Vendors rarely publish error margins. The one independent test I found, an older drone study, put area within 3% on all 13 roofs, worst where trees hid an edge; stale imagery or a misread pitch can do far worse. So order AI reports for five roofs you have measured yourself, compare area, pitch and lengths, and price each miss.
The biggest single trap is pitch. One step of pitch, 7/12 read as 6/12 for example, moves roof surface area by about 3.5%, so a report with a perfect footprint can still be several per cent out. Imagery that pre-dates an extension or a re-roof is the second trap, and it can cost far more.
What vendors publish about accuracy, and what independent tests exist
The vendor pages I read in November 2026 describe accuracy in words and give no error margin. Roofr's measurements page says each report goes through a multi-check system with up-to-date imagery and a calculation formula. EagleView's residential reports page calls its measurements highly accurate for material ordering. Hover's pricing page speaks of accurate measurements. None prints a percentage or a test method, and neither Roofr's nor EagleView's page says when its imagery was captured. That is not an accusation; it is the reason a test on your own roofs carries weight.
Independent evidence is thin. The only test I could find with per-roof numbers is old and small. Haag Engineering, an engineering consultancy, compared a drone-based measurement workflow with manual measurements by field surveyors on 13 roofs. A January 2018 account, written by the chief executive of the company whose software processed the images, reports roof area differences from +1.2% to -2.7% per roof, an average difference of 0.6% and a mean absolute variation of 1.1%. The largest miss was a roof whose front edge was covered by overhanging branches. It tested a drone workflow, not today's aerial AI reports, so treat it as a picture of how error behaves rather than a benchmark: small on average, worst where something hides the roof edge.
That leaves your own roofs as the only evidence that fits your vendors and roof types. For how these measurements are produced and what they cost per job, quoting from satellite images and the Roofr, EagleView and Hover comparison cover the background.
The Complete Excel Course with AI Integration
From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.
By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner
Buy nowPick five roofs that can embarrass the tool
Choose roofs where you already hold a measurement you trust: tape or drone figures from a job you won, or as-built quantities you reconciled against deliveries. The past-job rule keeps ladders out of the test. If you do go up for any reason, follow your normal fall-protection plan.
| Roof | Why it is in the test | What it exposes |
|---|---|---|
| 1. Simple gable, one pitch | The baseline | Whether the tool is close when nothing is difficult |
| 2. Hip roof, one pitch | Hips add length and facets | Hip length and area across four planes |
| 3. Cut-up roof with dormers and valleys | Many facets and junctions | Valley and ridge lengths, merged facets, pitch on small planes |
| 4. Steep roof under trees | Obstructed eaves | Pitch and eave length when the view is blocked |
| 5. A roof that changed after the likely imagery date | Stale imagery: an extension, re-roof or new dormer | Whether the report shows its imagery date, and how you would notice a gap |
Your own tape figures have error too. If they could be out by 2%, no tool can be judged more finely than that. Write a confidence note beside each ground-truth figure (high, medium or low) so a mismatch on a low-confidence roof prompts a re-measure instead of a verdict on the tool.
What to record, and the spreadsheet that does the sums
One spreadsheet, one row per roof. Record the area in squares (a square is 100 sq ft, about 9.3 square metres), the main pitch, the ridge, hip, valley, eave and rake lengths, the facet count, the report's suggested waste factor, the imagery date if the report shows one, the order date, and which vendor and report type you used. Work out error as the report figure minus the tape figure, divided by the tape figure, and keep the sign. A tool that runs short is a different problem from one that runs long.
Columns: A roof | B tape squares | C report squares | D true pitch (rise per 12) | E footprint sq ft | F installed price per square
Error % =(C2-B2)/B2
Slope factor =SQRT(1+(D2/12)^2)
Area from footprint (squares) =E2*SQRT(1+(D2/12)^2)/100
Report vs footprint check =(C2-E2*SQRT(1+(D2/12)^2)/100)/(E2*SQRT(1+(D2/12)^2)/100)
Cost of the miss =(C2-B2)*F2The footprint check is a second opinion that needs no ladder. If a report's area differs from footprint times slope factor by more than about 2%, either the pitch or the footprint is wrong, and you know which to look at. It will not catch a merged facet, because merging facets barely changes area, but it catches a misread pitch and a missing extension.
