Design to Fabrication BIM: From a Coordinated Model to Shop Drawings

The model everybody congratulates, then quietly abandons

Coordination signs off. The clash report is at zero on the critical zones, the design team has resolved the last of the services in the plant room, and somebody circulates a screenshot of the federated model with the caption “fully coordinated”. Four weeks later the ductwork contractor is on site with drawings that do not look much like that model, the sprinkler subcontractor has moved a main by 400 mm without telling anyone, and the steel fabricator has been working from a set of PDFs since the day they were appointed.

The model did not fail. It simply reached the end of what it was built to do, and nobody had planned what happened next.

This is the gap between design coordination and fabrication. It is where a large share of the value that BIM promises either gets realised or gets thrown away, and it is the part of the process least likely to be written down anywhere. This guide covers what actually has to happen: who models what, what gets handed over and in what form, which tools take the model past the point Revit’s design families stop being useful, how the second coordination round works, and how shop drawings, approvals and procurement hang off the result.

We are not re-covering the LOD ladder here. If you need the definitions, our guide to LOD levels in BIM sets out what each tier means. This post starts where that one stops, at the moment the coordinated LOD 350 model exists and someone has to build something from it.

Why a coordinated design model is not a fabrication model

The design model answers a different question than the fabrication model. Design asks whether the building works. Fabrication asks whether this specific piece can be made, delivered, lifted and fixed by these people on that date. Six differences do most of the damage when teams assume the first model can serve the second.

Geometry is representative, not real. A design duct is a rectangular extrusion with a nominal size. A fabrication duct is a set of sheet metal pieces with a gauge, a seam type, a joint type, stiffeners, a manufacturing length driven by the sheet size, and a connector at each end that has to physically mate with the next piece.

Connections do not exist. Design steel has beams meeting columns. Fabrication steel has bolts, plates, welds, holes, copes and a connection design that may change member sizes.

Nothing is split for delivery. A 40 m run of pipe is one element in the design model. In fabrication it is nine spools, each sized to fit a lorry, a lift shaft and a pair of hands.

Support and access are missing. Hangers, brackets, seismic restraint, valve access, filter withdrawal space and maintenance clearance are usually absent or generic in design models, and they are exactly what causes the late clashes.

Tolerance is unstated. Design models are drawn to nominal dimensions. Concrete does not arrive nominal. A fabrication model that assumes it will produces steel that does not fit.

The data is wrong for the job. Fabrication needs part numbers, material grades, weights, finishes, batch and heat numbers, and an assembly hierarchy. Design models carry specification-level data instead.

Handing a design model to a trade contractor and expecting shop drawings out of it is like handing an architect’s sketch to a joiner. It is useful. It is not the thing.

Who models what, and when the model of record changes hands

The single most useful decision on this whole workflow is made early and costs nothing: agree, per system, who owns the fabrication model and at what point the design model stops being the source of truth.

System	Design model author	Fabrication model author	Typical handover point
Structural steel	Structural engineer (Revit, analytical plus physical)	Steel fabricator (Tekla Structures, Advance Steel)	Connection design award, usually at tender
In-situ concrete and rebar	Structural engineer	Rebar detailer or contractor (Revit rebar, Tekla)	Following approval of the general arrangement
Ductwork	Mechanical designer (Revit MEP)	Mechanical trade contractor (Revit fabrication parts, CADmep, SysQue)	After design coordination sign-off per zone
Pipework	Mechanical designer	Mechanical trade contractor	Same, usually zone by zone
Electrical containment	Electrical designer	Electrical trade contractor	Often later, containment follows the others
Sprinklers	Fire engineer, performance spec only	Sprinkler contractor (design and build)	At appointment, there is rarely a design model
Facade	Architect and facade engineer	Facade manufacturer (Inventor, SolidWorks, Rhino)	Following the facade package award

Two patterns cause most of the trouble. The first is a handover point that nobody names, so the design team keeps modelling while the trade contractor also models, and two divergent versions of the same services exist for a month. The second is a design model that keeps being issued after handover, so the trade contractor rebases their work on a moving target.

