Service · BIM content · Working remotely
Parametric families for manufacturers and design offices, built to your product or your standard, kept light, and tested against the conditions that break content rather than the ones that flatter it.
01, Why libraries go wrong
A family that fails rarely fails at being drawn. It fails at being used, in someone else's template, at someone else's scale, on a deadline. These are the three failures worth naming.
Two type changes in the family editor is not a test. The parameters that break are the combinations nobody tried, the shortest length with the deepest offset, the one that inverts a constraint and takes the geometry with it.
Imported mesh geometry, full manufacturing detail, every fillet the CAD file had. Accurate, unusable, thirty of them in a model and the view regeneration is where the project day goes.
Looks right in a view, empty in a schedule. Shared parameters missing, classification absent, naming that doesn't match the office standard, so the model can be drawn from but not counted from.
02, Built work
Two families for one company, designed around their needs and around their own connection details, for two different areas of the project.
One curtain-wall system, two panel types. The frame, the brackets and the anchor detail are identical between them, what changes is the infill and the head condition, because the two sit in different parts of the elevation and answer to different requirements.
That is the case for building families rather than downloading them: the shared half is shared by construction, not by coincidence, so a change to the frame is one change and not two.
Drawn from the delivered families themselves, the outlines above are the real geometry, reduced to boundary and crease edges. Nothing here is a stand-in.
03, What moves the number
Drawing the shape is the short part. Four things decide how long the rest takes, and you can assess all four from your own files.
A fixed component is one thing. A family that flexes across a size range, swaps materials and hides parts by type is a matrix of combinations, and every one of them is a state that has to be checked, not assumed.
A dimensioned drawing is a specification. A CAD export full of imported mesh, or a PDF of a brochure, is a reconstruction job first, and reconstruction is where the hours actually go.
Content that only has to look right is quick. Content that has to schedule, price, classify and survive a client's own parameter file is doing a second job, and that job has its own scope.
A one-off is finished when it's handed over. A catalogue that tracks a real product line has versions, a Revit release range and a changelog, a different commitment, and better decided at the start than discovered later.
No estimate is given before the first two stages. A number produced before the problem is stated in full is a guess wearing a currency symbol.
04, Process
Four stages. The first two are where most of the risk gets removed.
We watch the real workflow, not the documented one. What's actually done by hand, by whom, and how often it goes wrong.
The problem gets stated in full, with a success measure that can be tested, before any tool is proposed.
Built against your standards and your users. Every added layer of complexity has to earn its place.
Systems are built to change shape as your work grows, so the tool doesn't become the next thing to work around.
05, Questions
Whatever actually exists, dimensioned drawings, a CAD file, a spec sheet, or a family that nearly works. The first stage establishes what is specification and what is a guess, because those are two very different jobs and they shouldn't be quoted as one.
By deciding what the geometry is for. Detail that only reads at 1:5 is modelled as detail components at that scale, not carried in 3D through every view. Imported mesh is rebuilt as native geometry rather than wrapped. The size target is set in stage two and tested against, not hoped for. The same discipline across a whole library is covered in library optimization.
That's the usual case. Naming, shared parameters, subcategories and line weights come from your standard, and where the standard is ambiguous, the ambiguity gets raised rather than silently resolved in a way you find out about later.
The ones your team and your clients run. Families are delivered in the oldest version in scope so they open forward, the range is agreed in the first stage, since it affects which features can be used at all.
Usually, yes, and it's usually cheaper. An audit establishes what is worth keeping, what is duplicated and what is genuinely broken. Rebuilding content that already works is billable and pointless.
It's tested against the combinations that break content, not the ones that flatter it, and the test cases are written before the family is. A family that flexed once in the editor hasn't proven it works, it's proven it worked once. The tests themselves are set out in the parametric family piece.
06, Start
Bring the file and the model it has to live in, not a specification of what you think it needs. If your library needs an audit rather than new content, we'd rather say so first.
We are not a fit if you want geometry traced from a brochure with no dimensions, or a family count quoted before anyone has looked at the source.