Write two separate acceptance questions
First ask whether the drawing, model, or coordinate schedule agrees with the physical features it claims to represent. Then, if the commission includes construction assessment, ask how those features compare with the nominated design. Identify the responsible reviewer for each question. The person preparing the record should not silently decide which construction differences are acceptable.
Define the features and surfaces being compared. A modelled column centreline, the visible concrete face, and a fitted axis are not interchangeable. Confirm the survey date, construction state, and design revision. Otherwise, a reported discrepancy may be caused by comparing different surfaces or different stages rather than by poor measurement.
Establish a dependable comparison reference
Review the coordinate system, units, height reference, and transformations before comparing geometry. Test known locations spread through the relevant area. An overall translation or rotation can produce an alarming deviation map even when the local shapes agree. Conversely, fitting two datasets freely can conceal a real positioning error that matters to the project.
Document whether the comparison uses the accepted project reference or a local alignment for a stated purpose. Both may answer useful questions, but they answer different ones. Preserve the original coordinates and record any transformation so another reviewer can reproduce the analysis and understand what has been removed by alignment.
Select checks that test the actual risk
Choose check locations across levels, work areas, and feature types. Include critical interfaces and areas where visibility or modelling was difficult. If only a sample is checked, describe the sampling rationale and its limits. A small number of easy measurements near one entrance cannot establish the quality of an entire complex building.
Where practical, use observations that were not used to create or adjust the record being tested. Repeatable comparisons against the same source can help find drafting mistakes, but they may share the source's underlying error. Plan the check method and its expected uncertainty so that it is capable of resolving the question being asked.
Inspect differences rather than only averages
Report signed differences where direction matters and use clear feature identifiers. An average near zero can hide positive and negative differences that cancel. Group results by useful areas or systems and investigate patterns. A whole level shifted in one direction suggests a different issue from isolated discrepancies on irregular surfaces.
For cloud or model comparisons, explain the comparison settings and exclusions. A colour map can make a report readable, but it should not be the only evidence. Moving objects, incomplete coverage, and ambiguous surfaces need considered treatment. Keep a traceable link between significant findings, their location, and the observations that support them.
Close findings with a clear outcome
Classify findings according to the agreed review process: confirmed record error, construction difference for engineering review, insufficient evidence, or no action required within the defined criteria. Update all affected deliverables when corrections are accepted. Record what changed and whether the correction requires another check.
The verification report should state its extent and avoid claiming universal accuracy from limited checks. It should explain unresolved areas and the uses for which the record was assessed. This gives future users a practical basis for deciding whether they can rely on the information or need targeted additional measurement.
Common questions
Does comparing a model with its source point cloud verify it?
It can check modelling consistency and reveal omissions. It does not independently prove the point cloud's position or measurement quality, so additional checks may be needed for the intended use.
What should happen to a borderline result?
Review the measurement uncertainty, comparison definition, and project criteria. Record the ambiguity and seek the responsible reviewer's decision or additional evidence rather than forcing the result into a simple pass or fail.
Further reading
This article was rewritten in October 2026 around the subject of the original Geobiro post. The original publication date is retained; the update date identifies the current text.