The moat wasn't the measurement. It was admitting that.
I'm building McQueen Roofing, an address-in, itemized-estimate-out tool for roof replacements, and the biggest lesson so far has nothing to do with roofs. It's about what I almost built instead.
The first version of the plan was the obvious one: pull aerial imagery, measure the roof, ship a report. Area, pitch, facet count. That's what AI roofing usually means, and it's what most of the category has converged on. I spent longer than I want to admit assuming that was the hard part, and that pricing would be a simple markup on top once the measurement was solid.
It wasn't. The measurement turned out to be the easy, mostly commoditized layer. Several companies already do it well. What nobody seemed to be taking seriously was the part after: turning that geometry into a number a homeowner could actually trust, one that itemizes tear-off, underlayment, ice and water shield, drip edge, ridge and starter courses, flashing, ventilation, field shingles, and labor separately, with a waste factor that reflects the real complexity of that specific roof instead of a flat percentage slapped on every job.
That reframe changed what I was actually building. The product isn't a measurement tool with a price attached. It's a pricing engine that happens to consume measurement as an input. Every number in the output is tagged with how it was produced, measured, derived, assumed, user-provided, or allowance, so the estimate shows its work instead of just asserting a total. It also flags when a quote, yours or a contractor's, looks like it's missing a line item, which turned out to be the actual answer to the question I got asked most in early conversations: why do roofing quotes for the same job vary so much? Almost never fraud. Almost always undisclosed differences in scope.
Pricing is currently: free for a rough, footprint-based estimate with no measurement, eighteen dollars one-off for a fully measured, itemized pass pulling current aerial imagery. I'm building a more precise tier, one that measures every ridge, hip, valley, rake, and eave individually instead of carrying them as a single allowance, and I'm being deliberate about not implying it's available before it actually is. It isn't yet. That's a small thing, but it's the same discipline the whole product is built on: don't assert a number you can't show your work for.
If you're building anything in a category where the obvious hard part is actually already solved by three other companies, it's worth asking what the unsolved layer above it looks like. That's usually where the actual product is.
