Chapter 35 — BIM for Construction: 3D Coordination, 4D Scheduling, 5D Estimating, and Digital Twins

The Hook: The Afternoon the Crane Got Trapped

It is a Thursday in the middle of December, Year 1, and Grace Lindqvist has given Dani Okonkwo an hour.

Grace runs virtual design and construction — VDC, which is the umbrella term for the modeling, coordination, sequencing, scanning, and layout work that used to be called "the BIM guy" — for Kestrel Construction Group. Her room is the smallest one in the trailer complex, it has three monitors and no window, and it is where the Northgate Outpatient Pavilion is being built for the first time.

I sent Dani over because Dani had said, in a meeting, "the model," in the tone people use when they mean "the pretty picture."

Grace does not open the pretty picture.

"Level two," she says, and puts up a view that looks like somebody dropped a bag of spaghetti into an architectural drawing. "Federated model. Structural, architectural, ceilings, duct, hydronic pipe, plumbing, medical gas, sprinkler, power, lighting, low voltage. Nine models, one file. Now watch the counter."

She runs the clash test. The counter climbs and stops.

1,840.

Dani makes a noise. You may have made the same noise reading Chapter 10, where that number opened the chapter and nine people sat around a table looking at it.

"That number is not the answer to anything," Grace says. "That number is the input. Watch."

She starts stripping. Insulation touching a hanger inside a tolerance she set herself: gone. Two Cardinal Mechanical hangers 0.2 inches apart — Cardinal's own model clashing with Cardinal's own model, which is Cardinal's problem and not a coordination meeting's problem: gone. Sprayed fireproofing brushing duct insulation by a tenth of an inch: gone. Sprinkler branch lines passing through the structural model's floor slabs, because at this level of development the structural model does not carry the penetrations: gone, and flagged as a modeling artifact rather than a conflict. Then she groups. One 60-foot duct main crossing one sprinkler main does not generate one clash; it generates fourteen, because the software reports every intersecting element pair, and a duct main is made of a lot of elements.

The counter goes 1,840. Then 704. Then 218.

Then 61.

"Sixty-one things that require a human being to make a decision," Grace says. "Thirty-eight of them will be closed by a detailer moving a line before Monday. Twenty-two are trade-to-trade negotiations we will run in Thursday's meeting. And one of them" — she zooms to grid line 6 — "is this."

It is the level-two corridor. It is the one you already know: thirty inches of ceiling cavity, thirty-nine inches of systems, and a transfer girder that cannot move.

"That's the whole job right there," Dani says.

"That is half of it," Grace says. "Sit down. I want to show you the other half, because nobody ever does."

She closes the clash view and opens a different one. Same building, but there is a date stamp in the corner and a slider along the bottom. She drags it back to June.

"This is the erection sequence as Wei Chen loaded it in May, when the precast erector's approved sequence came in. I am going to play it. Don't look at the steel. Tell me what goes wrong."

She plays it. Columns and beams grow out of the foundations in five colored sequences. The 275-ton crawler crane — the one Margo and I argued about for two hours in Chapter 21, the one that works inside the building footprint on a timber mat road and walks out through the last bay — creeps around inside its own building. Around week eight, grey panels start climbing the north elevation. The sequence runs to the end. The building is up.

"Nothing," Dani says. "It works."

"Watch week nine."

She plays it again and freezes it. The crawler is sitting at grid 9, line D, with the last four roof girders in the northeast bay still to set. The north elevation precast — twenty-two panels, set in week eight to protect the March 28 dried-in date — is complete from grid 1 all the way to grid 11.

"Now get the crane out of the building," Grace says.

Dani looks at it for maybe fifteen seconds. Then their face changes.

The mat road out of the building runs through the northeast bay. It is twenty-two feet of clear opening. With the precast panels set on their haunches and their erection bracing still in, that opening is eleven feet six inches. The crawler is fourteen feet eight inches wide over extended tracks.

The crane finishes the building and cannot leave it.

Nobody had seen that. Not Margo Deacon, who has run steel for thirty-one years. Not Wei Chen, whose schedule it was. Not Hank Duffy at Ironbridge Steel, whose erection drawings it was. Not the precast erector, who was doing exactly what his approved sequence told him to do. Not me. The sequence sat in the schedule for five weeks looking completely reasonable, because on a Gantt chart Set precast, north elevation and Erect steel, Sequence 5 are two bars that do not touch, and there is no logic relationship in the Critical Path Method — CPM — that can express the machine will not fit through the hole.

Grace found it in an afternoon in the last week of June. Week nine of steel erection, counting from the August 27 start, would have arrived in the last week of October. Four months.

Fixing it in June cost $33,900 and zero days.

Finding it in October would have cost, by our own estimate at the time, about $295,000 and eleven calendar days on the critical path.

That is the chapter.

