Scan to BIM: Turning a Point Cloud Into a Revit Model

Quick answer: Scan to BIM is the process of turning laser scan data of an existing building into a modelled Revit file. A survey crew captures the building as a point cloud, and a modelling team then builds walls, floors, columns, roofs, openings and whatever services were asked for, each one placed to match the scan within an agreed tolerance. The scan is measurement. The BIM model is interpretation. You are paying for the second part, and the quality of the result depends almost entirely on how precisely you specify it.

What the two halves of the job actually are

The phrase joins two separate pieces of work, and most disappointment comes from treating them as one.

The first half is capture. A surveyor sets a terrestrial laser scanner at a series of stations, registers them together, and produces a point cloud: many millions of coordinates in space, each one a place the laser struck a surface. The cloud is extremely accurate about where things were on the day of the visit. It is also completely dumb. It does not know a wall from a shelf, and it has no idea what sits behind the furniture.

The second half is modelling. Somebody opens the registered cloud in Revit, and builds real objects against it: a wall family with a thickness and a type, a floor with a level, a column on a grid, a roof with a pitch. Every object is a decision. Is this bulge in the plaster a wall that bows, or a flat wall with a rendered patch? Is that opening a door, a blocked door or a service hatch? The point cloud cannot tell you. A person has to look, judge and commit.

That is why two suppliers can return very different files from the same cloud, both of them technically correct. The cloud is the evidence. The model is the argument made from it.

What arrives from the survey

• The registered point cloud, usually as an RCP or RCS file for Revit, sometimes as E57 or LAS for other software. Registration is the step that stitches the separate scan positions into one coherent coordinate system, and poor registration shows up later as walls that refuse to line up.
• Panoramic photography, captured from each scan station. This matters far more than people expect, because it is how the modeller identifies what a cluster of points represents.
• A control report, stating the accuracy achieved and how the scan was tied to a benchmark or a site grid. Without it, nobody can say what the tolerance of the finished model means.
• A coverage note, listing what was not captured. Locked rooms, occupied tenancies, ceiling voids and roof planes are the usual gaps, and each one becomes a modelling assumption later.

If a survey arrives without the photography or the control report, the modelling will still proceed, but a share of the model becomes inference rather than measurement. That is worth knowing before anybody signs off on it as a record of the building.

Level of detail: the decision that sets the price

There is no single thing called a scan to BIM model. There is a series of increasingly expensive ones, and the gap between them is enormous.

• Shell and core. External walls, floors, structural columns, roof and principal openings. Enough to design into, to calculate areas, and to coordinate a new fit out against. The fastest and cheapest option, and the right one more often than clients assume.
• Architectural detail. Internal partitions, doors and windows as proper families, stairs, ceilings, mouldings and skirtings. This is the usual level for refurbishment design and for planning submissions.
• Structural and services. Beams, slab soffits, ductwork, pipework, cable trays, sprinklers and plant. The most expensive level by a wide margin, because services are visually chaotic, constantly cross each other and are frequently hidden behind the very surfaces that were scanned.
• Heritage and ornamental. Mouldings, carved detail and irregular historic fabric modelled as geometry rather than simplified to a type. Occasionally necessary, rarely necessary for the whole building.

Alongside level of detail sits tolerance, which is the other half of the specification. A model built to plus or minus 10mm against the cloud is a different quantity of work from one built to plus or minus 50mm, and for most refurbishment design the looser figure is perfectly adequate. Asking for tight tolerance everywhere is the most common way to pay for precision that nothing downstream uses.

The useful discipline is to vary both by zone. Tight tolerance and full detail around the areas where new work meets old fabric. Shell level everywhere else. A sensible supplier will propose exactly that; a supplier quoting one number for the whole building has not asked what the model is for.

What the finished model should contain

A model is only useful if the information in it is addressable. That means the objects have to be real Revit objects, correctly categorised, not geometry pushed into roughly the right shape.

Walls should be walls, with types and thicknesses, not generic masses. Levels and grids should exist and be named to match whatever convention the project uses. Rooms should be bounded and named, because area schedules are one of the main reasons the model gets commissioned. Doors and windows should be family instances with dimensions, not voids cut into walls.

