Interview grid for a Revit BIM modeller: 15 questions that reveal the real level

Hiring a BIM modeller without a BIM Manager in the firm is a blind exercise: the candidate talks about Revit, you nod, and the real level surfaces three weeks after the hire, when the model starts weighing 800 MB and the schedules stop working. The fifteen questions below do not require you to be an expert to ask them. Each comes with what a solid answer contains, and the typical warning sign.

Model organisation (questions 1 to 4)

• 1. "How do you decide to split a project into several Revit files?" Good answer: by building, by discipline or by zone, with a SITE file for georeferencing, and a file-size threshold in mind. Warning: "we put everything in one file".
• 2. "What is a workset and how do you name them?" Good answer: a stable convention, one workset per package or level, and the habit of checking which workset an element belongs to before creating it. Warning: confusion between worksets and phases.
• 3. "What do you do when a file becomes slow?" Good answer: file audit, purge, hunting heavy families, checking open views and links, before any hardware upgrade. Warning: "I ask for a more powerful PC".
• 4. "How do you handle Revit links and IFC links from other disciplines?" Good answer: links positioned by shared coordinates, never moved by hand, and a coordination meeting to validate updates. Warning: the candidate copies elements from links into their model.

Families and modelling (questions 5 to 8)

• 5. "When do you create a family and when do you use an in-place element?" Good answer: in-place stays exceptional (unique element, context-dependent geometry); anything that repeats becomes a loadable family. Warning: regular use of in-place "because it's faster".
• 6. "Explain the difference between a family parameter, a shared parameter and a project parameter." Good answer: clear, with the typical use of each (schedules, tags, exchanges). Warning: the candidate doesn't know why a parameter doesn't show in a schedule.
• 7. "How do you name your families and types?" Good answer: an existing convention (the firm's or a public one), applied systematically. Warning: "it depends on the project".
• 8. "Tell me about a complicated family you built." Good answer: constraints, formulas, tests in several configurations, and what broke. Warning: no concrete story.

Data, views and deliverables (questions 9 to 12)

• 9. "How do you build a door schedule the engineering office can use?" Good answer: fields defined upfront, shared parameters, sorting and filters, a check that every door is filled in. Warning: the schedule seen as a mere export.
• 10. "What are view templates for and why use them?" Good answer: graphic consistency, time saved, centralised control. Warning: the candidate sets every view by hand.
• 11. "How do you prepare a clean IFC export?" Good answer: category mapping, MVD choice, check in an independent viewer, shared coordinates. Warning: "I export with the default settings".
• 12. "How do you check that a model is ready to be issued?" Good answer: a checklist (warnings, unassigned elements, views on sheets, data filled in). Warning: no control routine.

Method and collaboration (questions 13 to 15)

• 13. "What does a BEP contain and what do you look for first?" Good answer: naming conventions, model breakdown, information levels per stage, exchange schedule. Warning: the candidate has never opened a BEP.
• 14. "How do you react when a colleague modifies an element in your area?" Good answer: communication, worksets, version history; no conflict. Warning: the candidate edits without warning or locks everything.
• 15. "What have you learnt about Revit in the last six months?" Good answer: a feature, a plug-in, a script, a convention; anything, as long as it is precise. Warning: nothing.

A good interview question does not test whether the candidate knows a button. It tests whether they know why they use it, and what happens when it goes wrong.

How to score without fooling yourself

The classic trap is to leave the interview with an overall impression ("they seemed comfortable") and convert it into a grade afterwards. To avoid that, score each question immediately, on a four-level described grid: 0 doesn't know, 1 has heard of it, 2 can do it, 3 masters it and can explain it. Fifteen ratings from 0 to 3, weighted by what matters for the position, give a score comparable from one candidate to the next. Two candidates close on the score are then separated on the skill profile, not on gut feeling.