TL;DR
TL;DR: BIMForum LOD sets what is modeled and how developed it is; USIBD LOA sets how closely the point cloud (measured accuracy) and the model (represented accuracy) match the building. Specify both per element group, with a confidence level and a validation method. Our ±5mm registered accuracy is a measured figure; ±10mm at LOD 300 and ±6mm at LOD 400 are represented figures.
A scan-to-BIM spec needs two separate designations for every element group: a BIMForum Level of Development (LOD), which says what gets modeled and how far each element is developed, and a USIBD Level of Accuracy (LOA), which says how closely the measurements and the finished model must match the real building. LOD answers "what is in the model." LOA answers "how far off can it be." Neither implies the other: the USIBD guide states that any given LOA may apply to any given LOD. Name only one and the other is left to the provider's judgment, which is where most scan-to-BIM disputes start.
What does BIMForum LOD actually define?
The BIMForum LOD Specification is a reference that lets project teams "specify and articulate with a high level of clarity the content and reliability" of a model. It builds on the AIA's LOD definitions, is organized by CSI Uniformat 2010, and is meant to be used alongside a BIM Execution Plan, not in place of one. BIMForum lists the 2025 edition as current. The definitions below come from the 2024 Part I edition, which carries the updated AIA E201-2022 wording.
| LOD | What the definition says | Where it shows up on existing-conditions work |
|---|---|---|
| LOD 100 | Symbol or generic representation; not necessarily geometric | Rarely requested from a scan |
| LOD 200 | Generic placeholder, recognizable as the component, with approximate quantity, size, shape, location and orientation | Feasibility, test fits, early pricing |
| LOD 300 | Quantity, size, shape, location and orientation can be measured | Design background, construction documents |
| LOD 350 | LOD 300 plus measurable interfaces with adjacent or dependent elements; supports construction-level coordination | Selected tie-ins and coordination zones |
| LOD 400 | Detail sufficient for fabrication, assembly and installation; roughly shop-drawing level | Targeted fabrication zones only |
| LOD 500 | Graphic representation of an existing or as-constructed condition from observation, field verification or interpolation; the level of accuracy must be noted | Field-verified as-built elements |
Two passages matter more than the table for scan work. First, the spec says "There is no such thing as an 'LOD ### model.'" "Give us an LOD 300 model" is not a spec; "LOD 300 for the shell and structure, LOD 200 for the storage wing" is. Second, LOD 500 is not a step above LOD 400. BIMForum explains that it "indicates that the element's geometry is determined through observation of an existing item rather than design of a future item," that the definition requires the element's accuracy to be specified, and that BIMForum recommends the USIBD LOA Specification for that purpose. The LOD standard itself hands the accuracy question to LOA.
What LOD never gives you is a tolerance. An LOD 300 wall's size and location "can be measured"; nothing says how close that measurement is to the real wall.
What does USIBD LOA define?
The USIBD Level of Accuracy Specification sets out five levels, LOA10 through LOA50, each defined as a range of standard deviation. The two-digit numbering is deliberate, so nobody confuses LOA30 with LOD 300. USIBD's current release is Version 3.1, downloadable with a free account. The ranges below are from the publicly posted C120 Guide, Version 2.0 (2016); check them against the version your contract cites.
| Level | Upper range | Lower range |
|---|---|---|
| LOA10 | User defined | 5cm |
| LOA20 | 5cm | 15mm |
| LOA30 | 15mm | 5mm |
| LOA40 | 5mm | 1mm |
| LOA50 | 1mm | 0 |
All ranges are specified at the 95 percent confidence level, which the guide ties to a standard deviation of 2 sigma. Anything tighter than the level you specify is presumed acceptable.
The part most specs skip is that LOA is two values, not one:
- Measured accuracy is the standard deviation achieved by the final measurements, whatever the method. On a scan project, that is the registered point cloud.
- Represented accuracy is the standard deviation achieved once that data is processed into line work or a model. That is your Revit model.
The guide is blunt about the relationship: represented accuracy "can only be even or less accurate than the measured accuracy," and when measurements are processed into a deliverable, both should be specified. The specifier also chooses whether accuracy is absolute (tied to a reference frame such as a building or state coordinate system) or relative (within a region or set of setups), and picks a validation method for each half. Measurement validation is A (no data check), B (check by overlapping data sets) or C (check by independent measurements or methods). Representation validation is A (no check), B (double check) or C (triple check).
The guide's notation packs this into one string: "30Brel-20Arel" is measured accuracy at LOA30, checked by overlapping data sets, relative; represented accuracy at LOA20, no check, relative.
Why does a spec need both, set per discipline?
Because they fail independently.
- Right LOD, wrong LOA. A steel beam is modeled at LOD 300: correct section, correct length. The modeler aligned it to the grid instead of the cloud. It is measurable, and it is in the wrong place.
- Right LOA, wrong LOD. The point cloud meets its measured accuracy, but ductwork was modeled at LOD 200 as generic placeholders. The accuracy never reached the element the mechanical engineer needed.
Both standards also expect levels to vary by element. The USIBD guide says "There is no such thing as an 'LOA XX model,'" because concrete cannot be controlled to the same degree as steel. Both are organized by Uniformat, so one element list can carry an LOD column and two LOA columns.
How do LOA and LOD relate to ZEALOT's published accuracy?