Write it into the BIM execution plan as a table like the one above. Add a column for the date and a column for the named individual. The abstraction “the contractor takes over” is not enough to run a project on.

The design transfer package

When the handover point arrives, the trade contractor needs more than a model file. A proper transfer package has six parts.

1. The model itself, in a format they can use. Native Revit if they work in Revit, IFC if they do not, and a DWG or SAT background if their tool is an AutoCAD derivative. Say which one is contractual.
2. The coordinate reference. The shared coordinate system, the survey point, and the setting-out reference. Get this wrong and every downstream model is offset. Confirm it in writing rather than assuming the model carries it.
3. The tolerance and assumption statement. What was assumed about slab levels, structural deflection, fireproofing thickness, finished floor build-ups and ceiling zones. These assumptions are invisible in geometry and they drive fabrication decisions.
4. The coordination record. The zone sign-off status, the outstanding issues, and the BCF issue log as it stands. A trade contractor inheriting a zone needs to know which clashes were resolved and which were parked.
5. An explicit exclusions list. What is deliberately not modelled. Hangers, small-bore pipe below a stated diameter, cable of a given size, insulation, access panels. Silence here reads as “it does not exist”.
6. The information requirements. Which parameters the fabrication model must carry back, in what format, and against which classification system. Specify this before they model, not after.

Issue the package through the common data environment with a suitability code, not by email. It is a contractual transfer, and it will be looked at again in a dispute.

The tooling: where Revit stops and trade software starts

Revit models design intent well and fabrication poorly, and the industry has settled into a set of workarounds and dedicated tools.

Tool	What it is for	Strength	Limitation
Revit fabrication parts	Duct, pipe and containment modelled from real manufacturer content inside Revit	Stays in one environment, links to design model	Configuration setup is heavy, editing is unlike normal Revit
Autodesk Fabrication CADmep	The long-established MEP detailing tool, ITM content library	Mature content, direct to CNC and plasma cutters	Separate environment, weaker coordination tools
Trimble SysQue	Real manufacturer parts modelled natively inside Revit	Revit-native workflow, no round trip	Licensed per seat, content library dependent
Tekla Structures	Steel and concrete fabrication detailing	Connection modelling, assembly and part marks, NC files	Not a design tool, steel-focused
Advance Steel	Steel detailing in the Autodesk stack	Revit interoperability	Smaller ecosystem than Tekla
Revit rebar and bar bending	Reinforcement detailing and schedules	Works with the concrete model directly	Heavy models, performance drops fast
Inventor, SolidWorks, Rhino	Facade and bespoke component manufacture	True manufacturing geometry	Far from the BIM toolchain, needs export discipline

The choice is rarely yours as the coordinator. The trade contractor already owns a toolchain and a content library, and asking them to change it mid-project buys nothing. Your job is to make sure the exchange between their tool and the federated model works, which usually means agreeing an IFC or NWC export schedule and testing it once before it matters.

One practical point on Revit fabrication parts: they are not ordinary families and they do not behave like them. Sizes come from a service configuration, connectors follow real manufacturer geometry, and the modelling gestures are different. If your team is moving from design MEP modelling into fabrication, budget training time. People who are fluent in design Revit are not automatically fluent here.

What actually changes when you model to fabrication

The move from LOD 350 to LOD 400 is not a geometric refinement pass. It is a rebuild against manufacturing constraints.

Real parts replace nominal ones. Every fitting becomes a catalogue item with a manufacturer, a size, a material and a price. Radiused bends replace the design model’s square elbows and they take more space.

Runs get split into manufacturable lengths. Sheet size, standard pipe length, lorry length, lift capacity and site access all set the split points. This is where prefabrication either becomes possible or does not.

Connections get designed. Flanged, welded, grooved, pressed, bolted. Each has a different length allowance and a different installation clearance.

Supports get modelled. Hangers, trapeze brackets, anchors and their fixing points into the structure. This is the single highest-value addition, because support clashes are the ones that stop work on site.

