Most maturity assessments begin with an interview. Somebody from outside arrives, asks the project manager how information is managed on the job, writes down the answer and converts it into a score with a colour attached. The project manager is not lying. They are describing the BEP, which is a document, and the question was really about a practice.
There is a cheaper way to find out, and the person best placed to run it is already on the payroll. A document controller can measure three things over ninety days on one live project, beside the day job, without buying a maturity model: whether people can retrieve the current issued set, whether a transmittal can be reconstructed, and whether suitability codes constrain action or simply sit on the metadata looking tidy.
Aggregated returns will get published here if enough independent scorecards come back to be worth pooling. That dataset does not exist yet, and saying so seems better than inventing one. Any vendor "industry benchmark" that arrives without a denominator deserves the same scepticism.
What the three tests actually measure
ISO 19650-1:2018 clause 12.1 asks for a status on each information container so the recipient can see the permitted use — and, by extension, the uses that are not permitted. The UK National Annex to BS EN ISO 19650-2 gives a standardised list (S0–S4 and the published An codes; the 2018 and 2021 annexes disagree about S5, S6, S7 and CR). The UK BIM Framework's Guidance Part 2 is blunt about the consequence: an S3 container is not for procuring materials, agreeing costs, or constructing the works.
Nobody needs to have memorised Table NA.1 for the audit to work. What matters is whether the project has an agreed list at all, whether four people can lay hands on the current issued set, whether a transmittal can be rebuilt from the system of record, and whether a status code has ever stopped somebody doing the wrong thing with a file.
A week would be too short — it would only measure how quickly the controller can find things, which was never in doubt. A consultancy quarter is too long and too expensive for a first measurement. Ninety days sits in between: long enough that a one-off tidy-up in week one cannot fake the result, short enough that the person who already runs the register can carry it alongside everything else.
So what would weaken the claim that this project has information management? Any of the three tests scoring 0, whatever the BEP says. What would weaken the opposite claim, that the job has none at all, is all three scoring 3. Report the three separately. Averaging them into a maturity percentage invents precision the method does not have, and percentages travel further than they deserve.
Who runs it
A named document controller or information manager on the job, not an external auditor. The controller may recruit four colleagues for the retrieval test but must not coach them on the path, and must not quietly tidy the common data environment the week before day one. That last temptation is real, and giving in to it wastes the ninety days.
Pick one live project rather than a closed archive. Name the BEP, information protocol or project information standard you will test against, including the status-code list that document actually publishes. If no such list exists, write that down on day one — you have your first result before the audit has properly started.
The scorecard
Print this page. One row per trial. The controller fills it in; the project manager does not edit it.
Test A — who can find the current issued set
Name one live package the site is building from this month: a floor of architectural GAs, a coordinated MEP set for a zone, the current structural issue for a frame. Then ask four roles, independently and with no coaching, to produce "the current issued set" in twenty minutes:
- the project manager
- a site engineer or supervisor
- a subcontractor who is installing from it
- the document controller, as a control
Record, for each person: elapsed minutes, the source they used (published location, email, messaging app, plotter tray, personal drive), the revision, and the status code they thought they had.
| Role | Minutes | Source | Revision | Status | Score |
|---|---|---|---|---|---|
| Project manager | |||||
| Site engineer / supervisor | |||||
| Installing subcontractor | |||||
| Document controller (control) |
Score each retrieval:
The project score for Test A is the worst of the four, because maturity is the weakest retrieval on the job rather than the controller's. Run it twice: once announced in week two, once as a cold ask at least a week later. Keep both sheets. If the cold ask comes back worse, the announced score was a performance, and most of us have put on that performance at some point.
Test B — who can reconstruct a transmittal
Pick three transmittals sent in the last ninety days, or the last three if the project is younger. For each, reconstruct six facts from the system of record rather than from memory:
- unique transmittal number and date
- named recipients, not "the architect"
- exact containers and revisions
- purpose, or the action requested
- dated response or acknowledgement
- what has superseded any item since
| Transmittal | Number and date | Recipients | Containers and revisions | Purpose | Response | Superseded by | Minutes | Score |
|---|---|---|---|---|---|---|---|---|
| 1 | ||||||||
| 2 | ||||||||
| 3 |
Score each transmittal:
Again the project score is the worst of the three. Where a transmittal cannot be reconstructed, stop and write down the fact that is missing. The instinct is to build a better process on the spot so the next one scores higher; resist it until the audit finishes.
