Where this came from
Clash detection was not invented in construction. It came from aerospace: the Boeing 777, in the mid-nineties, was the first commercial aircraft assembled entirely in the computer before it physically existed, with no physical mock-up. That was called digital pre-assembly, and construction imported it about fifteen years later.
The craft of coordinating is far older. Before the model it was done with transparent overlays on a light table: ductwork on one sheet, piping on another, electrical on a third. The clash was always visible. What changed is that the machine finds it now, instead of the tired eye of whoever is checking at eleven at night.
Coordination is not about finding clashes. It is about stopping decisions from being made too late.
The five types of clash
Sending every result to the same list is the most common opening mistake. A 4D clash is not the modeller's to fix, and an internal model clash should never reach the coordination meeting at all.
| Type | What it is | Example | Who resolves it |
|---|---|---|---|
| Hard | Two elements occupy the same physical space | A beam run through by a main duct | The author of whichever element can actually move, per the hierarchy |
| Soft or clearance |
They do not touch, but a space that must stay clear is invaded | Pipe four inches from a panel that requires code working clearance in front of it | The author, with input from the affected discipline's specialist |
| 4D or time |
They never coexist in the model, but they do in the same week on site | Two crews and a crane in the same work front on the same Tuesday | The scheduler, not the model coordinator |
| Model or internal |
The model clashes with itself: duplicated or badly modelled elements | Two identical walls stacked by a copy-paste | The modeller. It should never reach the meeting |
| False positive | Noise from tolerance, insulation or representation | Insulation «clashing» with the structure that supports it | Nobody. You adjust the rule, not the model |
Tolerance: separating the problem from the noise
Running detection at zero tolerance produces thousands of results, most of them irrelevant. The consequence is not extra work: it is that the team stops opening the reports by week two, and from then on coordination exists on paper and not on site.
Tolerance is set per system pairing, not globally. A two-millimetre overlap between conduits is not the same thing as two millimetres between a beam and a large duct.
How to prioritise
Working the list in the order it came out is working it in random order. The three criteria that actually rank the work are:
- When it gets built. What goes up in six weeks outranks what goes up in eight months, even when the second one is worse.
- What it costs to fix late. A clash that means cutting concrete is not worth the same as one solved by moving a cable tray.
- How many disciplines it touches. Three-party clashes take longer to decide, so they have to be opened earlier.
The hierarchy: whatever cannot move decides
Solving a clash by moving whatever is easy to move in the model — instead of what is correct to move on site — is the most common way to fix a drawing and create a problem. The order comes from a single question: which element has the least freedom?
What happens when one is found
This is what separates a coordinator from somebody who knows the software. There is a ladder, and skipping it is what produces models that no longer match the design.
A BIM coordinator has no authority. They have visibility. Confusing the two is the mistake that ruins a new coordinator.
The life cycle of a clash
A clash is not a finding. It is a task with states. Marking it resolved because somebody promised to fix it — without verifying it against the updated model — is the quietest leak in the whole process.
| State | What it means | Who moves it |
|---|---|---|
| New | Detected and classified, no owner yet | Coordinator |
| Assigned | Has an owner and a due date | Coordinator |
| In review | The author has proposed a solution | Element author |
| Approved | The fix has sign-off from whoever decides | Lead or designer |
| Resolved | The model now carries the change | Element author |
| Closed | Verified against the updated model, with evidence | Coordinator |
How a clash travels between different software
Emailing a screenshot does not work: whoever receives it does not know which element it is or where the camera was. That is what the BCF format has existed for since 2010 — an open buildingSMART exchange that carries the issue with its viewpoint, its identified elements, the comment, the assignee and the date, and opens in any tool along the chain.
It is the piece that turns «there's a clash somewhere over there» into a task somebody can pick up. Without something like it, «a clash has an owner and a date» is a wish.
Frequently asked questions
What is the difference between a hard clash and a soft clash?
A hard clash is geometry against geometry: two elements occupy the same space. A soft clash never touches, but it invades a clearance that has to stay free — the code-required working space in front of an electrical panel, the access needed to service a unit, the swing of a valve. Geometry alone does not find soft clashes: they are found by whoever knows the rule. That is why rule-based model checking and geometric clash detection are two different disciplines, and a good coordinator uses both.
Who resolves a BIM clash?
Never the coordinator. The coordinator detects, classifies, assigns and verifies closure, but does not move somebody else's elements. It is resolved by the author of whichever element can move; if the fix touches another discipline it escalates to the lead, and if it reveals a design conflict it becomes a formal RFI.
What tolerance should clash detection run at?
It depends on the system pairing, which is why you define a clash matrix: which discipline is tested against which, and with what clearance for each pair. A single global zero tolerance produces thousands of false positives and the team stops reading the reports, which is the worst possible outcome.
What is a 4D or time clash?
Two activities that never touch in the model but land in the same space in the same week: two crews in one front, a crane blocking access, material laydown where another trade needs to work. It is invisible in geometry; you see it by crossing the model with the schedule.
How many clashes are «too many»?
The right question is not how many there are, but how many have an owner and a date. Two thousand unassigned clashes are worth less than thirty well-assigned ones. Measuring the work by clash count is measuring the noise.