Insulation gets modelled as geometry. A 300 mm duct with 50 mm insulation is a 400 mm object. Design coordination that ignored this now produces clashes.

Access and maintenance zones get modelled. Valve handles, damper access, filter withdrawal, coil pull space. Model them as real geometry so they clash, because a clearance that only exists in a specification will be built over.

Data gets attached. Part marks, assembly marks, weights, batch references, and whatever the information requirements asked for.

Expect the model to grow substantially and expect performance to become a real constraint. Fabrication models are usually split much more aggressively by zone and by system than design models, for exactly this reason. Our guide to Revit performance problems is worth a read before the file becomes unworkable rather than after.

The second coordination round

Here is the step most programmes forget to allow time for. Fabrication modelling introduces new clashes, because the real geometry is bigger than the design geometry and the supports did not previously exist. A zone that was clash-free at design coordination will not be clash-free at fabrication coordination.

Run it as a proper round, not as an afterthought:

1. Federate the fabrication models from all trades plus the current structural and architectural models.
2. Run clash tests with fabrication-appropriate rules. Tolerances that were sensible at design (a 25 mm allowance, for example) are now too loose, because the objects are real.
3. Add clearance and access clash tests. These matter more than hard clashes at this stage.
4. Prioritise by installation sequence, not by clash count. The first trade into the ceiling void wins the space.
5. Resolve in the fabrication model, and record the resolution against the issue.
6. Re-issue and re-test.

The techniques are the same ones covered in our guide to MEP clash detection in Revit and Navisworks. What changes is who resolves the clash and how expensive it is. At design stage a clash costs a modeller an hour. After material has been ordered it costs a reorder, a delivery slot and a programme slip.

Sequence and space allocation matter as much as geometry here, and if the project is running a 4D construction sequence, this is the point where the two workflows should meet.

Spool drawings, assembly marks and the shop drawing set

Fabrication output is not a general arrangement drawing. It is a set of documents each aimed at one person doing one job.

Spool drawings show one prefabricated assembly: its parts, dimensions, cut lengths, weld and joint details, and a unique spool number. The fabricator makes the piece from this and nothing else.

Part and assembly marks are the numbering system that ties a physical object to the model, the delivery note, the drawing and the installation location. Agree the numbering convention before anyone starts, and make sure it survives an export. Steel fabricators have very established conventions here; MEP is looser and benefits from being told.

Installation drawings show the installer where the assembly goes, with setting-out dimensions from a known site reference rather than from a gridline the concrete did not quite follow.

Cut lists and NC data drive the machinery: DSTV files for steel, plasma cutting files for sheet metal, bar bending schedules for rebar. These come from the model automatically when the model is built properly and are retyped by hand when it is not.

The quality gate here is simple to state and easy to skip. Every drawing must be generated from the model, never drawn alongside it. The moment a detailer starts fixing a drawing in 2D rather than fixing the model, the model stops being trustworthy and the whole exercise reverts to CAD with extra steps.

The submittal loop

Shop drawings are contractually reviewed, and the review loop is where fabrication BIM most often reverts to paper.

The pattern that works: submit the drawings as PDF for the formal record, and submit the model alongside them for the review. Reviewers comment on the model using BCF, not by marking up a print, so comments carry a viewpoint and a element reference. Approved-with-comments is tracked as an issue with an owner and a date, not as an annotation somebody has to notice.

Two things to hold firm on. First, the model and the drawing set must be the same revision, or a reviewer approves geometry that no longer exists. Second, the review turnaround has to be in the programme with a real duration. A two-week review cycle on a fabrication package with a six-week lead time is a schedule problem masquerading as an administrative one.

Procurement, prefabrication and the site

Once the fabrication model carries real parts and real assembly marks, it starts driving things other than drawings.

Material take-off becomes procurement. Quantities come from actual catalogue items rather than estimated lengths, which is what makes model-based ordering credible. This is the same data path the 5D cost workflow uses, at a much higher confidence level.