Test C — whether suitability codes work as codes
Sample forty containers that left work-in-progress in the last ninety days, or all of them if there are fewer. Use the project's own agreed list, not a list you happen to prefer. For each container, answer four yes/no questions:
- Does it carry a status from that agreed list?
- Does the status match the CDE state (S0 is not in Published; a published
Anis not still editable by the originator without a new revision)? - Can the file be downloaded, plotted, or issued regardless of the code?
- Has anyone on this job recorded an action that the code required — a hold, a rejection of use, a request for a new revision?
| Sample | Status on file | Agreed-list match? | State matches code? | Download ignores code? | Recorded action? |
|---|---|---|---|---|---|
| 1–40 | Y/N | Y/N | Y/N | Y/N |
Then have one conversation and write it down: find whoever plots or issues to site and ask whether they look at the code before they print. The answer takes a minute and is often the most informative part of the whole test.
Score the project, not each file:
A code earns the name when it changes a subsequent action. It is decoration when the BEP required the field, the download button ignores it, and the unofficial set in the trailer does not carry it at all.
The ninety days
Days 1–7 — Scope. Name the project, the protocol, the status-code list, the four roles, the package, and the three transmittals. Export or photograph the current register page as a baseline. Do not tidy first. If a colleague refuses to take part in the retrieval test, record the refusal; that is a 0 for the role.
Days 8–28 — Issued-set retrievals. Run Test A announced, then cold. Note the unofficial sets as you go: trailer folders, messaging threads, the plotter tray, a QS working file. Those sources are part of the measurement rather than local colour.
Days 29–56 — Transmittal reconstruction. Run Test B. If the only complete trail turns out to be somebody's personal mailbox, write that down. Migrating the mail mid-test so the second transmittal looks cleaner than the first defeats the point.
Days 57–77 — Suitability. Run Test C. Count how many of the forty would be legal to construct from if the National Annex rule were applied as written. That count is an observation about this sample, not an industry rate.
Days 78–90 — One page. Three scores, three holes, and one change you will make without buying software. Things that stay inside the method: publish one issued-set location and tell the four roles where it is; stop plotting anything that is not in it; refuse to send a transmittal that cannot later produce facts 1–4. Software that cannot answer those facts will not lift a 0 to a 3 by being purchased.
Returning a result
If you run the audit, send the three project scores — not the drawings — to hello@aecovault.ai. This shape makes returns comparable:
Project type (building / civils / other):
Region (country only):
CDE / store (product name or “shared drive”):
Test A (worst of four): 0–3
Test B (worst of three): 0–3
Test C (project): 0–3
Status-code list used (2018 NA / 2021 NA / project-specific / none):
One hole you could not close:
Please do not send project names, employer names, or files. There is no subscriber list behind this site; a return is a one-off email. A league table will not be built out of a handful of replies, and no benchmark will be invented to fill the wait. If enough independent returns arrive to treat as a dataset, the count, the filtering and the places where the method still breaks will be published alongside it.
Where the scores stop
Three integers cannot show that a named platform, a mandate or a model will raise them. They do not certify ISO 19650, and any consultant who says otherwise is selling. Model federation, clash closure and handover completeness are all later tests that this one deliberately leaves alone. And no arithmetic here produces "thirty per cent of lifecycle data" or any similar figure that turned up without a denominator.
The information manager — or whoever already owns the protocol on this job — should write one change on day 90 that does not wait for a purchase: one issued-set location the four roles can name, a stop on plotting anything outside it, and a refusal to send a transmittal that cannot later produce facts 1–4. Ninety days of ordinary time is the price; a consultant and a new store are not required, and neither will lift a 0 while the unofficial set in the trailer is still the one people print from.
Read next: The AECO information lifecycle in one diagram, Monthly Register — October 2024, The most expensive thing on your project is the document nobody can find.