.
The capture date is the most important number in the report
Everything above assumes the image shows the building as it is now. Imagery is not continuous: a property is photographed on specific days, and the gap between captures varies from months in dense markets to years in rural ones.
So the report should state when the image was taken, and the user should read it. Three cases where it decides whether the measurement is usable at all.
After a storm. Imagery captured before the event shows the roof that no longer exists. For claim work this is sometimes exactly what you want, as evidence of prior condition, and it is never what you want as a measurement of the current structure.
After a renovation. An addition built last summer is not in imagery from two summers ago, and the measurement will be confidently wrong by the area of the addition.
New construction. A house completed this year may not be in any public imagery, and the measurement will be of the lot.
Public programmes publish their currency, and the USGS National Map is the reference point for what is available and how recent it is. Commercial providers refresh more often in populated areas and much less often elsewhere, which means the accuracy of a remote measurement varies with the rural-ness of the address in a way nobody mentions in the marketing.
When to fly it yourself
The gap between "imagery is too old or too obstructed" and "send somebody with a tape" is filled by flying the property, and that is a regulated activity.
Commercial drone operation in the United States is governed by 14 CFR part 107, which covers pilot certification, aircraft registration, operating limits and airspace authorisation. The requirements are not onerous and they are not optional, and "we only fly our own customers' properties" is not an exemption. If a business is flying as part of its commercial work, part 107 applies.
The operational rules that most often catch people out are airspace authorisation near controlled airports, the rules on flight over people, and the requirement to keep the aircraft within visual line of sight. All three are checkable before the truck leaves.
What a good report does about all this
It does not eliminate the limitations. It declares them.
Per-facet output rather than a single total, so a questionable plane can be identified rather than being averaged into a number. The capture date, stated, not buried. A flag on facets where the edge was obstructed or the pitch was inferred rather than measured. And a statement of method for pitch.
A report that gives one confident number and no provenance is asking to be trusted. A report that shows its working is one you can quote from, and argue from later.