±0.8% on commercial roof geometry.
This is the work behind the number.
Forge publishes a single accuracy figure for Hyperion: ±0.8% on commercial flat and low-slope roof geometry, ±2% on small-residential structures. This page documents the comparison baselines, the sample plan, and the limitations stated up front. Independent third-party verification is not yet complete; that work is the next milestone for this page, and the published figure stays hedged until it lands.
Measurement pending.
What the number means.
Gross roof area on commercial single-ply, modified-bitumen, EPDM, TPO, and ballast roofs. The category Hyperion was built for first.
Pitched roofs on residential structures. The depth-resolution gap on these is wider; the published number stays honest about that.
The unit is gross roof area. Per-facet accuracy on complex geometry (dormers, valleys, parapet returns) is tracked separately and will publish its own line when the sample is large enough to be honest about it.
How the number is measured.
- COMPARISON BASELINE A
- EagleView Premium reports for the same structure on the same date, where available.
- COMPARISON BASELINE B
- Certified manual takeoffs by a licensed roofing measurement firm — total station survey for ground truth on a subset of sites.
- CAPTURE DEVICE
- iPad Pro M-series, LiDAR sensor, ARKit fused depth. Recent iPhone Pro models with LiDAR are tracked separately to avoid mixing depth-resolution categories.
- CAPTURE PROTOCOL
- Walk pattern documented per site. Each scan logged with operator, weather, time-of-day, and elapsed scan duration.
- COMPARISON METRIC
- Gross roof area, computed by Hyperion vs. the baseline. Reported as percent delta from baseline.
- OUTLIER HANDLING
- No trimming. All scans count toward the published figure. Outlier sites are listed separately so the reader can see what the published number includes.
- REPORTING CADENCE
- This page updates when the sample crosses a threshold or the figure changes by more than 0.2%. Last-reviewed date is stamped at the top.
What it takes for the number to graduate.
- TARGET SAMPLE SIZE
- ≥30 sites for the published figure to graduate from methodology to measurement.
- VERTICAL MIX
- Primary sample is commercial roofing. Solar PV scans contribute to a separate page; security/AV/HVAC scans run on the same instrument but report on different geometry classes.
- GEOGRAPHIC DISTRIBUTION
- No single CBSA exceeds 40% of the sample. Climate variation matters for membrane behavior; the sample plan reflects that.
- OPERATOR COVERAGE
- ≥3 operators. The published figure has to hold across operators, not lean on one steady hand.
- INDEPENDENT VERIFICATION
- Third-party review by a licensed measurement firm before the figure graduates to a measured benchmark.
The honest line.
Hyperion is engineered to deliver these figures, and the methodology above is what proves it. The proof is not yet published as a measured benchmark. Until the first independent third-party measurement set completes, the published figure stays hedged.
- —The number on the Hyperion page and in vertical pages cites this page for methodology. It does not claim the number is independently verified yet.
- —The first measurement set lands here when complete. The figure may move; if it does, both this page and every place that cites it move with it.
- —Charter Members are the first cohort whose scans contribute to the sample. The covenant runs both ways: the rate is locked, the proof comes from the work.
The page's history.
- APRIL 27, 2026
- Methodology and sample plan published. Figure remains hedged. Independent third-party verification not yet scheduled.
A claim with no methodology page behind it doesn't belong on the site. The proof page is where the work shows.