Five reports at list prices cost little. Roofr's pages show $19 a report on its free Starter plan and $13 on Essentials and Measure+, so five cost $95 or $65. Hover's pricing page shows $29 for a simple roof on its pay-per-project plan, so five simple roofs cost $145, with complex roofs priced higher. EagleView's residential reports page lists its preliminary Bid Perfect report from $18 and a discounted Premium Roof Report upgrade at $24.25, and its separate pricing grid ran from $24.25 to $87 a report by roof size as of October 2026. Delivery differs as well: Roofr says Starter reports arrive within 24 hours and Essentials reports within 2, while Hover lists a typical 12 hours on Starter. Test the kind of report you would really bid from, because some tools return a measure in seconds from an address and others deliver a checked report hours later.
The test on five illustrative roofs
A six-person roofing firm orders reports for five roofs it re-roofed in the last two years and holds tape figures for. It prices work at an illustrative $520 a square installed. Every figure below is invented to show the method, not a result from any vendor.
| Roof | Tape (squares) | Report (squares) | Error | Effect on a bid at $520 a square | What the firm found |
|---|---|---|---|---|---|
| 1. Simple gable, 6/12 | 24.6 | 24.8 | +0.8% | $104 high | Nothing to explain |
| 2. Hip, 7/12 | 31.2 | 31.9 | +2.2% | $364 high | Report included a small overhang the tape run skipped |
| 3. Cut-up with dormers, 8/12 | 38.5 | 36.4 | -5.5% | $1,092 short | Pitch read as 7/12, and one dormer facet merged into its neighbour |
| 4. Steep, 10/12, under trees | 27.0 | 25.9 | -4.1% | $572 short | Branches hid part of the front eave line |
| 5. Extended after the imagery, 6/12 | 33.4 | 28.9 | -13.5% | $2,340 short | Rear extension built after the imagery; the report measured the old footprint |
The mean absolute error is 5.2% across all five and 3.1% without the extended roof, but an average hides the finding. The errors that cost money all ran short. Over-measuring by 0.8% or 2.2% costs a little competitiveness; under-measuring by 5.5% on roof 3 means a bid $1,092 too low on a job worth about $20,000, which is more than a 3% contingency (about $600 on that job) would cover. Added together, the five reports left the firm 6.8 squares short, about $3,540 at its rate, and 4.5 of those squares came from one roof the imagery had never seen.
Now the firm can write rules instead of opinions. Reports are good enough for budget pricing on simple and hip roofs, where they landed within 3%. They are not good enough on cut-up roofs, steep roofs under trees, or any roof that may have been extended, and each of those gets a re-measure or a pitch check before the number goes in a quote.
Check the lengths: the valley figure usually gives the game away
Area is only part of a roofing quote. Ridge, hip, valley, eave and rake lengths drive the cap, flashing, underlayment and edge metal. On roof 3, the firm compared lengths too:
| Length (ft) | Tape | Report | Error |
|---|---|---|---|
| Ridge | 96 | 94 | -2.1% |
| Valley | 62 | 49 | -21.0% |
| Eave | 118 | 117 | -0.8% |
| Rake | 84 | 80 | -4.8% |
Valley length wanders most, because every dormer adds a junction where two planes meet at an angle, and a tool that merges two small facets loses the valley between them. Thirteen feet of valley is small money. Its value is as a clue: a valley total 21% short on a roof with dormers points at a merged facet, and the area figure should be doubted too. The report listed 16 facets and the firm counted 17 planes from a street photo. A facet count that does not match your own count is a free alarm, so check it every time.
Why one step of pitch moves the area by 3% to 4%
A roof's surface area is its footprint multiplied by a slope factor, and the slope factor is the square root of 1 plus (pitch divided by 12) squared. Pitch is quoted as rise per 12 units of run. The steeper the roof, the more each extra step adds:
| Pitch | Slope factor | Area added by this step | Angle |
|---|---|---|---|
| 4/12 | 1.054 | 2.3% | 18.4 degrees |
| 5/12 | 1.083 | 2.8% | 22.6 degrees |
| 6/12 | 1.118 | 3.2% | 26.6 degrees |
| 7/12 | 1.158 | 3.5% | 30.3 degrees |
| 8/12 | 1.202 | 3.8% | 33.7 degrees |
| 9/12 | 1.250 | 4.0% | 36.9 degrees |
| 10/12 | 1.302 | 4.1% | 39.8 degrees |
| 12/12 | 1.414 | 4.2% per step above 10/12 | 45.0 degrees |
Take a house with a 2,150 sq ft footprint. At 6/12 the roof is 2,150 x 1.118 = 2,404 sq ft, or 24.0 squares. At 7/12 it is 2,150 x 1.158 = 2,489 sq ft, or 24.9 squares. The gap is about 0.85 of a square, roughly $440 at the example rate, from a pitch difference nobody would notice standing in the street. This is why pitch deserves its own column in the log and why a report with the right footprint and the wrong pitch is still wrong.