Three of our published accuracy figures map directly onto LOA: ±5mm registered accuracy for the point cloud, and model faces within ±10mm of the registered cloud at LOD 300 and ±6mm at LOD 400. In LOA vocabulary, the first is a measured accuracy figure and the other two are represented accuracy figures. Model tolerance looser than cloud tolerance is exactly what USIBD formalizes: modeling adds error on top of measurement. Our guide to what ±5mm accuracy means walks through that chain.
Confidence level is where people slip. LOA ranges are stated at 2 sigma. The manufacturer's accuracy statement behind our ±5mm figure is given at 1 sigma, as listed on our NavVis VLX3 page. A 1-sigma figure and a 2-sigma range are on different scales, so you cannot read one straight into the other's table. Before you map any provider's number to an LOA band, ours included, get four answers in writing:
- Is it measured or represented accuracy?
- Is it absolute or relative, and relative to what control?
- At what confidence level is it stated?
- How was it validated, and where are the check results?
A good registration report answers all four. If it does not, have that conversation before the model is built.
What does a sample spec clause look like?
The levels below are illustrative, not a recommendation for your building; set them with your structural and MEP engineers, using the Suggested LOAs the USIBD guide gives for measured accuracy as a starting point.
3.2 Level of Development. Model elements shall be developed in accordance with the BIMForum LOD Specification, [2025] edition, at the levels listed in Table 3.2. Elements not listed shall not be modeled.
>
3.3 Level of Accuracy. Measured and represented accuracy shall conform to the USIBD Level of Accuracy Specification, Version [3.1], at the levels listed in Table 3.2, stated at the 95 percent confidence level. Measured accuracy is relative to the project control network.
>
3.4 Validation. Measured accuracy shall be validated by independent measurements (method C). Represented accuracy shall be validated by double check (method B). The registration and QC report shall list check locations, residuals and the confidence level used.
>
3.5 Concealed conditions. Elements not visible in the point cloud shall be modeled only where identified in Table 3.2, and shall be flagged as not field-verified.
| Element group (Uniformat category) | LOD | Measured LOA | Represented LOA |
|---|---|---|---|
| B Shell: structure, floors, roof | 300 | LOA30 | LOA30 |
| B Shell: exterior walls and openings | 300 | LOA30 | LOA20 |
| C Interiors: partitions and doors | 200 | LOA20 | LOA20 |
| D Services: mains in Mechanical Room 1 | 350 | LOA30 | LOA30 |
| C Interiors: Stair A (new fabrication) | 400 | LOA40 | LOA30 |
Every row names both halves of the LOA, and no row asks for a represented accuracy tighter than its measured accuracy, which the standard does not allow. Tighter measured levels in a fabrication zone usually mean adding tripod scanning or independent measurement there, which is why the table limits them to that zone. Our scope of work template and deliverable spec guide show where this table sits in a full scope.
What are the most common LOA and LOD mistakes?
LOD 350 everywhere. LOD 350 is defined by interfaces with adjacent or dependent elements, and BIMForum notes it "usually requires craft knowledge." In an existing building many of those interfaces sit behind finishes and insulation, and the USIBD guide notes that concealed conditions can make certain LODs and LOAs difficult or impossible to document. Ask for 350 at the tie-ins where new work connects to old, and LOD 300 elsewhere. Our LOD 200 vs 300 vs 400 comparison covers how to split levels by area.
LOA with no represented accuracy. "Deliver to LOA30" alone binds the point cloud at best and leaves everyone guessing at worst. Designers use the model, not the cloud, so write down the represented accuracy.
"An LOA40 model" or "an LOD 300 model." Both standards reject a single level for a whole model. One number invites the cheapest reading of it, or the most expensive price.
Mixing confidence levels. A tolerance, a 1-sigma spec sheet value and a 2-sigma LOA range are three different things; the USIBD guide treats a tolerance as allowing no exceedance at all. Say which one you mean.
No edition or version. BIMForum updates the LOD Specification periodically and lets a project adopt a specific version; older LOA guides still circulate. Name both.
No validation method. "No data check" and "no check" are legitimate options in the LOA framework. If the spec does not choose, that may be what you get.
Treating LOD 500 as "better than 400." It is a statement about where the geometry came from, not how developed it is, and it only means something when an accuracy is attached.
How should you use this on your next scan-to-BIM RFP?
Start from the element list, not the model. For each element group, write the LOD the design decision needs, the measured and represented LOA it can tolerate, and how each will be checked. Name each standard's edition and the confidence level, send the table with the RFP, and compare quotes against it line by line.
For a second opinion before you issue it, send us the scope. We return scan-to-BIM quotes within one business day. For wider context, see our building scanning guide or contact us.
Sources
- U.S. Institute of Building Documentation, Level of Accuracy (LOA): Version 3.1 release notes and download.
- U.S. Institute of Building Documentation, Document C120, Level of Accuracy (LOA) Specification Guide, Version 2.0 (2016): LOA ranges, confidence level, measured vs represented accuracy, validation methods, notation.
- BIMForum, Level of Development (LOD) Specification: purpose, Uniformat organization, 2025 edition.
- BIMForum, Level of Development Specification, 2024 Part I: LOD 100 to 500 definitions, LOD 500 and LOA, "no such thing as an LOD ### model."