This chapter opens Part VII

Everything before this point has been the general case, learned through one commercial building. Part VII does two things with it. It tests the principles against different markets — housing, where the unit of production is a house and the schedule is a cycle time; heavy civil, where the contract is unit-price and the owner is a department of transportation. And it looks at the tools as management instruments with costs, returns, and failure modes — the same way Chapter 21 treated a crane. Not as a technology showcase. Every tool in this part gets asked the same question: what problem does this solve, what does it cost, what does it return, and what has to be true before it works at all?

Building information modeling gets asked that question first, because it is the tool that most consistently produces value and most consistently gets bought badly.

Here is the argument of this chapter, stated once, plainly, so you can hold me to it:

BIM in construction is not a visualization technology. It is a coordination and scheduling instrument. Its return comes almost entirely from rework avoided and sequence problems found early — and it produces that return only when somebody plans it, budgets it, and pays the trades to participate.

Everything else — the renderings, the walkthroughs, the marketing — is a byproduct. Useful, sometimes valuable, never the reason.

🏃 Fast Track: If you have run coordination before, skim §35.1 and §35.4 (you know the workflow from Chapter 10) and go straight to §35.2 (level of development as a reliance rule, which most practitioners get wrong), §35.5 (4D and its honest limits), §35.6 (the quantity-extraction comparison — the $306,700 error), and §35.9 (the full cost-benefit accounting, which is the section that will actually change how you buy this).

🔬 Deep Dive: The coordination problem this chapter tools up for is Chapter 10. The quantity discipline it depends on is Chapter 12. The schedule 4D visualizes is Chapter 14, and the look-ahead planning it feeds is Chapter 27. Reality capture, prefabrication, and robotics get their own treatment in Chapter 39. Turnover and the owner's data package are Chapter 40. The tool landscape is Appendix H.

35.1 What a Building Information Model Actually Is

Building information modeling (BIM) is a database with a geometric interface.

Read that again, because everything in this chapter follows from it. The information is the point. The geometry is how you navigate to the information. A wall in a model is not a picture of a wall; it is a record — type, thickness, height, fire rating, acoustic rating, layer count, cost code, schedule activity, room boundaries on both sides — that happens to render as a rectangle so that you can find it by pointing at it.

When somebody on your job says "send me the model," the useful question back is: which model, and what are you going to do with it? Because there are at least three of them, they are produced by different parties for different purposes, and confusing them is the single most expensive mistake in this domain.

That last cell is the whole of §35.2, so hold it.

The three models, which are not the same model

The design model is a wish. The construction model is a plan. The fabrication model is an instruction.

I have watched a project manager download an architect's design model, extract a quantity, and put it in a guaranteed maximum price. I have watched a superintendent open a design model on a tablet in a mechanical room and try to lay out hanger inserts from it. Both of those are the same error: taking a model produced for one purpose and relying on it for another.

🔍 Why this works. The reason the design model cannot be used for coordination is not that designers are careless. It is that the design model is an expression of intent, and intent does not have a hanger. Trellis Engineering sized Northgate's supply trunk at 42 inches by 16 inches because that is the free area the air needs. They did not model the two inches of external insulation, the trapeze hardware, the one inch of installation clearance, or the fact that Cardinal Mechanical's shop builds that duct in ten-foot sections with a specific seam. Nobody paid them to, nobody asked them to, and their professional obligation is to specify a duct that moves the air, not to decide how Cardinal will hang it. The insulated, hung, serviceable envelope is roughly 21 inches tall, not 16 — a 31 percent difference in the dimension that decides whether the ceiling fits. That difference belongs to the construction model, which is why it exists.

🔄 Check your understanding. An owner's representative asks why Kestrel is paying subcontractors to model systems "the engineer already modeled." Answer in three sentences.

35.2 Level of Development: The Rule That Makes Any of This Safe

Here is the concept that turns BIM from a liability into a management tool, and it is the one most often misunderstood.

Level of development (LOD) describes how much you are entitled to rely on an element. It does not describe how detailed the element looks.

That distinction is not academic. A beautifully rendered chiller downloaded from a manufacturer's website, with every louver and bolt head, can be worthless for coordination if it is not the model number that was actually bought. A crude grey box can be entirely reliable if the box is the exact outside dimension of the unit that is on the truck, with its exact service clearances, sitting at its exact coordinates. Detail is a property of the picture. Level of development is a property of the promise.

The industry uses a numbered framework, commonly running from 100 through 500, maintained and periodically updated by industry bodies and referenced in standard contract exhibits. Get the current edition of whatever framework your contract names; do not work from memory or from this book. What follows is the function of each level, which is stable even as the documents evolve.

The rule, and it is a hard rule

You may not extract a quantity from, or coordinate a clearance from, an element whose level of development does not support it.