The deliverable list should also state what is deliberately absent. A model with no ceilings, no services and no furniture is a legitimate product if that was the agreement. It becomes a dispute only when the scope never said so in writing.

One more item belongs in the handover: a short note recording the assumptions made where the cloud was silent. Floor build ups that were never visible. A wall thickness inferred from one measured reveal. A roof plane reconstructed from below. Three paragraphs of honest notes make the model trustworthy for the next person who opens it. Their absence makes every number in it a guess of unknown provenance.

Where it fits against other ways of documenting a building

Scanning is not the only route, and it is not always the correct one.

For a small or simple building, a measured survey by hand or with a laser distance meter, drawn up into 2D CAD, gives you an accurate record at a fraction of the cost. A two bedroom flat does not need a point cloud. See as-built floor plans for existing buildings for how that route works.

Where drawings already exist on paper or in an old file format, the work is conversion rather than capture. Redrawing from existing documents is far quicker, with the obvious caveat that it inherits every error the old drawings contain. See old plans to CAD conversion and redraw.

Where the question is which existing conditions drawing you actually need, the distinction between a survey and a drawing set is the one that catches people out. See as-built survey versus as-built drawings.

Scanning earns its cost when the building is large, geometrically irregular, hard to access safely, or when the consequence of a dimensional error is expensive. Towers, industrial plant, heritage fabric, anything where a prefabricated element has to fit a space measured once.

What it costs, and what moves the number

Rendimension quotes scan to BIM by scope after reviewing the cloud, because the same floor area can represent wildly different amounts of modelling. Published pricing on the 3D rendering cost page covers visualization deliverables, and the modelling that feeds them follows the same logic: the model is built once, and everything produced from it afterwards is comparatively cheap.

Four factors move a scan to BIM quotation more than floor area does.

• Level of detail, above all. Adding services to an architectural model can double the work.
• Tolerance, because tight tolerance means checking every object against the cloud rather than placing it sensibly.
• Cloud quality, meaning registration accuracy, station density and whether the panoramic photography exists. A clean cloud models quickly. A noisy one with occlusions is slow and produces more assumptions.
• Repetition, which cuts hard the other way. Forty near identical hotel rooms are a fraction of the work of forty different ones, because the family and the type are built once.

How to specify it so you get what you wanted

Write six lines before you ask anybody to price it. They shorten the schedule and the quotation more than any conversation about scanner models.

• What the model is for. Refurbishment design, area schedules, clash coordination, planning submission, facilities handover. Each implies a different model.
• Level of detail per zone, not one figure for the building. Name the zones.
• Tolerance per zone, stated as a number.
• Which disciplines, architectural, structural, mechanical, electrical, and which of them stop at principal runs.
• Software and version, because a Revit file saved forward cannot be opened back, and a consultant on last year's version will not thank you.
• Who signs it off, one person, with the date they need it.

Once the model exists, it is also the cheapest source of everything else. Visualization, floor plans and marketing imagery all come from the same geometry without rebuilding it. See the Revit to 3D rendering workflow and BIM visualization explained. If clash detection is the goal, the relevant scope is covered in BIM coordination and clash detection.

The honest summary

Scan to BIM is a reliable way to get a trustworthy model of a building that already exists, and it is routinely oversold. The scanning is the easy, well understood, highly accurate part. The modelling is where the cost, the judgement and the risk live, and it is bounded entirely by the specification you hand over.

Specify the purpose, vary the detail by zone, state the tolerance as a number, demand the assumptions note, and insist on real Revit objects rather than decorative geometry. Do those five things and the model will serve the project for years. Skip them and you will own a very expensive file that nobody downstream trusts.

Common questions

What does scan to BIM mean?

It means converting laser scan data of an existing building into a modelled BIM file, usually in Revit. The scan produces a point cloud of measured coordinates, and a modelling team then builds walls, floors, columns, roofs and openings against that cloud to an agreed level of detail and tolerance.

