As-built verification checks that the model represents what was actually built, within an agreed tolerance, before it is handed over as the record model. Buildings never go up exactly as designed: services move around obstructions, products are substituted, site instructions change details. A record model that silently shows the design instead of the building is worse than no model, because people trust it.
At a glance
| Who does it | Contractors and trade contractors (update the model), surveyors (measure), the information manager or client's team (check) |
| Against | Laser scans, site measurements, commissioning records, red-lined drawings, product submittals |
| Tolerance | Set per element type in the requirements, often with level of accuracy (LOA) classes |
| Also verified | Data, not just geometry: installed product, serial number, commissioning date, warranty |
| Output | A record model and data the client accepts, with a statement of how it was verified |
What changes between design and as-built
| Change | Example | What to update |
|---|---|---|
| Rerouted services | A drainage pipe drops under a beam instead of going through it | Geometry, plus access points if they moved |
| Product substitution | Fan coil unit from a different manufacturer | Type, properties, maintenance data, weight |
| Site instructions | A door moved 400 mm for a column | Geometry and room boundaries |
| Omitted or added items | An extra isolation valve fitted for commissioning | New element with its data |
| Hidden works | Fire-stopping, cast-in items | Records and photos linked to elements |
How to verify
- Agree the tolerance and the method in the requirements: what is scanned, what is measured, what is accepted from installers' records.
- Capture: scans or measurements before ceilings and walls close, by zone.
- Compare: model against scan (deviation analysis), or model against installed records.
- Update: the trade that installed the work updates its model; the coordinator federates.
- Check the data: installed products, tags and properties against asset requirements.
- Record the verification: what was checked, how and by whom, so operations know how far to trust it.
Worked example
Scans of a plant room are compared with the MEP model. 92% of ductwork is within 50 mm of its modelled position; three runs deviate by up to 400 mm where they were rerouted around a beam on site. Those runs are remodelled. The comparison also shows a pump that is in the model but not on site: it had been omitted by a design change that never reached the model. Without the scan, the FM team would have looked for a pump that doesn't exist.
Common mistakes
- Renaming the final design model "as-built" without checking anything.
- Verifying geometry but not the installed product data.
- Scanning after ceilings close, when the services are hidden.
- No stated tolerance, so nobody knows what "accurate" means.
Where to read more
ISO 19650-2 covers acceptance of the delivered information model; ISO 19650-3 covers its use in operation. Survey specifications from professional surveying bodies describe scan accuracy and level of accuracy classes.