BIM skills self-assessment: how to make it honest

"On a scale of 1 to 10, what's your level in Revit?" Asked like that, the question produces just about anything: the cautious modeller says 6, the self-assured coordinator says 9, and neither number tells you what the person can actually do. Self-assessment is not a bad idea, though. It is even the cheapest way to map the skills of an entire team without tying up a manager for hours. Provided it is built to be honest.

Why a self-assessment goes off the rails

Three biases show up every time. The first is the Dunning-Kruger effect: the less you master a domain, the less able you are to measure what you lack. A modeller who has never managed complex parametric families does not know what they don't know. The second is the stakes bias: if the team member believes the rating will weigh on their annual review, they round up. The third is the opposite, surface modesty, common among juniors or recent hires who don't want to seem arrogant.

• Abstract scale ("out of 10", "beginner / intermediate / expert") with no definition of what each level means.
• Question about an entire tool ("Revit") rather than a precise task ("create a family with shared parameters").
• Result visible to the manager with no discussion planned, so perceived as a conduct grade.
• No cross-check: the team member's rating is the only source.

Rule 1: describe each level by an observable behaviour

The most effective fix is also the simplest. Instead of a numeric scale, each level is described by what the person can do, in a concrete situation. For clash detection, for example: "0: I have never run a clash detection", "1: I can run a detection in Navisworks but I can't sort the results", "2: I produce a usable clash report on my projects", "3: I define the detection rules and lead the coordination meeting". Faced with these four sentences, the team member no longer picks a number, they recognise themselves in a description. The stakes bias shrinks, because it becomes hard to tick "I lead the coordination meeting" when you never have.

Rule 2: break it down by task, not by software

"Revit level" means nothing: someone can be excellent at architectural modelling and lost as soon as they need to georeference a model or export a clean IFC. A useful grid breaks the skill into units that can actually be observed on a project. It takes longer to fill in, but it is the only way to get a profile that says where to train the person, not just whether they are "good" or "average".

A self-assessment is not a grade. It is a map: it is there to know where to go, not to rank people.

Rule 3: say what the results are for

Framing changes everything. If the team understands that the self-assessment is used to choose the year's training, to allocate roles on upcoming projects and to identify who can coach whom, they have no reason to inflate their answers. If they suspect a link with pay rises, they have every reason to. That framing must be written in black and white in the invitation, before the first question, and honoured afterwards.

Rule 4: cross-check with a second opinion

A self-assessment on its own is still a declaration. Comparing it with the BIM Manager's or project lead's view on the same skills, using the same grid, immediately reveals the interesting gaps: the points where the person underrates themselves (often skills they exercise without naming them) and those where they overrate themselves. These are not faults, they are the topics of the next conversation. In a ten-person firm, this comparison takes an hour and is worth more than the score itself.

Rule 5: repeat at a fixed interval, on the same grid

The value of a self-assessment shows on the second pass. Six or twelve months later, on exactly the same grid, the evolution of each skill becomes readable, and the question "did the training help?" finally has an answer. Changing the grid between two passes destroys that comparison; it is the most common mistake of firms that cobble together a spreadsheet and improve it every year.

• Levels described by a behaviour, not a number.
• Skills broken down by observable task, not by software.
• Purpose stated in writing before the first question.
• Manager's second opinion on the same grid, then a discussion of the gaps.
• Same grid, at a fixed interval, to read progress.