You can get pitch without going on the roof. Three ways:
- Gable end, from the ground: measure the ridge height above the wall plate and half the span. With a 24 ft span and a ridge 7 ft above the plate, pitch is 7 divided by 12, times 12, which is 7/12.
- From the loft: rest a pitch gauge or digital level on a rafter and read the angle, then convert with the table (30.3 degrees is 7/12).
- From a photo: a straight-on photo of the gable end with a known width in frame (a door, say) lets you scale the rise and run. It is rougher, so use it to confirm, not to settle a close call.
Also check whether a report gives one pitch for the whole roof or one per facet. A cut-up roof often mixes pitches, and a single predominant pitch applied to every plane is a quiet source of error. A report that lists each facet's pitch lets you spot the odd one out.
Waste is a separate number, and it is yours to set
Measured squares are not squares to order. Waste covers cuts at hips, valleys and edges, plus offcuts. EagleView's residential reports page lists a waste factor on its Bid Perfect report for steep-slope roofs only, and Roofr's measurements page says its reports include a waste factor and work out materials from your preferred one. Ask which method a report applies and whether you can override it. Keep two columns in the log: report squares and squares to order.
Build your own waste table from jobs you have finished. Waste is squares ordered divided by squares measured, minus one. Your history might read like this: the last eight simple gables ran at 9% to 11%, the last six hips at 12% to 14% and the last five cut-up roofs at 16% to 19%. These are made-up figures; yours will differ. Then the order for roof 2 is 31.9 squares x 1.13, about 36.0 squares, instead of whatever default the report suggests. Roof 3 at 16% to 19% needs a wider margin than roof 1 because it has more cuts, however good the measurement.
The temptation after a bad report is to add a blanket 5% to everything. On a 30-square roof that adds 1.5 squares, about $780 at the example rate, to every bid, including the simple gables the report measured within 1%. In the test above, a 5% pad would have nearly covered roofs 3 and 4 and still left roof 5 about 9% short. A targeted rule costs less and fixes more: measure the roofs that trigger a check, and leave the rest alone.
Imagery date: the roof that changed after the photo was taken
Aerial and satellite images are photographs taken on a particular day. Roofs change afterwards: an extension, a new dormer, a re-roof, solar panels, a tree felled or grown. A report built on older imagery measures the roof that used to exist, and it will look just as polished as a correct one. This is the cause of the biggest miss in the test, and no amount of pitch checking would have caught it.
Four ways to catch a stale roof:
- Look for an imagery date in the report. If it shows one, record it in the log. If it does not, ask the vendor in writing and keep the reply.
- Put a question in your intake form: "Has any part of the house or roof been extended, re-roofed or fitted with solar panels in the last ten years? If yes, roughly when?"
- Ask the customer for a phone photo of each elevation, or take a drive-by photo, and compare the outline with the report diagram.
- Compare the report footprint with any plan the customer holds, such as an older survey or extension drawings.
On roof 5 in the example, the clue came during the booking call. The customer mentioned that the kitchen at the back had been added "a couple of years ago". The report diagram showed a plain rectangle with no rear projection. That one sentence, taken during the call, was worth $2,340. Without it the quote would have gone out on 28.9 squares and the firm would have found the other 4.5 when the scaffold went up.
When to re-measure: a rule you can apply on the phone
Re-measure by tape, drone or a second report if any one of these is true. They come straight from the five-roof test, so swap in the triggers your own results suggest:
- The report shows no imagery date, or the customer mentions work on the house since a plausible date.
- The facet count differs from your own count of planes on a photo.
- Pitch is 9/12 or steeper, or the report gives one pitch for a roof with obviously different planes.
- The footprint check differs from the report area by more than about 2%.
- Trees overhang any eave in the photos.
- The valley total looks short for the number of dormers.
- A 5% miss would be bigger than the contingency you carry on that job.
