Look around before you draw
Most buildings repeat themselves. The corridor on level 3 has the same beams as the corridor on level 4; the heating branches in one ward run like the branches in the next. When one run clashes, its neighbours have usually already solved the same problem, and the way they solved it is the project's convention: the height under the beams, the side of the corridor, the distance from the wall.
A fix that ignores that convention may clear the clash and still be wrong. It drops 300 mm where every other heating pipe drops 250; it hugs the wrong wall; it lands in front of a valve someone will need to reach. ClashFree reads the neighbours so the fix fits in, not just fits.
What counts as a precedent
A precedent is a nearby run of the same system classification, size and layout as the run being fixed, with no clash in the current report. For each one, ClashFree measures how it passes the same obstacle:
- its height relative to the beam's bottom or the obstacle's face;
- the walls it crosses, and which it avoids;
- its distance from the nearest wall.
Those measurements go into the context the AI reads, as evidence it can cite, and into the router as targets: a preferred height or line for the reroute to aim at.
A bonus, not a pass
An option that sits where the precedents pass, within 50 mm of their height or line, earns a bonus of −800 points. To put that in scale: an added fitting costs 300, so matching the neighbours is worth a little over two and a half elbows. It is enough to break a tie between reasonable fixes, and not enough to make an unreasonable one win.
A precedent is never a rule. It cannot make an option feasible, it cannot lift a hard limit, and a precedent that itself breaks one is shown with that limit, so the AI knows not to follow it. (The neighbours may well hang below the new ceiling line because they were modelled before it was set.) When the AI follows a precedent, it says so: "matched the three heating runs at 2650 under the beam".
Reads same-system, same-size runs near the clash and on twin rooms of other levels, measures how they pass the obstacle, gives a matching option a −800 bonus within 50 mm, and passes the measurements to the AI and the router as evidence. Nothing about it bypasses the validator or Revit's verification.
Twin rooms on other levels
Sometimes there are no neighbours on the same floor: the clash is the only heating branch in that room. But the room above it is often the same room, built the same way, with the same branch already coordinated. ClashFree reads that too. It finds the room the clash is in, looks for the same room on other levels, and takes their runs of the same system as precedents.
Clash points do not always fall inside a room; a pipe in a ceiling void over a corridor wall belongs to nobody. In that case ClashFree probes along and beside the run to find the room it serves. And when the other level has no matching room, it falls back to runs stacked directly above or below on other levels.
The corridor: who else lives here
Neighbours are not only precedents; some are obstacles with needs. ClashFree's context includes the corridor: the runs beside the one being fixed, and the access zones of valves, dampers, terminals and equipment along it. An access zone is the space a person needs to reach and service the device. A fix that routes through one is not an obvious clash in a clash test, which is exactly why it gets missed. The router prices crossing an access zone, and the AI is told to keep the zones clear.
When the neighbour is a wall
One finding comes before all the others. When the run being fixed lies inside a wall along most of its length, it was almost certainly dragged there by accident in modelling. The clash is a symptom; the fix is to drag the run back out. ClashFree shows that finding first, before any option, and the drag that undoes it is usually the best scored: the run moves with its end fittings and the neighbouring segments stretch, the way the duct next door was placed.
When B is a copy of a neighbour
The same habit of looking around settles one more question: is B real? A duct of your trade that appears in another trade's model is often a copy someone linked in for reference. ClashFree treats B as a probable duplicate only when it finds a twin in your host model: the same category, system and size, parallel and overlapping. Then the clash is flagged as a duplicate and no fix is attempted. Without a twin, B is a real obstacle, and the context says the check was made.
A walk-through
Here is how the pieces combine on one clash. A DN80 heating branch on level 2 runs through a downstand beam in a ward corridor. The option list holds the usual candidates: drops of different depths with 45° and 90° fittings, a sideways offset towards the corridor wall, a reroute. The precedents add two facts. Three DN80 heating runs a few metres along pass the same beam 250 mm below this branch's level. And the same ward on level 3, its twin room, has its branch passing that beam the same way.
The drop of 250 mm matches the corridor neighbours and earns the bonus. The sideways offset would sit in front of an isolation valve's access zone, so the router's price for it goes up and the AI's context says why. The option at the neighbours' level comes out on top, Revit verifies it, and the explanation reads like a coordinator's note:
Dropped 250 mm with 45° fittings to the level of the three DN80 heating runs under this beam (precedent); sideways would block the access zone of valve V-2.07; ceiling line 2700 (project) kept.
Nothing in that note is invented. Each clause points at a measured precedent, a corridor entry or a profile value, which is what makes the explanation checkable. (An illustrative example; the ids and heights are made up for this article.)
Why evidence beats rules here
It would be easy to write "heating pipes pass beams at 2650" into a project rule. It would also be wrong on the floor where the beams are deeper, in the plant room, and on the day the ceiling is lowered. Precedents adapt because they are measured from the model each time, and they stay honest because they are only ever a bonus. When a precedent and a hard limit disagree, the limit wins and the AI is told why.
See how the bonus sits among the other terms in Five valid fixes, one right answer, or let ClashFree read the neighbours in your own model in the free beta.