Grace enforces this with one sentence stamped at the top of every quantity extraction that leaves her room. You met it in Chapter 12:

"Extracted at LOD 300 — geometry and location reliable, materials and assemblies not."

Twelve words. They are the difference between a useful tool and a $306,700 mistake, which is a real number and which we will work in §35.6.

🧩 Productive struggle. Before you read on. You are handed a coordination model of a hospital floor at LOD 350. Someone asks you to use it for four things. Which of the four are legitimate, and which are not — and why?

1. Confirm that the corridor ceiling cavity can accommodate the duct, sprinkler main, and lights.
2. Extract the total linear feet of 4-inch sprinkler pipe for a change-order price.
3. Extract the square feet of gypsum board for the drywall subcontract buyout.
4. Generate hanger insert coordinates for the layout robot to mark on the deck.

Take three minutes. Write down your answers and your reason for each before you look.

The LOD matrix, which is the core of the execution plan

Saying "we will use LOD 350" is meaningless, because different systems need different levels at different times, and the whole value of the concept is in the specificity. What you actually need is a matrix: every system, every phase, one number, one owner.

Here is Northgate's, abbreviated. Read the coordination column especially — notice which systems Grace pushed all the way to 400 for coordination, and ask yourself why.

Three things to notice, because they are the reasons a matrix beats a slogan.

Ceilings are at 350 for coordination, not 300. The ceiling plane and every soffit and bulkhead is a hard boundary in a coordination model. Model ceilings at 300 and you get a plane with no thickness and no grid, and the cable tray that sits one inch below it passes clash detection.

Fire protection is at 200 in the construction documents and 400 in coordination. That is not sloppiness. Sprinkler layout is delegated design: the engineer specifies performance and hazard classification, and a licensed fire-protection engineer working for the sprinkler contractor does the hydraulic calculation and the layout, stamps it, and submits it. So the design documents genuinely cannot be better than 200, and the coordination model genuinely has to be 400. Anyone who does not understand delegated design will look at that row and think somebody failed.

Site utilities are delivered at 500. Because when the City comes back in fifteen years to cut the street, the only thing that matters is where the pipe actually is, and the only way to know that is a surveyed as-built taken while the trench was open. That is a field discipline, not a modeling one.

🔄 Check your understanding. A design model shows a 30-inch chilled-water pipe main running through a mechanical room at what looks like a fully detailed level — you can see the flanges. The BIM execution plan says piping is at LOD 200 in design development. Can you use it to confirm the pipe clears the air handler? Why or why not?

35.3 The BIM Execution Plan: The Document That Decides Whether Any of This Pays

Most failed BIM efforts failed here, before anybody modeled anything.

The BIM execution plan — written BxP or BEP depending on whose template you inherited — is the document that says what you are actually going to do with the model, who does it, to what level, by when, on whose coordinates, and what happens at the end. It is short. On Northgate it is nineteen pages. It is also the highest-leverage document on the project after the subcontracts, because a project with a BIM requirement and no execution plan will spend real money and produce a picture.

First decision: which model uses you are actually pursuing

There are a couple of dozen recognized "model uses" floating around the industry. The single most common mistake in a BIM execution plan is to list all of them, because listing them is free and doing them is not.

Pursue the ones that pay. Write down the ones you are not pursuing and why, because that sentence will save an argument in month eight.

| Model use | Northgate ($47.5M healthcare) | Willow Street ($6.8M community center) | Why | |---|:--:|:--:|---| | Design authoring | ✔ design team | ✔ design team | Not the contractor's decision | | Energy and code analysis | ✔ design team | ✔ design team | Design-phase; contractor consumes results | | 3D coordination / clash detection | ✔ full — all systems, 4 rounds per level | ✔ limited — gym roof, kitchen, mechanical room only | The core value. Scope it to where systems are actually congested | | 4D sequencing | ✔ structure, enclosure, MEP zones, site logistics | ✔ gym roof structure and RTU setting only | Pays where sequence is spatially constrained | | 5D estimating | ✔ limited — steel, concrete, curtain wall, precast | ✘ | Pays only where quantity is geometric; see §35.6 | | Model-based layout | ✔ full | ✔ partitions, sleeves, anchor bolts | Cheap once the model exists; real productivity | | Prefabrication / multi-trade racks | ✔ corridor racks, levels 2 and 3 | ✘ | Needs volume and repetition to pay | | Site logistics and phasing | ✔ | ✔ | Nearly free; high communication value | | 4D safety planning | ✔ | ✔ (gym roof, RTU picks) | Cheap add-on to an existing 4D | | Reality capture / scan-to-BIM | ✔ existing clinic tie-in + 6 verification scans | ✘ except utility as-built | Essential on renovation; optional on greenfield | | As-built model + asset data handover | ✔ | ✔ reduced scope | Owner-driven; must be bought, not assumed | | Digital twin / live operational integration | ✘ | ✘ | Meridian's facilities group is not staffed for it. Say so. | | Owner and public communication animations | ✔ (2 deliverables) | ✔ (1 deliverable, public meeting) | Cheap, and it buys you goodwill you will spend later |

Notice that the two projects have genuinely different answers, and that on the smaller one most of the boxes are unchecked. That is not a failure of ambition. That is §35.10.

