What Accuracy Do Architects Need on Site?

A 20 mm discrepancy may be irrelevant when testing a broad office layout. The same discrepancy can prevent a bespoke stair from fitting between listed walls, compromise a façade package, or create a difficult conversation on site. That is why the question, what accuracy do architects need, has no single numerical answer. The required accuracy should follow the decision the information must support.

For existing buildings, the aim is not to pursue the smallest possible number for its own sake. It is to establish dependable geometry at the points where design, coordination and construction depend on it. A survey that is accurate enough for feasibility may not be accurate enough for planning a complex roof alteration, setting out new steelwork, or recording historic fabric.

What accuracy do architects need for design work?

Architects need accuracy that is proportionate to project stage, building complexity and consequence of error. Early-stage massing can often work with lower-detail information, provided the overall footprint, levels and principal structural grid are reliable. Detailed design requires a clearer record of wall positions, openings, soffits, service zones and changes in level. Fabrication, conservation and interface-heavy work may require verification of individual critical dimensions beyond the general survey tolerance.

The useful question is therefore: where would an error change the design, cost, programme or consent position? Those locations deserve closer capture and clearer reporting.

For a typical measured building survey, an agreed accuracy in the region of +/-10 mm is often suitable for architectural design documentation. This is not a universal rule. Irregular stone walls, complex historic roofs, tight fit-out interfaces and concealed or inaccessible areas all need a more considered approach. A drawing can be produced neatly at a stated scale while still omitting the feature that matters most to the scheme.

Indicative tolerances by use

| Intended use | Typical accuracy approach | What must be clear | | --- | --- | --- | | Feasibility and area studies | Reliable overall geometry, often with lower detail | Footprint, principal dimensions, floor levels and major constraints | | Planning and concept design | Measured plans, elevations and sections with dependable building form | Openings, ridge and eaves lines, neighbouring context and level changes | | Detailed design and coordination | Tighter control of geometry and coordinated 2D or BIM outputs | Structural zones, soffits, service routes, interfaces and irregular walls | | Heritage, retrofit and fabrication-sensitive work | Project-specific tolerances with focused verification | Historic fabric, deformation, junctions, bespoke elements and critical clearances |

These ranges are a starting point for a brief, not a substitute for one. The appropriate tolerance should be agreed before site capture, particularly where a design team is relying on the information to resolve a tight interface.

Accuracy is more than a laser scanner specification

A high-accuracy scanner is valuable, but scanner accuracy alone does not define the accuracy of a finished plan or Revit model. The final result depends on the whole documentation chain: control, scan registration, coverage, interpretation, modelling conventions, quality assurance and the clarity of the deliverable specification.

A point cloud records sampled surfaces, not design intent. It may show the face of plaster, the edge of a skirting board or the exposed face of masonry. The modeller must decide which condition is relevant to the required output. On a listed building, a wall may bow, taper and step across its height. Modelling it as a straight line can make a plan appear cleaner, but it can remove geometry that is essential to a conservation or fit-out decision.

This is why accuracy and precision should be separated. Precision is the repeatability or fineness of a measurement. Accuracy is how closely the documented geometry represents the real building condition. A dense point cloud may be highly precise, yet a model can still be inaccurate if it simplifies the wrong feature or is aligned poorly.

Set the tolerance around risk, not drawing scale

Drawing scale is a helpful communication convention, but it should not be the sole basis for accuracy requirements. A 1:100 planning drawing and a 1:50 construction drawing have different purposes, yet either may contain a critical dimension that needs close confidence.

Start with the risk profile of the project. A regular, modern commercial floorplate usually allows a more straightforward survey strategy than a converted mill, a Georgian townhouse with uneven party walls, or a roofscape altered repeatedly over centuries. The latter may need more scan positions, carefully coordinated control, additional sections and explicit recording of non-orthogonal geometry.

Consider also the cost of discovering an error late. If the information will only support an initial options appraisal, broad reliable geometry may be the right investment. If it will drive a package of bespoke joinery, structural interventions or prefabricated components, it is usually more economical to define critical tolerances at the outset and verify them before manufacture.

The dimensions that are most often missed

Most costly survey issues do not arise from a room being a few millimetres out in isolation. They arise where incomplete information hides an interface. Ceiling void depths, beam undersides, floor build-ups, chimney breasts, window reveals, roof slopes and offsets between floors can all affect whether a design can be built as drawn.

Levels require particular attention. Architects commonly need more than a single finished floor level for each storey. Existing thresholds, landings, slab changes, external ground levels, head heights and drainage falls may all influence compliance, accessibility and coordination. If levels are captured without a stated datum or are not carried consistently across drawings and model views, teams can reach different conclusions from the same survey.

External geometry deserves the same care. Roof plans are frequently treated as secondary until planning, drainage or access design begins. On complex properties, the roof may be the most irregular part of the building. Ridge lines, valleys, parapets, abutments, chimney stacks and concealed gutters should be documented at the level required by the proposed work rather than inferred from a simplified elevation.

Match the output to the people using it

A dependable survey output is not simply a file that opens. It should be structured for the recipient's workflow. For CAD users, this means clear layers, sensible linework, coordinated plans, elevations and sections, and dimensions that can be trusted. For BIM teams, it means an agreed level of development, logical element categorisation, usable model origin and a clear distinction between observed geometry and assumptions.

LOD should not be treated as a guarantee of geometric accuracy. A model at LOD300 or LOD400 can contain a great deal of detail, but that detail is only useful if its location and interpretation are appropriate. Conversely, an LOD200 model can be highly dependable for its intended purpose when its geometry and limitations are clearly defined.

The brief should state whether the team needs a registered point cloud, 2D measured drawings, a Revit model, or a combination. It should also identify the areas requiring enhanced detail. A focused instruction such as “verify headroom beneath all retained beams” is far more useful than a generic request for “high accuracy”.

Build verification into the workflow

No existing-building survey can see through every finish, ceiling or locked door. The professional standard is not to pretend uncertainty does not exist. It is to identify limitations, distinguish observed from inferred conditions, and arrange targeted verification where it will affect the next decision.

A sound workflow begins with a short technical conversation before the visit. The documentation team should understand the intended design use, proposed interventions, access constraints and required outputs. During capture, difficult areas can receive additional coverage rather than being treated as standard rooms. During production, checks between plans, sections, elevations and the point cloud help prevent inconsistencies from entering the final issue.

For complex, irregular or heritage buildings, this approach is particularly valuable. Space Captures uses precision-first laser capture and structured CAD or Revit documentation to make difficult existing conditions usable for design teams, while being clear about the scope and limitations of each deliverable.

Specify accuracy in a way that can be delivered

The best survey briefs describe outcomes, not just a headline tolerance. State the intended stage of work, the deliverables, the coordinate or level datum where relevant, and the zones where fit is critical. Identify whether walls should represent finished faces, structural faces or a measured centreline convention. Ask how irregular surfaces, inaccessible areas and assumptions will be represented.

It is equally useful to agree what is outside scope. A measured survey is not automatically a structural investigation, services survey or opening-up exercise. Clear boundaries prevent a design team from treating absent information as confirmed information.

The right accuracy gives architects confidence to move forward without paying for unnecessary detail everywhere. If a decision depends on a dimension, level or irregular form, make it explicit before capture. That small act of definition is often what turns existing-building information from a potential risk into a dependable basis for design.