A re-measure costs money, so the rule needs a sum behind it. Price a visit at 45 minutes of staff time at $60 an hour plus $15 of fuel, $60 in all. A 5% miss on a $20,000 roof is $1,000. If a roof with one trigger has even a 10% chance of a miss that size, the expected loss is $100, more than the visit costs. A roof with no triggers has a far smaller chance, so the visit is not worth it. That is the whole logic: visit where the odds justify it.
A second report is not always a second opinion. Two vendors may draw on the same imagery, so before you treat a $13 or $29 re-order as a check, find out what imagery each uses and from when. A tape, a drone flight or a pitch gauge in the loft is independent; another report may only repeat the first one's mistake.
Reading your own results
When the five rows are in, look for patterns before looking at averages:
- Errors under 3% on roofs 1 and 2, no consistent sign: use reports for budget pricing on simple roofs and keep the triggers for the rest.
- Pitch wrong on more than one roof: measure pitch yourself on every job; it is the cheapest check there is.
- Every report short: find the cause (usually pitch, merged facets or stale imagery) instead of padding; a blanket pad hides the cause and costs you bids.
- Stale imagery on roof 5: add the intake question and the elevation photos to your standard process.
- A vendor that cannot tell you its imagery date: weigh that when choosing between tools, because you cannot judge what you cannot date.
Repeat the test after a vendor changes its product, and again once a year with five new roofs. The quote that goes out carries your name, so a report is an input and never the final word. Detecting damage on a roof is a different job from measuring one, and what drone and photo inspection tools actually detect needs its own test. If a report feeds an order, turning squares into waste and bundles is the next step, and an instant estimate on your website depends on the same measurement being right before a customer sees a price.
Roof measurement accuracy: questions from the test
How close should an AI roof measurement be to a tape measure?
There is no universal figure, so decide by money. Convert the error into squares and dollars at your installed rate and compare it with the contingency you carry. As a starting rule, accept area within 3% on simple roofs and treat any pitch difference as a reason to re-measure, then tighten it once your own results are in.
Does a bigger waste factor fix a measurement error?
No. Waste covers cuts and offcuts during installation. Hiding measurement doubt inside it makes every bid dearer and leaves surplus material on the drive. Fix the measurement first, then set waste from your own history of squares ordered against squares measured, by roof type.
How do I find out which imagery date a report used?
Look in the report header and notes. If no date appears, ask the vendor in writing and treat the report as unverified on any roof that may have changed since, such as an extension, a re-roof or new solar panels. A current photo of the front and gable ends settles most doubts.
Can I test AI roof measurements without climbing?
Yes. Use past jobs where you hold tape or drone figures, or measure the footprint and eaves from the ground and apply the slope factor. Pitch can come from a gable-end measurement: rise divided by half-span, times 12. Keep ladders for roofs you were going to visit anyway.
Further reads
- Which AI Tools Do Roofers Need? Measurements, Calls, Estimates — Which AI tools a roofer needs, from measurements to calls.
- Roofing Proposal Software: Adding AI-Written Scopes and Warranty Wording — Turning a checked measurement into a proposal with scope and warranty wording.
- Roof Inspection Reports: Photos to a Customer-Ready Report With AI — From roof photos to a customer-ready inspection report.
- How to Use AI for a Roofing Business: Where to Start, and What Waits — Where a roofing firm should start with AI, and what can wait.
- ChatGPT for Roofing: 10 Prompts for Estimates, Inspections and Warranty — Ten prompts for estimates, inspections and warranty follow-up.
- CompanyCam Alternatives for Roofers: AI Photo Reports and Checklists — Photo documentation tools for roofers, with AI reports compared.
- AI Automation for Roofers: Six Workflows Worth Setting Up First — Call answering, estimate follow-up, measurements, instant estimates, inspection reports and job close: what each costs, how to start and what to watch.
- Siding Contractor Software: What to Pay For and Where AI Helps — CRM, estimating, measurement and photo software for a small siding firm compared on price, with a worked five-person choice and the AI that earns its place.
- Roof Damage Reports for Insurance Claims: Let AI Document, Don't Adjust — Photo order, caption wording and a measurements log that let AI organise roof damage evidence without drifting into coverage opinions.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Roofr measurements and pricing pages, EagleView residential reports page, Hover pricing page, a January 2018 Geo Week News account of Haag Engineering's 13-roof measurement test, all read in November 2026.