The execution plan outline, as an artifact you can use

The shared origin, and the weeks it costs

Section 6 of that outline looks like housekeeping. It is not.

Every model has an origin — a 0,0,0 point — and a rotation. The architect's model may be placed on a real-world survey coordinate system. The structural model may be built on an internal origin at the intersection of grid A and grid 1. The mechanical detailer, working in a different office on a different contract, may have started at a different internal origin. All three models are internally correct. Federate them without an agreed origin and the ductwork appears in a parking lot two thousand feet away, or eleven feet below the slab, or rotated ninety degrees.

A project with no agreed origin will waste weeks. I have seen three. Every one of them was discovered in the first federation, which is the good case; the bad case is a detailer who "corrected" it by moving their model by hand, so that it looks right and every extracted coordinate is wrong by the amount of the hand correction.

The fix is one paragraph written in week one: a stated coordinate system, a stated project base point with its real-world coordinates, a stated survey point, stated units, a stated project north and true north, and a small shared "origin test" file — three cubes at known coordinates — that every party inserts into their model to prove alignment before they model anything else. It takes an hour.

⚖️ What the contract says. The execution plan should be incorporated by reference into the subcontracts, not filed on a server. A BIM execution plan that no subcontract references is a recommendation. A BIM execution plan named in the subcontract as a contract document is an obligation with a scope, a schedule, and a remedy. On Northgate the mechanical, electrical, plumbing, fire protection, low-voltage, precast, and curtain-wall subcontracts each name it, each carry a stated detailing allowance, and each tie the fabrication-release milestone to coordination sign-off. That last clause is worth its weight: without it, a subcontractor under schedule pressure will fabricate ahead of sign-off to protect itself, and then hand you the bill when the model moves.

35.4 3D Coordination: The Workflow That Produces the Money

You saw the shape of this in Chapter 10. Here we run it as an operation.

The six steps

1. Trade models produced to the agreed level of development, on the agreed origin. Every trade models its own work — the installer, not the engineer. Cardinal Mechanical's detailers model duct that Cardinal's shop will fabricate, at the sizes and fitting types Cardinal actually builds. That is the entire economic point: the model is a fabrication instruction, not a picture.

2. Federation. Combining the trade models into one file. Whoever federates — on Northgate, Grace — does not own the trade models and does not edit them. She assembles them, and the assembly is disposable; the trade models are the source of truth for their own scope.

3. Clash detection runs. And here the distinctions matter more than most people realize.

The second row is where most coordination programs are quietly mediocre. A soft clash cannot be found by a machine that has not been told what the clearance is. Somebody has to model an access zone in front of a panel, above a filter rack, around a valve, under a coil-pull, and somebody has to write a rule that tests for intrusion into it. That is skilled, unglamorous work, and it is why an experienced VDC coordinator is worth what they cost.

4. Triage. This is what Grace showed Dani in the hook, and it deserves to be taught rather than implied.

   RAW GEOMETRIC RESULT  ......................................  1,840
     │
     ├─ TOLERANCE RULES ────────────────────────────────────────  −  611
     │    insulation-to-hanger contact under 0.5"
     │    fireproofing-to-insulation contact under 0.25"
     │    sprinkler branch through structural slab (no
     │      penetrations modeled at this LOD — artifact)
     │
     ├─ SELF-CLASH FILTER ──────────────────────────────────────  −  525
     │    Cardinal vs Cardinal, Halcyon vs Halcyon.
     │    Real, but resolved inside one company's model —
     │    not a coordination-meeting item
     │
     ├─ GROUPING / CLUSTERING ──────────────────────────────────  −  486
     │    one duct main crossing one sprinkler main reports
     │    as 14 element-pair clashes. It is ONE decision.
     │
     ├─ ALREADY-ASSIGNED (open from round 2) ───────────────────  −  157
     │
     ▼
   ITEMS REQUIRING A HUMAN DECISION THIS WEEK  .................     61
        38 closed by a detailer moving a line
        22 trade-to-trade negotiations for Thursday
         1 escalated to a design decision  ◄── the level-2 corridor

Check the arithmetic: 1,840 − 611 − 525 − 486 − 157 = 61.