Prefabrication becomes plannable. Assemblies built in a workshop and delivered complete need a delivery sequence, a storage plan and a lifting plan, all of which come off the model.

Status tracking becomes possible. Each assembly can carry a state: modelled, approved, fabricated, delivered, installed. Filtered views of the model become a genuinely useful progress report.

Installation becomes model-based. Setting-out points exported to a robotic total station, and the model available to installers on a tablet. That is the subject of our guide to field BIM, and the fabrication model is what makes it worth doing.

Closing the loop to as-built

The last step is the one that gets cut when the programme tightens, and it is the one the client actually paid for. Fabrication models are the best possible starting point for a record model, because they already describe what was made rather than what was intended.

Capture deviations as they happen rather than in a single exercise at the end. A weekly update of the fabrication model against site reality is achievable; a full as-built survey in the final fortnight is not. Verify against a scan or a 360 capture where the deviation matters, and be honest in the deliverable about which parts are field-verified and which are assumed. A record model that quietly claims accuracy it does not have is worse than one that states its limits.

Whether that model then feeds an asset register or a facilities system depends on the information requirements, which should have been set in the transfer package at the very start of this workflow rather than discovered at handover.

Contract and liability questions worth settling early

Fabrication BIM makes design responsibility visible in a way drawings do not, so settle these before they become a claim.

• Who is responsible if the fabrication model is built correctly from a design model that was wrong? Usually the design team, but only if the transfer package documented what was assumed.
• Is the model or the drawing the contract deliverable? Most contracts still say drawing. Say so explicitly, and keep the drawing generated from the model.
• What is the tolerance regime? Agreed, stated, and consistent between the structural and the fitting-out trades.
• Who owns the federated fabrication model? Someone has to run the second coordination round, and it is a real scope with a real cost.
• What are the model reliance limits? Trade contractors will often accept a model for coordination but not for setting out. That distinction belongs in writing.

Common mistakes

1. No named handover point. Two models of the same system evolve in parallel and diverge quietly.
2. Handing over geometry with no assumptions. The trade contractor fabricates against slab levels that were never real.
3. Skipping the second coordination round. Fabrication geometry is bigger than design geometry. New clashes are guaranteed, not possible.
4. Not modelling supports and access. These cause more site stoppages than duct-to-duct clashes ever do.
5. Drawing over the model. A detailer fixing a drawing rather than the model destroys the model’s authority in a single afternoon.
6. Agreeing part numbering after modelling starts. Renumbering an approved package is expensive and error prone.
7. Treating IFC export as a formality. Test the exchange once, early, on a real zone, before the deadline depends on it.
8. No time in the programme for the submittal loop. Approval duration is a programme item with a lead time attached.
9. Leaving the record model to the end. It becomes a fiction assembled under pressure.

How to start

If your projects currently stop at design coordination, you do not need to solve all of this at once. Take one system on one zone.

1. Pick a single service in a single zone with a cooperative trade contractor.
2. Write the handover table for that system: who models, from when, in what tool.
3. Assemble a real transfer package, including the assumptions and exclusions. Issue it through the CDE with a suitability code.
4. Test the model exchange with them before any real work depends on it. Check coordinates first, geometry second, data third.
5. Let them model to fabrication, then run a proper second coordination round with clearance tests, not just hard clash tests.
6. Generate the spool drawings from that model and check that nothing was drawn by hand.
7. Track the submittal loop as issues with owners and dates.
8. At the end of the zone, hold an honest review with the trade contractor about what did not work, and write it into the BIM execution plan for the next zone.

One zone done properly teaches a team more than a year of coordination meetings. The model quality checks you already run at design stage mostly transfer; what is new is the discipline of treating another company’s model as a first-class part of your federated set.

The point of all of this is straightforward. A coordinated design model proves the building can be built. A fabrication model is how it gets built. The distance between those two sentences is where BIM either pays for itself or becomes an expensive drawing exercise.

If you want to build these workflows properly, from Revit coordination through to the handover disciplines that make a model useful downstream, our Revit and BIM courses teach them the way real project teams run them.