Point clouds and digital twins: turning a scan into a model teams can use

A site scan can capture a building in a few hours and return millions of measured points, each one a precise, three-dimensional coordinate recorded by a laser that measures the time light takes to bounce off a surface and return. On its own, that point cloud is not a model, it is closer to a very accurate but completely unlabelled photograph. Turning it into something a design team can actually use, open in their own software, measure against, and build on, is where the real work happens, and it is usually underestimated by anyone who has not done it before.

What a scan actually gives you

The raw output of a laser scan is a dense set of coordinates, accurate typically to a few millimetres, but entirely unstructured. Walls, pipes and beams are not labelled as such by the scanner; they exist in the data only as clusters of points that happen to line up in a way a human, or specialised software, can recognise as a wall, a pipe or a beam. A single scan of a modest building can easily produce hundreds of millions of individual points, far more data than any design software can sensibly work with directly, which is exactly why the next stage exists.

From points to a model

Modellers, working either manually or with the help of increasingly capable automated recognition software, trace the geometry the scan implies, checking it against reference photos taken during the scan and against known dimensions from existing drawings where they are available and trustworthy. The result is a proper 3D model, made of walls, floors, structural elements and often building services, tied to real-world coordinates through the same control points used to register the original scan. That model can then be opened in ordinary design software and compared directly against the design intent, something a raw point cloud simply cannot do.

Where the value actually shows up

Comparing an as-built scan-derived model against the original design model is where most of the financial value of scanning is realised, since it catches clashes, deviations and construction errors before they become expensive to fix, sometimes before the next trade has even started work on top of the mistake. It is also, separately, the fastest and often the only reliable way to document a building that has no trustworthy existing drawings, which is common in refurbishment work on older buildings where original drawings, if they still exist at all, rarely match what was actually built decades ago.

What drives the cost of a scanning project

The biggest cost driver is rarely the scanning itself, which is fast, but the modelling that follows it, since a human or a piece of software still has to interpret those millions of points into meaningful, labelled geometry. The level of detail requested has an outsized effect on that cost: a model that only needs to show wall positions and floor levels is far cheaper to produce than one that needs to show every pipe, duct and cable tray, so agreeing the required level of detail before the scan is commissioned, rather than after, avoids paying for detail nobody asked for or discovering too late that the model is missing detail the project actually needed.

Keep the scope tight

Scanning every part of a site or a building to the same level of detail is rarely worth it, since areas of straightforward, unchanging geometry add cost without adding useful information. Agree in advance which areas genuinely need survey-grade accuracy, typically the areas where new work will connect to existing structure, and which areas just need a general record for context, and the project stays affordable without losing the accuracy that actually matters to the design team.