Three things about that funnel:

• The tolerance rules are a judgment call with a signature on it. Set them too tight and you drown the meeting in noise; set them too loose and you suppress real conflicts. They belong in the execution plan, by system pair, agreed by everybody, and changed only in writing. A coordinator who quietly loosens a tolerance to make a report look better has committed the modeling equivalent of hiding a nonconforming weld.
• The self-clash filter is not "ignoring" anything. Cardinal's hanger hitting Cardinal's hanger is a real problem that Cardinal fixes. It is simply not a problem that eight companies need to sit in a room to discuss.
• A coordinator who reports raw counts is either inexperienced or hiding. The raw number is a measure of how many elements are in the model, not how much trouble you are in.

5. Resolution meetings. Run by the VDC manager, with trade foremen and detailers both present — the detailer because they will make the change, the foreman because they know whether the change can actually be installed. Two hours, weekly, screen on the wall, and a hard rule that a company sending someone without authority to move their own work has not attended.

The meeting applies the routing-priority convention from Chapter 10. Say it out loud at the start of the first meeting and never argue it again:

And above all six: the structure, which is not a routable system at all. When a system conflicts with a brace or a girder, the system moves — or, if it cannot, you write an RFI to the structural engineer, and that is a design question, not a routing decision.

6. Sign-off. Per level, in writing, all parties. From that moment an installer who deviates from the coordinated model without written approval owns the consequence.

Reading a clash matrix, which is a diagnostic instrument

A single total tells you nothing. A clash matrix — every trade against every other trade — tells you which conversation to have. Here is Northgate level three, round 2:

Total: 3,195. Now stop looking at the total and read the table.

• Structure ↔ duct at 386 and structure ↔ sprinkler at 204. Two trades are modeling without the structural model loaded, or loaded from an old version. That is a five-minute fix and a file-management conversation, not a coordination problem.
• Duct ↔ electrical at 296 and arch/ceilings ↔ electrical at 214. Halcyon laid out lighting from the reflected ceiling plan without the coordinated duct in front of them. Also a process problem, not a design problem.
• Duct ↔ sprinkler at 231. This one is real congestion, and it is where Thursday's two hours should go.
• Structure ↔ architecture at 14. Healthy. Those two models are talking to each other.

The matrix converts a meaningless number into four specific phone calls, three of which are about process and one of which is about the building. That is the whole skill.

The resolution log

Every open item gets a row. No exceptions, no verbal closures.

Notice the fifth row. It is dated five and a half months earlier than the rest and it is not a geometric clash at all. That is the point of §35.5.

📋 Try it. (Do this one on paper before you open the answer. It is the drill that separates people who have run coordination from people who have watched it.)

Grace hands you the level-three clash report the afternoon before the round-two meeting. Fourteen items. Here they are.

Your tasks:

(a) Triage all fourteen into must resolve, resolve if convenient / route elsewhere, and false positive.

(b) Four of these are true routing conflicts. Identify them, apply the routing-priority convention, and state which system gives way and why.

(c) Identify the one item that is a 4D conflict, not a geometric one, and say how you know.

(d) Take the worst item on the list and estimate what it would have cost had it been found in the field instead of in the model. Show your arithmetic.

🔄 Check your understanding. Your clash matrix shows structure ↔ sprinkler at 204 and structure ↔ architecture at 14. What is the most likely explanation, and what do you do about it before the meeting?

35.5 4D: The Schedule You Can Watch

4D is the model with the schedule attached: each model element is linked to the CPM activity that builds it, and time becomes a slider.

How it is actually built

Not by magic and not by a button. It is built by a mapping, and the mapping is only as good as the schedule's structure.

Step 2 is the one that decides whether 4D is possible at all, and it happens months earlier, in the work breakdown structure. A schedule whose activities are Erect structural steel and Set precast panels cannot be linked to anything useful, because those activities cover the whole building. A schedule whose activities are Erect structural steel — Sequence 4, grid 7–8, lines A–F can. If you want 4D, you have to want it before you build the schedule.

What CPM can and cannot express

This table is the argument of the section.

The practical rule: 4D finds spatial and sequence conflicts that CPM logic cannot express.

And its corollary, which is the honest limit:

4D does not calculate anything the CPM did not already calculate. It is a visualization of the schedule, not a schedule.

A 4D simulation will not level your resources, will not compute float, will not find your critical path, and will happily and beautifully animate a schedule that is complete fiction. If the CPM says the curtain wall goes up in six weeks and it actually takes eleven, the 4D will show you a confident, smooth, wrong six weeks. Garbage in, cinematic garbage out.

