On 9 invoiced gutter jobs, Measured Bid's quoted gutter footage had a median absolute error of 14.2% and a mean absolute error of 16.2%. We also used these jobs to tune the engine, so new houses may miss by more. Homeowners see a range that widens when the data is weaker, and every job should be confirmed on site.
By the Measured Bid teamUpdated September 27, 2026
Quoted footage against invoiced footage on each of the 9 jobs. 6 came in short and 3 came in long.
| Job | Stories | Invoiced | Quoted | Error | Confidence | Invoice inside the quoted band | Imagery |
|---|---|---|---|---|---|---|---|
| A | 1 | 130 ft | 161 ft | +23.8% | Medium | No | 2024 |
| B | 2 | 395 ft | 243 ft | -38.5% | LowImagery from 2015 | No | 2015 |
| C | 1 | 160 ft | 167 ft | +4.4% | Medium | Yes | 2024 |
| D | 1 | 204 ft | 146 ft | -28.4% | Medium | No | 2024 |
| E | 1 | 217 ft | 222 ft | +2.3% | Medium | Yes | 2024 |
| F | 1 | 233 ft | 200 ft | -14.2% | Medium | No | 2024 |
| G | 2 | 184 ft | 135 ft | -26.6% | LowOnly about 90% of the roof was modeled | No | 2024 |
| H | 1 | 130 ft | 125 ft | -3.8% | LowOnly about 88% of the roof was modeled | Yes | 2024 |
| I | 2 | 210 ft | 202 ft | -3.8% | LowImagery from 2018; about 89% of the roof modeled | Yes | 2018 |
Download the table as CSVFree to reuse with credit (CC BY 4.0).
The misses have causes you can check on the house before you quote.
Aerial data can't see roof under heavy trees, and the roof model sometimes leaves a section out. Missing roof shows up as short footage. When less than 90% of the roof is modeled, confidence drops to low and the range widens.
In this test Jobs G, H and I had about 88 to 90% of the roof modeled. Job B's gap is mostly roof under trees.
When the imagery is five or more years old, the estimate shows its date and confidence drops to low. A roof or addition built since then isn't in it.
In this test Job B's imagery is from 2015 and job I's imagery is from 2018.
A two-story house often has porch or first-floor roofs that need their own gutters. When the homeowner says the house has two stories, lower roofs are found by their eave height and priced at the lower story rate.
In this test The 3 two-story jobs came in 38.5%, 26.6% and 3.8% short.
Where an upper gable's eave ends above a lower roof, a short gutter can sit at the peak. The model doesn't find these yet. Your team or the homeowner can draw them in the run panel.
Status Not modeled yet.
The model builds each roof plane as a rectangle, so triangles, valleys and mixed roof directions are approximate.
In this test All 9 test houses were complex roofs.
Nine jobs is a small sample from one region. One unusual house moves the mean by more than a point.
Next 30 or more jobs before any broader claim.
Every estimate gets a confidence level. The homeowner's range is the estimate plus or minus a spread for that level. These are the defaults, and you can change them. The nine-job test checked gutter footage only, not roof area, wall area or prices.
High
±10% price range
Sharp imagery on a simple gable or hip roof that the footprint explains.
No test jobs at this level.
Medium
±15% price range
Complex roofs with many planes, which covers most houses we have tested, or an irregular footprint.
In this test: 5 jobs, invoice inside the band on 2.
Low
±25% price range
Less than 90% of the roof modeled, imagery five or more years old, or a roof the model couldn't resolve.
In this test: 4 jobs, invoice inside the band on 2.
The band in this test is quoted footage plus or minus the spread. A homeowner's price range also carries your minimums and add-ons, so price coverage is tracked separately, from won jobs. On 4 of 9 jobs the invoiced footage fell inside the band.
Each won job shows where the sold price landed against the range the homeowner saw.
From the Overview page in the dashboard, shown on the sample house with sample rates.
Engine changes are tested on the same nine jobs. Two changes that made the result worse, a hybrid outline rule and a fitted correction factor, were not shipped.
| Change | Median | Mean |
|---|---|---|
| First back-test on the nine jobsSeptember 23, 2026 | 15.0% | 17.0% |
| Geometry fixes: exact clipping, separate roof levels, 45° roofsSeptember 23, 2026 | 14.2% | 17.0% |
| Lower roofs counted on two-story homesSeptember 23, 2026 | 14.2% | 16.2% |
| Re-run after partial-roof, 45° and flat-roof checks from a 120-home test; footage unchanged on every jobSeptember 27, 2026 | 14.2% | 16.2% |
In our test on 9 invoiced jobs, quoted footage had a median absolute error of 14.2%: the middle job missed by that much. 4 of 9 were within 10%. The biggest miss was 38.5% short, on a two-story house with 2015 imagery and roof under trees. Results depend on the roof, the trees and the imagery date.
We compare the footage we quote with the footage on the invoice for the whole job, and we show every miss. Figures measured another way aren't directly comparable. Ask how any number was tested.
Not directly. This test covers gutter footage only. Permanent lighting uses the same roofline runs. Roof area and wall area haven't been tested against independent measurements yet.
A gutter company in the Sacramento region of California that uses Measured Bid. Addresses and customer details stay private. The per-job numbers are in the table and the CSV download.
Short more often. 6 of 9 jobs came in short and 3 came in long. The median job was 3.8% short. Missing roof shows up as missing footage.
Confirm the measurements on site first. The estimate is a starting price, and the range is there because aerial data can miss.
Run the demo on a past job and compare the footage with your invoice. No account needed.