A BIM quantity take-off is the fastest way to measure a repetitive building and the fastest way to produce a confident wrong answer. Both come from the same property: the software will schedule whatever is in the model, completely and consistently, without any view on whether the model describes the building that will be built.
Manual measurement is slower and error prone, but the errors are usually visible as gaps. Model errors present as clean tables of numbers, and a clean table is very hard to distrust.
Design intent models are not measurement models
Most architectural models exist to communicate design intent, coordinate with other disciplines and produce drawings. They are built to be looked at and to be checked for clashes. In practice they are not built to be counted.
That produces a specific set of consequences. Objects are often generic placeholders standing in for products not yet selected. Assemblies are modelled as single elements where the real construction has several layers. Elements are drawn to convenient boundaries rather than to construction joints.
None of that is a failing on the designer's part. It is what the model was for. It simply means the schedule it produces answers a different question from the one a cost consultant is asking.
What is usually absent
Builder's work in connection is the classic omission. Holes, chases, fire stopping, supports and the making good around services are rarely modelled and represent real cost.
Temporary works are absent almost by definition. Propping, scaffolding, edge protection, temporary roofs and access are construction items, not design items, and no design model contains them.
Below ground, models thin out quickly. Drainage runs may be indicative, service diversions may be missing entirely, and the substructure may be a simplified representation of a design that has not yet been engineered.
Finishes and small components tend to be modelled where they are visible and omitted where they are not. Ironmongery, sealants, trims and fixings almost never appear.
What is present but wrong for measurement
Wall areas commonly run through floors and openings in ways that suit drawing production but do not match how the work will be built or priced. Layered construction may be modelled as a single element carrying a thickness but no build up.
Duplicate and overlapping geometry produces double counting that is invisible in a schedule. Objects placed on the wrong level, or with the wrong type assigned, produce quantities in the right total and the wrong place.
Level of detail varies within a single model, often between disciplines and often between the parts a designer has developed and the parts they have not. An export averages across all of it and presents the result with equal confidence.
The reconciliation nobody wants to do
The only reliable way to use a model for measurement is to reconcile it against the drawings and against a manual check of the significant elements. Not everything, but the items that carry the value.
On a residential scheme that usually means the envelope, the frame, the internal division and the repeated unit fit out. Check one instance manually against the model output, and the reliability of the whole schedule for that element is established in an hour.
Where the two disagree, the disagreement is information. It usually reveals either a modelling convention that needs adjusting or a design issue nobody had noticed. Both are worth finding before tender rather than after. Our approach to this sits under quantities and take-off.
Where model measurement genuinely wins
None of the above is an argument against model based take-off. On a scheme with repetition it is transformative, because a design change can be requantified across every instance in hours rather than weeks.
That speed changes what is possible commercially. Option testing becomes affordable. Design changes can be priced while the decision is still open rather than after it has been taken. Reconciliation between design issues becomes a routine exercise rather than a project in itself.
The condition is that the model has been checked once, properly, at the start. After that the checking is incremental. It is the first pass that costs, and skipping it is what turns the advantage into a liability.
Mixed use makes it harder
On a mixed use scheme the model problem compounds, because different parts of the building are at different stages of design development and are often modelled by different teams to different conventions.
A residential element may be fully developed while the commercial shell is indicative and the public realm exists only as a site plan. A single export across all of it produces a quantity set whose reliability varies enormously between elements, with nothing in the output to indicate which is which.
Holding those in one schedule on one basis of measurement is a discipline rather than a software setting, and it is set out further under mixed use and regeneration.
What this means for you
Ask two questions of anybody presenting model derived quantities. What was reconciled against the drawings, and what is known to be absent from the model. A measurer who has done the work will answer both immediately.
If the answer is that the schedule came straight out of the software, the quantities are a starting point rather than a measurement. They may well be close. Equally, they are not yet defensible, and defensible is the standard a tender document has to meet. The failure mode when it does not is described in incomplete design information.
Have a model and no idea whether it can be measured?
Send the file and the drawing set. We will tell you what the model supports, what it does not, and what that means for the cost position.