What it is genuinely good for

1. Sequence validation. The crane in the hook. This is the highest-value use, and it is worth more on constrained sites than anywhere else.
2. Site logistics and phasing. Where the laydown is in month four versus month eleven, when the construction entrance has to move, when the temporary road gets built over.
3. Trade flow and work-zone planning. Watching how many trades are in a zone in a given week is the fastest trade-stacking check there is.
4. Communicating a sequence to people who do not read Gantt charts. Which is most people, including a great many owners, most public boards, and a fair number of foremen who can build anything and have never been taught to read a network diagram. This is not a soft benefit. A sequence people understand is a sequence people follow.
5. Look-ahead visualization for lean planning. In Chapter 27 you build a six-week look-ahead and a constraint log. Putting the next six weeks on a screen, by zone, with each trade in its own color, turns a pull-planning session from an argument about memory into an argument about the building. Grace runs the level-2 and level-3 interior sequence this way every month, and it is the single most-used 4D product on Northgate — far more used than the erection study that saved the most money.
6. Safety planning. See the alert below.

The crane, worked

Here is the finding from the hook, priced the way you would have to price it for me.

                      N ↑          ARDMORE CLINIC   (stays open; access required)
  ══════════════════════ NORTH PROPERTY LINE ═══════════════════════════════
   ▓▓  ▓▓  ▓▓  ▓▓  ▓▓  ▓▓  ▓▓  ▓▓  ▓▓  ▓▓  ▓▓  ▓▓  ▓▓  ▓▓  ▓▓  ▓▓  ▓▓  ▓▓
                                     └──── 22 precast panels, set week 8 ────┘
  ┌────┬────┬────┬────┬────┬────┬────┬────┬────┬────╥═════════╗
A │ S1 │ S1 │ S1 │ S2 │ S2 │ S2 │ S3 │ S3 │ S5 │ S5 ║  ░░░░░  ║ ← WALK-OUT BAY
B │ S1 │ S1 │ S1 │ S2 │ S2 │ S2 │ S3 │ S3 │ S5 │ S5 ║  ░░░░░  ║
  ├────┼────┼────┼────┼────┼────┼────┼────┼────┼────╨═════════╢
C │ S1 │ S1 │ S1 │ S2 │ S2 │ S2 │ S4 │ S4 │ S5 │ S5 │   S5    │
D │ S1 │ S1 │ S1 │ S2 │ S2 │ S2 │ S4 │ S4 │ ■■ │ S5 │   S5    │  ■■ = 275-ton
E │ S1 │ S1 │ S1 │ S2 │ S2 │ S2 │ S4 │ S4 │ S5 │ S5 │   S5    │       crawler,
  └────┴────┴────┴────┴────┴────┴────┴────┴────┴────┴─────────┘       week 9
    1    2    3    4    5    6    7    8    9   10   11

    S1..S5  = steel erection sequence          ▓ = precast panel, set week 8
    ░ = timber mat walk-out road

    MAT ROAD CLEAR WIDTH, as planned  ..................  22'-0"
    CLEAR WIDTH with panels set and erection bracing in  11'-6"
    CRAWLER WIDTH over extended tracks  .................  14'-8"
    ─────────────────────────────────────────────────────────────
    RESULT: the machine finishes the building and cannot leave it.

The fix, taken in June:

The same problem, discovered in week nine of erection (last week of October, Year 1):

💰 Money check. The arithmetic, stated the way you would state it to a project executive:

$295,150 − $33,900 = $261,250 avoided, from one afternoon of work in the last week of June.

That single finding is 41 percent of Northgate's entire VDC program cost for the whole job (see §35.9: $643,500). One conflict. Found by playing a movie.

And the thing to sit with: the schedule was not wrong. Wei's logic was correct. Every activity had a defensible duration and a defensible predecessor. Ironbridge's erection drawings were correct. The precast erector's approved sequence was correct. Every single document was right, and the job was still going to stop, because no document in the standard set expresses the sentence "the machine will not fit through the hole." That sentence lives in three dimensions plus time, and until somebody puts three dimensions plus time on a screen, it is invisible to everyone.

That is Theme 3 in its purest form: the project is built twice, and the first build determines the second.

⚠️ Safety alert. The cheapest safety product a 4D model makes is a crane and pick-path study with the exclusion zones turned on. Play the sequence and watch what is under the load path in each week: an active clinic entrance, a pedestrian route, a trade working below, a laydown where people eat lunch. On Northgate, the same 4D that caught the crane trap also showed that in Sequence 3 the pick path for the west bay crossed directly over the temporary walkway Ardmore's patients use, on four specific days. That walkway was rerouted for those four days, at a cost of about $2,100 in temporary fencing and signage. Nobody will ever be able to prove that decision prevented anything, which is the permanent frustration of safety work and no reason at all to skip it. Fall exposure, leading-edge work, and access routes can be visualized the same way, and should be, in the same session where you plan the sequence — not in a separate meeting that nobody from the field attends.

35.6 5D: Cost Linked to the Model

5D attaches cost to model elements: each element carries a cost code, quantities flow to an estimate, and the estimate updates as the model changes.

The promise is seductive and the reality is narrow. Here is the honest version.

Model-based quantity takeoff: what it does well and badly

The comparison, worked on Northgate

This is the exercise that will change how you use a model. Take a single element — interior gypsum board — and put three numbers side by side.

