Software for Roofing Contractors

Measurement accuracy has a resolution cliff, and big roofs fall off it

Automated roof measurement works beautifully on a 20-square suburban gable and then quietly fails on the jobs worth the most money. The reason is resolution: tracing runs against imagery sampled at a fixed scale, and past a certain roof radius the sampling can no longer resolve the edges. The trace either collapses, bleeds into the neighbour's roof, or returns an area confidently wrong by a quarter.

That is a brutal failure mode, because it fails on the large, complex, high-value roofs — exactly the ones where an estimator most wants help and least wants to double-check a machine.

Fixing it is unglamorous engineering: scale the sampling with the roof, cap the vertex budget so the outline stays a roof and not a noisy blob, clip the bleed into adjacent structures, and heal the gaps only where the geometry genuinely justifies it. Then verify against real addresses rather than a demo, comparing traced area and corner deviation before and after every change. We keep a harness that does exactly that, because a measurement change that looks fine on one house can be a systematic 25% error on another.

A roof is not one number

Most estimating software models a job as a single roof: one type, one square count, one price. Real jobs are not shaped like that. There is a main house in architectural shingle, a patio cover in rolled torch-down, and a detached garage that is somehow both.

When the software only holds one number, the rep works around it — quoting the main roof and writing the rest into the notes, or averaging everything into a blended rate that is wrong for both parts. Either way the margin is a guess.

So we model a job as sections. Each carries its own type, squares, pitch and pricing behaviour, and the total is the sum rather than an approximation. Job-level facts — stories, layers, decking, permits, add-ons — sit above the sections and apply once.

Do not show a choice that is not a choice

Tiered good/better/best pricing works when the homeowner genuinely has options: different shingle systems, different warranties, a real decision to make. It is powerful there, and it raises average job value.

It is actively harmful everywhere else. A flat rolled-torch roof or a repair has one sensible answer, and presenting three padded tiers around it reads as exactly what it is — an upsell ladder with two fake rungs. Homeowners notice, and it costs the trust the rest of the proposal depends on.

So the proposal follows the work: tiered types show the ladder, single-package types show one price card and no ladder at all, and repairs show the price the rep entered for the scope they actually inspected. Whether tiers appear is a property of the roof type, not a sales setting somebody can leave switched on.

How a job moves through the system

1. 1

   Lead lands and follow-up starts

   The moment a lead arrives, automated contact begins — because the contractor who speaks to them first usually wins, and that is a scheduling problem rather than a sales-skill problem.

2. 2

   Measure from above

   The roof is traced from aerial and elevation data into facets with pitch, giving a square count before anyone climbs anything.

3. 3

   Inspect on site

   Photos captured by stage with the conditions that change the price — layers, decking, penetrations. Images are compressed on the device, because raw phone photos are large enough to fail an upload silently.

4. 4

   Build the proposal from what already exists

   Measurements, photos and customer details carry into the builder. The rep confirms scope and prices rather than re-entering a job they just inspected.

5. 5

   Present and sign

   A proposal the homeowner reads on their phone, with the pricing shape that matches the work, and a signature at the end of it.