The Forensic BIM score™ and its grades
Summary
Every report carries two headline numbers. The Overall quality score looks at the whole file: can it be opened, is it organised, is the geometry usable, is the data there and is it coherent. The Forensic BIM score™ looks only at the data the objects carry: how complete, rich, reliable and well-defined the information is. It runs from 0 to 100, is built from four factors with a fixed number of points each, and gets a grade from A to E.
The two are linked. A model can be tidy and valid and still carry almost no usable information, so the Forensic BIM score™ puts a ceiling on the Overall quality score.
Where you see it in the report
| Place | What it shows |
|---|---|
| Model header, at the top of every page | A pill reading Forensic BIM score™ with the score and the grade, next to the Quality pill. Both take you to the Overview. |
| Overview, score strip | The score as a ring with its grade and status, captioned How complete, rich, reliable and well-defined the data is. It links to the Scores page. |
| Scores page | The score and grade next to Overall quality, and the card What makes up the Forensic BIM score™ with the points earned per factor. |
| Benchmark page | The percentile of the score among comparable models, the scatter Overall vs. Forensic BIM score™, and the table Forensic BIM score™ factors. |
| Capabilities page | The AI ready capability requires Forensic BIM score™ ≥ 60, among other conditions. |
A tour of all the pages is in How to read your report.
The four factors
The card What makes up the Forensic BIM score™ shows each factor as a bar of points earned out of points available, with the figures behind it underneath (for example materials 85% · quantities 60% · classification 92%). Hover a factor to read its description. The points add up to the score.
| Factor | Points | Description in the report | Measured from |
|---|---|---|---|
| Completeness | 30 | Average of material, quantity and classification coverage. | The checks Material coverage, Quantity coverage and Classification coverage. A coverage that could not be measured is left out of the average. |
| Richness | 25 | Effective facts per element on a log scale (20 facts = full marks). | The check Effective facts per element: property values that say something, after empty and system values are discounted. |
| Reliability | 25 | Share of informative values and alignment continuity, reduced for filler values, duplicate GlobalIds and negative quantities. | The checks Informative property values, Disguised missing values (‘-’, ‘n/a’, ‘tbd’), Duplicate GlobalIds and Negative areas or volumes, plus the alignments of road and rail models. |
| Semantics | 20 | Properties with defined meaning (standard or URI), classification URIs and type objects. | The checks Properties with a defined meaning, Classification linked to URIs and Elements linked to a type. |
How each factor fills up
Completeness is the plain average of the coverages. A model where every element has a material and a classification but none has quantities earns two thirds of the 30 points.
Richness uses a logarithmic scale, so the first facts on an element count most. About 5 effective facts per element earn roughly 15 of the 25 points, about 10 earn roughly 20, and 20 or more earn all 25.
Reliability starts from the share of informative property values (values that are not empty, filler or placeholder). That share is reduced by the share of disguised missing values such as ‘-’, ‘n/a’ or ‘tbd’. If the file has any duplicate GlobalIds, the factor is cut by 30%; if it has any negative areas or volumes, by a further 10%. For models with alignments, alignment continuity adds to the factor: a fifth of it when every alignment is perfect, less when they have notes or warnings. A model with perfect alignments but no properties still earns most of this factor. See The alignment check.
Semantics weighs three things: properties with a defined meaning (45%), classification references with a URI (25%) and elements linked to a type object (30%). A file without any properties can only earn the type part.
When the data is unusable
When a file has elements but all of its properties are undefined, it has no materials, no quantities, no facilities (no site, building, storey or other facility) and no alignments, the report tags it useless data. Every score in the report, the Forensic BIM score™ included, is then 0, and the file gets no Benchmark page.
Grades
The grade is a short label for the score. It is shown in the header pill, on the Overview and on the Scores page (grade B after the total on the factors card).
| Grade | Forensic BIM score™ |
|---|---|
| A | 80 or more |
| B | 65 to 79 |
| C | 50 to 64 |
| D | 35 to 49 |
| E | below 35 |
Next to the grade the report shows the same status label it uses for every score: Good (or Pretty good below 100) from 80, Needs attention from 55 to 79, Weak from 30 to 54 and Critical below 30. The two use different bands, so a score of 60 is grade C and Needs attention.
How it differs from Overall quality
Overall quality
A weighted average of the five dimensions: Validity, Structure, Geometry, Information and Consistency. It answers whether others can reuse the model with little rework. Information counts double; Geometry and Consistency count three quarters.
Forensic BIM score™
Only the information on the objects, in four factors. It ignores whether the file opens cleanly, how the spatial structure is organised or how heavy the geometry is. It asks whether the data is there, says something, can be trusted and has a defined meaning.
A file can do well on one and badly on the other. A geometry-only export from a design tool can pass every validity and geometry check and still score near 0 on the Forensic BIM score™.
The limit: “Limited by its data”
Since scoring version 2 the Overall quality score can be at most 35 + 0.7 × the Forensic BIM score™. A tidy model without usable data cannot score high on quality.
| Forensic BIM score™ | Highest possible Overall quality |
|---|---|
| 0 | 35 |
| 20 | 49 |
| 40 | 63 |
| 60 | 77 |
| 80 | 91 |
| 93 or more | no limit |
When the limit applies, the Overview score strip and the Scores page say Limited by its data, followed by the score the five dimensions would have given and the score after the limit, for example Limited by its data: 74 → 63. On the Benchmark page, the scatter Overall vs. Forensic BIM score™ draws the limit as a dashed line; models held back by it are shown as purple dots.
The points per action on the Potential page are calculated before the limit. If your model is limited by its data, fixing validity or geometry issues will not lift the Overall quality score while the limit applies; raising the Forensic BIM score™ is what moves the limit up. The rules and their history are in Scoring versions and what changed.
A quality score, not a valuation
The score tells you how usable the information in the file is. It does not put a price on it. The tooltip on the header pill calls it the value of the data itself; read that as the usefulness of the data, separate from anything the model is worth in money. The financial figures are on the Valuation and Savings pages, and the valuation has its own method.
How to raise it
Look at the factor with the most points missing on the Scores page, then at the checks it is built from on the Information and Consistency pages. Usually that means asking whoever delivers the model to:
- assign materials, quantities and a classification to every element;
- fill in the properties you need and remove empty ones and filler values;
- use standard property sets or properties linked to a URI, so their meaning is defined (see Property sets and Classification in IFC);
- link elements to type objects;
- remove duplicate GlobalIds and negative quantities (see GlobalIds).
Frequently asked questions
My file passes every validity check. Why is its grade E?
Validity says the file can be read. The Forensic BIM score™ asks what the objects tell you once it is read. A file with shapes, names and little else has a low score whatever its validity.
Is the score the same for buildings and for bridges, roads or tunnels?
The factors and points are the same for every file. Road and rail models can earn part of Reliability from their alignments. On the Benchmark page the score is compared with models of the same kind; see Benchmark: peer groups and percentiles.
Does a higher grade mean the model is worth more?
Not directly. The grade describes the quality of the data. The worth of the data is a separate question, answered on the Valuation page.