Chapter 12 built the manual takeoff, and it built it as a derivation with its assumptions on the page: 18,600 LF of partition, a weighted average height of 10.5 feet, two faces, a second layer on rated corridor and shaft walls, a deduction convention for openings over 50 square feet, and an allowance for soffits and furring. That produced 412,000 SF.

Now extract the same thing from the coordination model. The coordination model is at LOD 300 for partitions — accurate thickness, accurate location, which is exactly what coordination needs.

73,200 square feet. The model is 17.8 percent low.

Now explain the difference line by line, because that is the actual skill:

1. Wall face area is 24,200 SF low. The model's average wall height works out to 9.84 feet against the takeoff's weighted 10.5. Why? Because for coordination purposes, what matters about a partition is where its top is relative to the ceiling cavity. A block of partitions in the imaging suite was modeled to the underside of the ceiling rather than to deck, and 137 walls were modeled at a placeholder unconnected height of 10 feet 0 inches and never updated. Both were entirely correct decisions for coordination. Both are wrong for quantity.

2. Layer count is 25,200 SF low, and this is the big one. The model reports wall surface area — two faces. It does not multiply by the number of board layers, because the wall type carries the layer information as data, and the area parameter is geometry. On rated corridor and shaft walls that is a whole additional layer of board, 2,400 LF of it. To get this right you have to query by wall type and apply layer counts by type — which is possible, and which nobody does unless they were taught to.

3. Opening deductions are 13,000 SF too generous. The model deducts every opening, including the 3'-0"×7'-0" doors. The estimator does not deduct those, because the board is cut from a full sheet and the drop is waste that gets carried to a dumpster. The model is geometrically right and commercially wrong.

4. Soffits and furring are 10,800 SF low. They live in different model categories, and the extraction query did not include them.

The dollar value of the error. Chapter 12 gives us the check: 18,600 SF of installed board is roughly $78,000, which is $4.19 per square foot installed — board, hang, tape, finish.

73,200 SF × $4.19/SF = $306,708

Round it: $306,700 of installed gypsum board, missing from a number that had three decimal places on it. On a job with a $1,320,000 construction contingency, that is 23 percent of your contingency, consumed by a query.

🔍 Why this works — or rather, why it fails. Notice that not one of those four differences is a modeling error. Every partition is exactly where it should be, exactly as thick as it should be. The model is correct. What failed is the assumption that a model built to answer "does it fit?" can also answer "how much board do I buy?" Those are different questions with different tolerances and different definitions of the same word. Chapter 12's threshold concept was that an estimate is a priced bundle of assumptions — and a model-extracted quantity is a bundle of assumptions too, except that nobody wrote them down and the software displayed the result to two decimal places. A precise wrong number is more dangerous than a rough right one, because nobody checks it.

Where 5D genuinely works

Now the counter-example, so the section is fair.

Structural steel, from Ironbridge's fabrication model. Canonical Northgate quantity: 985 tons erected. The extraction from the LOD 400 fabrication model returns 984.6 tons of members plus connection material. The difference from the takeoff is four-hundredths of a percent, and it is better than the takeoff.

Why does this one work when the drywall one failed? Because the fabrication model and the product are the same artifact. Every piece in the model is a piece that will be cut from a specific length of a specific shape, with a specific mark number, and shipped on a specific load. There is no gap between the representation and the thing. That is what LOD 400 means.

Slab on grade — and the honest limit even where it works. Canonical: 33,000 SF at 5 inches = 510 CY. The model returns 510.4 CY. The takeoff returns 510 CY. Perfect agreement.

And you will still be short, because you do not order 510 cubic yards. Subgrade tolerance means a nominal 5-inch slab places at something closer to a 5.4-inch average:

510 CY × (5.4 ÷ 5.0) = 551 CY ordered

That is 8 percent more than both the model and the takeoff, and it is not an error in either of them. It is a means-and-methods fact about how a slab on a graded subbase actually behaves. The model cannot know it, the takeoff cannot derive it, and only an estimator who has stood next to a concrete pour puts it in the number.

Real-time cost feedback and target value design

The genuinely valuable version of 5D is not the bid. It is the design phase, and it is the engine underneath the target value design work in Chapter 11.

Target value design inverts the normal process: instead of designing and then pricing, you set the allowable cost and design to it. That only works if the cost feedback loop is fast. When the design team can move a floor-to-floor dimension, or change a curtain-wall module, or swap a structural bay spacing, and see a cost consequence in hours instead of the three weeks a full re-estimate takes, the conversation changes character entirely. It stops being "here is what your design costs" — which arrives too late and feels like an accusation — and becomes "here are four options and what each one does to the number," which is a design conversation.

