What Scan-to-BIM Means

Start with the distinction that the whole module turns on, because almost every misunderstanding about scan-to-BIM comes from blurring it. A point cloud is a set of measured points — each just a position in space, perhaps with a colour and an intensity value. That is *all* it is. Thirty million points describing a room carry no idea of "room", "wall", "door" or "floor"; they are thirty million independent dots that happen, to your eye, to form recognisable surfaces. The cloud is geometry without meaning. You can measure distances between points, slice through them, and admire them, but the software has no notion that this cluster of dots is a load-bearing column and that cluster is a curtain.

A BIM model is the opposite kind of thing. It is built from objects — a wall object, a floor object, a column, a door, a pipe — and each object carries *semantics* (it knows what category of thing it is) and *parameters* (it knows its dimensions, its material, its type, its level, its relationships to other objects). A wall in a BIM model knows it is 230 mm brick, that it runs from this grid to that one, that it hosts a particular door, that it belongs to the ground floor. That meaning is exactly what lets you schedule every door automatically, run a clash check against the new services, calculate areas, or swap a wall type across the whole project in one move.

The gulf between them is the gulf between description and understanding. The cloud *describes* where surfaces are, faithfully and densely. The model *understands* what the building is made of and how it is organised. Scan-to-BIM is the human-and-software process of reading the description and authoring the understanding: looking at a plane of points and deciding "that is a wall, of this type, here", then creating the object to match. Crucially, this is an act of interpretation, not a mechanical conversion — which is why, as later lessons show, it is still largely done by skilled people rather than a single button. Hold onto the core image: a cloud is a measured ghost of the building; a BIM model is a thinking replica you can design and build against.

The deliverable of scan-to-BIM has a name worth getting right: an as-built model, though many practitioners prefer the more honest as-is model. "As-built" strictly means how the building was constructed; "as-is" means how it stands *today*, after decades of alterations, settlement, repairs and quiet undocumented changes. For renovation, retrofit, heritage and fit-out work, as-is is what you actually have and what you must design against — the building as it really is now, captured and modelled, rather than the fiction preserved in the original drawings (if any survive at all).

What does that model give you that the cloud could not? First, a reliable basis for design: you draw the new scheme directly onto a model of the true geometry — the out-of-plumb walls, the real floor-to-floor heights, the column positions as they actually sit — so the design fits the building instead of fighting it. Second, coordination and clash detection: because the existing structure and services are now objects, the new MEP and structure can be checked against them automatically, catching the clash on screen rather than on site. Third, quantification and scheduling: objects can be counted, measured and scheduled — areas, door lists, quantities — which feeds cost and planning. Fourth, a single shared source of truth: architect, engineer, contractor and client all work from one measured model rather than a drawer of contradictory old drawings. Fifth, a foundation for the building's whole life: the as-is model can become the seed of a facilities model or a digital twin, carrying forward into operation.

There is an honest limit to state up front. The model is only ever as good as the decisions made building it, and it represents the building to a chosen *level of accuracy and detail* — the subject of lesson 6.3. It is not automatically "the truth"; it is a deliberate, specified interpretation of the captured truth. And the binding accuracy of the underlying survey — whether the cloud is genuinely good to the tolerance claimed — remains the domain of the surveyor, the verified equipment specification and the governing standards, not something the modeller conjures. What scan-to-BIM delivers is an intelligent, usable model of the existing building; what it does not deliver is a guarantee of survey-grade accuracy unless the capture behind it was itself survey-grade and checked.

Scan-to-BIM matters for the same reason the whole course opened with: most architecture and interior work happens in relation to something that already exists, and the quality of everything downstream depends on how faithfully you understand that existing thing. A point cloud alone, for all its density, is awkward to design *inside*. You cannot host a new door in a cloud, cannot let a wall drive a schedule, cannot run the coordination checks a modern project depends on. The cloud is the measured evidence; the BIM model is the working basis. Scan-to-BIM is what converts evidence into a basis.

The payoff is avoiding the classic, expensive failure mode of work on existing buildings: designing against assumed or inaccurate information. The old drawings say the room is 4.2 m wide and the walls are square; the building says 4.13 m at one end, 4.19 m at the other, with a wall 40 mm out of plumb and a beam nobody drew. Design on the drawings and the joinery fouls, the steel is short, the ducts clash. Design on an as-is BIM model built from a good scan and those surprises are on your screen weeks before they would have been on site. In renovation, fit-out, heritage and adaptive-reuse work — and in the low-carbon push to reuse rather than demolish — this is transformative. It is also why scan-to-BIM pairs so naturally with prefabrication: you cannot safely manufacture a module to slot into an existing structure you have only guessed at, but you can against an accurate as-is model.

This is the point in the course where capture and BIM formally meet, and it is worth placing scan-to-BIM in that larger picture. BIM is the discipline of modelling buildings as structured object-based data across their whole life; scan-to-BIM is simply how you get an *existing* building into that discipline when it was never modelled in the first place. If you want to understand what a BIM model is, how it is structured, and how a project team works within it, that is the subject of the Studio Matrx course on Building Information Modelling — a natural companion to this module. Here we focus on the capture-to-model bridge; there you will find the world the model lives in once it exists.

Pull the threads together. Scan-to-BIM is the process of turning a registered, cleaned point cloud of an existing building into a structured BIM model — an *as-is* model in which the building's elements exist as intelligent objects with parameters, rather than as undifferentiated points. It is a bridge with three piers: the captured reality (the cloud, from photogrammetry, laser scanning or neural capture, registered and cleaned as Module 5 described), the interpretation (a skilled person deciding what each part of the cloud represents and to what level it should be modelled), and the model (the object-based deliverable the team designs, coordinates and builds against). Get all three right and you have replaced a drawer of stale, contradictory drawings with a single, measured, intelligent basis for the whole project.

This module walks the bridge end to end. This lesson set the concept: what scan-to-BIM means — a cloud is not a model, and the deliverable is an intelligent as-is model. Lesson 6.2, modelling from the point cloud, gets practical: how you actually reference the cloud and trace and fit walls, floors, columns and MEP to it, and the constant judgement of idealisation — when to model a wavy real wall as a clean plane, and when the waviness is the point. Lesson 6.3, levels of accuracy and detail, introduces the two specification concepts that govern the whole job — Level of Accuracy (LOA: how faithful to the real geometry) and Level of Detail/Development (LOD: how much detail and information) — and why both are driven by the *use*, with over-modelling wasting money and under-modelling misleading. Lesson 6.4, automating scan-to-BIM, gives the honest state of play: why this is still largely a manual, skilled craft, what the emerging AI and machine-learning tools can and cannot yet do, and a hype-guarded outlook.

One boundary runs through all of it, as through the whole course. Scan-to-BIM is a modelling and interpretation skill, and this module teaches it as such. The *binding accuracy* of the survey behind the model — whether the data is genuinely good to a stated tolerance, whether it is correctly georeferenced, whether it is fit for a legally or structurally binding purpose — is deferred to licensed surveyors and geospatial professionals, verified equipment specifications, and the governing standards and specifications. You will learn to build an excellent model from good data and to specify the accuracy and detail a job needs; you will also learn to know, and respect, where the survey professional's responsibility begins.