What LOD actually describes
LOD stands for Level of Development, not Level of Detail — the distinction matters. The framework in common use across the architecture and construction industry, maintained by BIMForum and built on the original AIA documentation of the concept, defines LOD as how much of a model element's geometry and attached information a downstream team can reliably rely on for a specific use, at a specific stage.
A wall modeled at LOD 100 might be a placeholder mass with no real dimensions. The same wall at LOD 300 has accurate size, shape, location, and orientation, modeled as a specific system. At LOD 400 it carries fabrication-level detail: connections, penetrations, and the information a shop would need to build from it directly. LOD is about what the model element can be trusted to tell you and what decisions it can support — not about how many points went into measuring it.
LOD 100 through 400 in practice
For a model built from an existing-conditions 3D scan, the lower tiers are rarely the deliverable — you are documenting something that already exists, not conceptualizing something that doesn't, so LOD 100 and often LOD 200 show up more in new design than in as-built work. Here is what each tier typically means when the starting point is a registered point cloud rather than a design concept.
| LOD | What it means | Typical use in scan-to-BIM |
|---|---|---|
| LOD 100 | Conceptual — generic massing, no reliable geometry or size. | Rare as a scan-to-BIM deliverable; occasionally used for a rough context shell around a detailed area. |
| LOD 200 | Approximate geometry — generalized size, shape, and location; not system-specific. | Common for secondary or low-priority zones where an as-built needs to exist but won't drive design decisions. |
| LOD 300 | Accurate geometry — specific size, shape, location, and orientation, modeled as the actual system. | The default for most design-facing as-built work: walls, structure, ductwork, and equipment a design team will work against. |
| LOD 350 | LOD 300 plus interfaces to other systems — how elements connect, clash, and coordinate. | Used where MEP coordination or clash detection between trades is part of the deliverable's purpose. |
| LOD 400 | Fabrication-level detail — assembly, connection, and installation information. | Reserved for the specific zones and systems a shop will build or install from directly, not the whole building. |
LOD is not accuracy — meet Level of Accuracy
LOD tells you how developed an element is. It says nothing about how closely that element matches the physical building it represents. That second question — how well does the model geometry agree with the measured reality — is what the USIBD Level of Accuracy specification addresses, and it is a separate axis entirely.
The USIBD (U.S. Institute of Building Documentation) Level of Accuracy specification defines a series of tiers, generally referred to as LOA, that describe the permitted deviation between a modeled element and the point cloud it was built from — running from coarse, walk-and-sketch-level agreement up to survey-grade tolerances tight enough for fabrication. The specification lays out the exact tolerance ranges and measurement methods tier by tier; rather than restate specific numeric thresholds here, the right move for a real project is to pull the current USIBD spec and reference the tier by name in the contract, because that is the document a QA report gets checked against.
The practical consequence: a model can be LOD 300 and LOA10 (developed as a real system, but loosely tied to measured reality), or LOD 300 and a tight LOA tier (developed as a real system, and verified against the scan to a defined tolerance). A buyer who asks only for an LOD number and never specifies an LOA target has left the accuracy of the model undefined — which is exactly the gap a low-cost modeler can fill with a model that looks right and isn't.
What to actually specify when you request scan-to-BIM
A request that only says "LOD 300, whole building" under-specifies the deliverable. What actually controls cost, schedule, and whether the model does its job is a modeling matrix: which systems, in which zones, at which LOD, verified to which LOA.
- LOD by system, not one number for the whole building — structure and walls at one tier, ductwork or process piping at another, if that's where the design decisions live
- LOD by zone — most projects need one or two areas modeled to fabrication-grade detail and the rest at a lower, cheaper tier
- An LOA target, referenced by tier name against the current USIBD specification, so accuracy is something a registration and modeling QA report can be checked against
- A minimum size threshold — the diameter or dimension below which an element is not modeled at all, which keeps small conduit and cable tray from bloating the model
- The native file format the model has to open in (RVT, IFC, or both) and the software version your team is standardized on
- Whether the point cloud stays linked in the delivered file, so any modeled element can be checked back against the measurement that produced it
How LOD choice changes cost and schedule
LOD and LOA both move price and time, but not in the way people expect. Field capture cost is driven mostly by area, geometric density, and access — not by the LOD the model will eventually be built to, because the scanner captures what is physically there regardless of how it gets modeled afterward. Modeling cost is where LOD matters: LOD 300 modeling takes substantially more time per element than LOD 200, because the modeler is now representing the actual system geometry rather than a generalized shape, and LOD 400 adds fabrication detail on top of that.
A tighter LOA target adds time differently — it can mean more scan setups to eliminate occlusion and reduce noise in a given area, more control points to hold registration tolerance, and more QA time checking modeled elements against the cloud. The honest way to control the number is to be narrow about where the highest tiers apply, not to negotiate a blanket LOD down across a whole project.
Common mistakes worth avoiding
The single most expensive mistake is specifying a uniform high LOD across an entire building when only a few systems or zones actually drive a decision. A design team doing a tenant fit-out generally needs LOD 300 walls and structure and does not need LOD 400 ductwork three floors away from the scope. Specifying LOD 400 everywhere for what is fundamentally an as-built record — rather than a fabrication package — pays for detail nobody downstream will use.
The second is treating LOD as a proxy for accuracy and skipping the LOA conversation entirely. A model can hit every LOD 300 requirement on paper — real systems, correctly categorized, plausible sizes — while sitting well outside a usable tolerance of the actual building, because nothing in the LOD definition itself checks that. The fix is simple: specify both, and ask for the registration and QA report that shows the LOA target was actually met.
The third is assuming LOD is fixed once modeling starts. A design that changes scope midstream, or a zone that turns out to matter more than expected, should trigger a conversation about revising the modeling matrix — not a silent decision by whoever is at the keyboard that day.