Is a point cloud the same thing as a BIM model?

No. A point cloud is raw measurement: millions of coordinates with no meaning attached. A BIM model is interpretation, where each object has been identified, categorised and given properties. You can measure from a cloud, but you cannot schedule, coordinate or design from it.

What level of detail should I ask for?

Vary it by zone rather than specifying one level for the whole building. Full architectural detail and tight tolerance where new work meets existing fabric, shell and core elsewhere. Services modelling is the single largest cost increase, so only include it where coordination genuinely requires it.

What tolerance is normal for a scan to BIM model?

Refurbishment design is usually served well by plus or minus 25mm to 50mm against the cloud, with tighter figures reserved for interfaces where prefabricated elements have to fit. Asking for tight tolerance across an entire building is the commonest way to pay for accuracy that nothing downstream uses.

What do I need to supply to get a model built?

The registered point cloud as RCP, RCS or E57, the panoramic photography from each scan station, the surveyor control report stating achieved accuracy, and a coverage note listing what was not captured. The photography matters most, because it is how the modeller identifies what the points represent.

Do I always need scanning, or will a measured survey do?

A measured survey drawn into 2D CAD is cheaper and perfectly adequate for small, regular buildings. Scanning earns its cost on large, irregular, hard to access or heritage buildings, and anywhere a dimensional error would be expensive to discover on site.

How much does scan to BIM cost?

It is quoted by scope after reviewing the cloud, because the same floor area can mean very different amounts of modelling. Level of detail, tolerance, cloud quality and repetition move the number far more than floor area does. Forty identical hotel rooms cost a fraction of forty different ones.

Can the model be used for rendering and marketing imagery afterwards?

Yes, and that is one of the strongest arguments for building it properly. Once the geometry exists, visualization, 3D floor plans and marketing imagery come from the same model without rebuilding it, which is why additional deliverables from an existing model are substantially cheaper than the first one.

Frequently asked questions

What does scan to BIM mean?

It means converting laser scan data of an existing building into a modelled BIM file, usually in Revit. The scan produces a point cloud of measured coordinates, and a modelling team then builds walls, floors, columns, roofs and openings against that cloud to an agreed level of detail and tolerance.

Is a point cloud the same thing as a BIM model?

No. A point cloud is raw measurement: millions of coordinates with no meaning attached. A BIM model is interpretation, where each object has been identified, categorised and given properties. You can measure from a cloud, but you cannot schedule, coordinate or design from it.

What level of detail should I ask for?

Vary it by zone rather than specifying one level for the whole building. Full architectural detail and tight tolerance where new work meets existing fabric, shell and core elsewhere. Services modelling is the single largest cost increase, so only include it where coordination genuinely requires it.

What tolerance is normal for a scan to BIM model?

Refurbishment design is usually served well by plus or minus 25mm to 50mm against the cloud, with tighter figures reserved for interfaces where prefabricated elements have to fit. Asking for tight tolerance across an entire building is the commonest way to pay for accuracy that nothing downstream uses.

What do I need to supply to get a model built?

The registered point cloud as RCP, RCS or E57, the panoramic photography from each scan station, the surveyor control report stating achieved accuracy, and a coverage note listing what was not captured. The photography matters most, because it is how the modeller identifies what the points represent.

Do I always need scanning, or will a measured survey do?

A measured survey drawn into 2D CAD is cheaper and perfectly adequate for small, regular buildings. Scanning earns its cost on large, irregular, hard to access or heritage buildings, and anywhere a dimensional error would be expensive to discover on site.

How much does scan to BIM cost?

It is quoted by scope after reviewing the cloud, because the same floor area can mean very different amounts of modelling. Level of detail, tolerance, cloud quality and repetition move the number far more than floor area does. Forty identical hotel rooms cost a fraction of forty different ones.

Can the model be used for rendering and marketing imagery afterwards?

Yes, and that is one of the strongest arguments for building it properly. Once the geometry exists, visualization, 3D floor plans and marketing imagery come from the same model without rebuilding it, which is why additional deliverables from an existing model are substantially cheaper than the first one.