Accuracy & Limitations

Review the quantities alongside their source references, measurement units and calculation basis. Check that repeated elements belong to the right building, level and location. A usable draft may still contain unresolved counts, dimensions or missing scope. Those items must remain visible for review rather than being treated as confirmed quantities. Deterministic CPM scheduling with P6-compatible XER export and IPC payment arithmetic in Contract Administration are separate services.

The table above lists current upload capability. Experimental means a processing route is available but quantity accuracy has not completed independent revalidation. Unsupported formats need an export to an accepted format. Structured model data can help identify elements and units, but it can also contain incomplete or conflicting quantities. Neither the file extension nor successful conversion establishes accuracy.

Every current QTO report requires review by a qualified engineer before contract or construction use. Check: - Source pages, models and any omitted scope - Counts, dimensions, units and measurement rules - Repeated elements across buildings, floors and drawings - Openings, voids, net/gross basis and deductions - Missing information, assumptions and unresolved disagreements - Specification-driven work and items absent from the supplied source Providing project measurement rules helps this review; it does not approve the output.

An AI confidence score is a review aid, not a measured probability that a quantity is correct. Agreement between extraction attempts is also not an independent accuracy test. Distinguish measured values, calculated values, assumptions and missing information. Check the source evidence and applicable measurement rules before accepting a quantity. An incomplete or uncertain result requires review even when its confidence score is high.

QTO revalidation is pending. Before publishing new accuracy claims, validation must compare output quantities against independently reviewed reference quantities with a declared scope, units, measurement rules and tolerances. It must assess missing and extra items, page and element coverage, and repeat runs of the same source. Software tests and report completeness checks help find regressions. They do not by themselves establish engineering quantity accuracy. Earlier QTO percentages and benchmark claims are withdrawn pending this work. Separate pricing benchmark (unchanged): - Pricing (case R18): the same villa's unpriced BOQ priced blind, then compared to the contractor's 2021 rates. All 12 items priced; 3 of 12 landed inside the contractor's rate bands, and composite structural rates currently over-price (total ~2.4x the 2021 contract). Calibration work continues with each step re-measured.

Current QTO limitations include: - Scanned, handwritten or unclear dimensions and symbols - Missing or contradictory scales and units - Pages, tiles or model elements that were not analyzed - Ambiguous schedules, repeated instances and complex MEP routes - Missing properties, unsupported geometry or conversion losses - Unspecified measurement rules and specification-driven quantities - Variation when the same source is measured again Unresolved quantities and incomplete coverage require review. A fresh measurement revision preserves the earlier result for comparison. - Composite "all-inclusive" structural pricing calibration (currently over-prices)

- Accuracy questions or disputes: [email protected] - Benchmark / dataset contributions: [email protected] - Vendor / enterprise procurement evaluations: contact us via /contact?topic=enterprise We're transparent: if an output doesn't match your expectations, send us the file and we'll either explain the discrepancy or fix the pipeline.