How to Check CAD Accuracy Before Design Starts

A drawing can look credible and still be unsafe to design from. A clean floor plan may conceal an assumed wall thickness, a shifted grid, an inconsistent level datum or a survey that simply did not capture the area where a new stair is proposed. Knowing how to check CAD accuracy means testing the evidence behind the drawing, not just its appearance.

For existing buildings, this is particularly important before planning alterations, coordinating services or pricing work. Small discrepancies can become expensive once they affect structural openings, façade interfaces, listed fabric or prefabricated components. The right checking process gives the design team a clear understanding of what is measured, what is interpreted and what needs confirming on site.

Start with the drawing’s stated purpose and tolerance

CAD accuracy is not one universal number. A landlord’s marketing plan, a feasibility drawing and a detailed measured survey may all represent the same building, but they are produced for different decisions and to different tolerances.

Before reviewing dimensions, establish what the file is intended to support. If it is being used to test massing or net internal areas, a lower level of detail may be appropriate. If it will inform conservation repairs, joinery, steelwork or coordination with a Revit model, the required confidence is much higher. The acceptable tolerance should be agreed before the drawing is relied upon.

Ask for the survey specification or delivery notes. These should explain the survey method, the areas included, the output scale, the coordinate or level datum, and any known exclusions. A dependable documentation provider will be clear about limitations rather than presenting every line as equally certain.

It also helps to separate three related issues:

• Capture accuracy is the reliability of the site measurement or laser scan data.

• Registration accuracy is how well multiple scans or survey stations align with one another.

• Drawing accuracy is how faithfully the CAD geometry has been extracted and interpreted from the captured information.

• Completeness is whether the drawing includes all the geometry needed for the intended design decision.

A highly accurate point cloud does not automatically produce a fully accurate drawing. Openings, obscured junctions, curved walls and historic distortions still need skilled interpretation.

How to check CAD accuracy against site control

The fastest useful check is to compare the CAD file with independent control dimensions. Choose dimensions that matter to the work ahead rather than measuring only convenient clear spans. A room width, a façade length, a structural grid bay and a floor-to-ceiling height will usually reveal more than several small internal dimensions.

Take measurements between stable, identifiable points. Finished wall faces can vary, particularly in older buildings with uneven plaster or panelling, so use the same reference points shown in the drawing wherever possible. Record where each check was taken and whether it is a clear measurement, an offset or an approximate reading constrained by access.

Then compare these checks to the CAD geometry. A single mismatch may be caused by the way a dimension has been presented. Repeated differences in one direction are more concerning. For example, if several rooms are consistently 30 mm narrower than shown, the issue may be a scale error, an incorrect wall reference line or a drawing that has been adjusted without revisiting the wider geometry.

Do not rely only on chained dimensions. A set of individual room dimensions can all appear correct while the overall building length is wrong. Check both local dimensions and longer control runs across the plan. The same principle applies vertically: verify floor levels, ceiling heights, soffits and roof geometry against a consistent datum.

Check units, scale and coordinate behaviour

Unit errors remain a common cause of CAD problems. Confirm whether the drawing is in millimetres, metres or another unit, and check that dimensions report correctly in the file rather than only on a plotted sheet. A 1,000-fold unit issue is obvious, but smaller scaling errors can survive unnoticed when a drawing has been copied between consultant files.

Open the drawing at 1:1 in model space and measure a known dimension. Check that text, blocks and hatches have not been used to disguise geometry that is incorrectly scaled. If the drawing will be linked into Revit or used with survey control, confirm the origin, northing and easting values, rotation and project datum before anyone begins modelling.

For multi-storey projects, test whether each floor aligns to a common reference. A plan may be accurate in isolation yet be offset from the floor below. That can create misleading riser positions, façade lines and structural coordination issues later.

Use point clouds properly, not as a visual backdrop

Where laser scan data is available, overlay the CAD drawing on the point cloud in the correct coordinate system. This is one of the strongest ways to assess whether walls, openings, columns and roof elements follow the captured condition.

Check the overlay at more than one scale. At a wider view, look for systematic drift across a wing or long corridor. At a closer view, inspect critical junctions such as wall returns, door reveals, beam soffits, chimney breasts and changes in level. Geometry should correspond to the appropriate surface in the point cloud, not simply sit near it.

Be careful with what the cloud can and cannot prove. Reflective surfaces, deep shadows, concealed edges, dense vegetation and inaccessible roof areas can leave incomplete data. Historic buildings also introduce genuine irregularity: walls may taper, corners may not be square and floors may fall noticeably across a room. Simplifying these conditions may be appropriate for an early-stage model, but it should be intentional and documented.

A useful question is: does the CAD show a measured condition, a sensible simplification, or an assumption? Each has a place, but they should not be confused.

Review critical geometry, not every line equally

Accuracy checking should be risk-led. Give the closest scrutiny to areas where an error would change the design, programme or cost. These commonly include party walls, stair openings, structural zones, roof junctions, service risers, façade set-backs and interfaces with retained fabric.

On heritage and complex-geometry projects, inspect non-orthogonal spaces with particular care. A plan that forces an irregular room into square geometry may be acceptable for a broad area schedule, but not for fitting new cabinetry, setting out repairs or producing coordinated elevations. Curved walls should be checked using multiple points or arcs that reflect the measured profile, rather than a single approximate curve.

Elevations and sections deserve the same discipline as plans. A plan can be dependable while the vertical information remains assumed. Check window head heights, cill levels, ridge lines, parapets, floor build-ups and changes in ceiling level. If these elements affect planning, conservation consent or technical coordination, they should be supported by measured evidence.

Inspect the CAD file as a working document

A technically accurate drawing can still be difficult to use if its file structure is poor. Before issuing it to a wider team, inspect layers, xrefs, blocks, linework and annotations. The drawing should be legible at the intended plotted scale, with a clear distinction between existing, proposed and survey information where relevant.

Look for duplicate lines, exploded blocks, untrimmed geometry and unexplained hatches. These do not always alter measured accuracy, but they can undermine quantities, model conversion and coordination. Check that external references are included or properly bound, and that no critical geometry is sitting on frozen, non-plotting or unnamed layers.

For BIM workflows, confirm the agreed exchange requirements early. A CAD base may need consistent storey naming, survey levels and reference grids before it becomes a dependable Revit model. The cleaner the source information, the less time is spent correcting assumptions during modelling.

Record discrepancies and agree the next action

The outcome of an accuracy check should not be a vague judgement that the drawing is “close enough”. Record the checks made, the dimensions tested, the discrepancies found and the areas that remain unverified. This creates an audit trail and lets the project team make proportionate decisions.

Minor discrepancies in low-risk areas may be accepted with a note. Material differences should trigger a targeted site revisit, additional scan coverage or a revised drawing issue. It is usually more efficient to resolve a doubtful opening or roof junction before design develops around it than to correct several consultant packages later.

For projects involving irregular, listed or difficult-to-access buildings, the strongest approach is to commission documentation at the level the design genuinely requires. Precision-first survey capture, clear tolerances and structured CAD outputs give the team a dependable starting point without adding unnecessary detail everywhere.

The most useful drawing is not the one that claims perfection. It is the one that makes its accuracy, limits and evidence clear enough for the next design decision to be made with confidence.