Two honest caveats. The feedback is only as good as the unit costs behind it, which are historical and must be maintained. And it is only reliable for the geometric portion of the cost — the elements whose quantity comes from shape. It will tell you what six inches of floor-to-floor costs in skin, steel, and stair. It will not tell you what it costs in coordination, which is exactly the number that killed value-engineering item VE-06 on Northgate when Sofia Marchetti of Cardinal Mechanical put Grace's model on the screen and took four minutes to explain why eight fewer inches of ceiling cavity costs more than it saves.

🔄 Check your understanding. Your model extraction reports 1,238.7 CY of footing concrete. Your manual takeoff says 1,240 CY. Which do you put in the estimate, and what do you write next to it?

35.7 The Other Model Uses: Layout, Prefabrication, and Reality Capture

Each of these deserves a chapter and gets a page. Chapter 39 takes the hardware side further.

Model-based layout

A robotic total station or a layout robot takes coordinates directly from the coordinated model and marks points on the deck — hanger inserts, sleeve centers, wall lines, anchor bolts, equipment pads. One person operates it. A layout robot drives itself.

The advantage is not speed, though it is faster. The advantage is that the point comes from the coordinated model, so the sleeve you drill is the sleeve the mechanical contractor's duct is actually going to use, and the hanger insert you set is the one the trapeze in the fabrication model was built for. You are transferring a coordinated decision to the floor, rather than transferring a drawing dimension and hoping.

The prerequisites are unglamorous and absolute: a signed-off model at fabrication level, a verified control network on the site, and a building whose as-built dimensions have been checked against the model. Skip the third and you will lay out three hundred perfectly accurate points in the wrong place.

Grace set Northgate's layout workflow up in preconstruction, which is the only time it is cheap to set up. The site control, the model coordinate system, and the survey benchmarks are the same decision, made once. Set it up in month six and you are retrofitting a coordinate system onto a building that already exists — see Chapter 17 and Chapter 18.

Prefabrication and multi-trade racks

Chapter 10 made the case; here is the part that belongs to this chapter. Prefabrication is a model use, and it changes what the model has to be.

A stick-built installation tolerates a model that is 95 percent right, because the last five percent gets solved by a fitter with a tape measure. A multi-trade rack — a forty-foot section of corridor ceiling with duct, sprinkler, medical gas, domestic water, conduit, hangers, and support steel, assembled in a shop on a jig and lifted in as a unit — does not. If the model is wrong, or if the building does not match the model, you have manufactured the error forty feet at a time.

So prefabrication converts a coordination risk into a fabrication risk, and it is appropriate on jobs with disciplined coordination and genuinely dangerous on jobs without it. The prerequisite is not enthusiasm. It is a signed model and a verified building.

Reality capture and progressive verification

Laser scanning and photogrammetry capture what actually exists, as a point cloud — millions of measured points — which can be brought into the model environment and compared against it.

Three uses, in descending order of how settled they are:

1. Existing-conditions capture. On any renovation or tie-in this is close to essential, and it pays immediately. Northgate connects to an existing Meridian clinic that stays open; the tie-in was scanned before design development, and the scan found that an existing corridor slab was 2.75 inches higher than the record drawings showed. Discovered in design, that is a threshold detail. Discovered during construction, that is an accessibility problem, a change order, and an argument.

2. Scan-to-BIM. Converting a point cloud into modeled elements. Useful, labor-intensive, and frequently over-scoped: you rarely need a full model of an existing building. You usually need the structure, the primary systems, and the interface surfaces. Scope it.

3. Progressive verification. Scanning installed work and comparing it to the model to catch deviation before it is buried. Northgate ran six scan events: post-foundation, three deck scans as the frame rose, one after level-2 and level-3 overhead rough-in, and one before ceiling close-in.

And here is the honest limit, which cost Kestrel real money and appears in Case Study 1: a verification program has a frequency, and things happen between scans. The level-3 slab deflected under the wet lightweight topping and finished about 1¼ inches high over a forty-foot stretch at grid lines 6 to 9. That happened between scan three and scan four. The multi-trade rack for that run had already been released for fabrication against scan three. The rack was built correctly, to a model that was correct, for a building that had moved. It cost $18,400 and four days.

Coordinate the model. Verify the building. They are two different verbs and you need both.

35.8 Digital Twins and What the Owner Actually Gets

This is where the marketing is thickest, so let us be careful.

A digital twin is a model connected to live operational data — building automation, sensors, meters, the work-order system — so that it reflects the building's current state rather than its designed state. Not a 3D model. Not an as-built model. A model with a live feed.

Almost nothing sold as a digital twin on a construction project is one. What is actually delivered sits somewhere on a spectrum, and it is worth knowing which rung you are buying:

Rung 3 and rung 4 are where the value is today. Not because rung 5 is fake — it is not, and hospitals, campuses, airports, and data centers do operate real twins — but because rung 5 requires an owner with a facilities organization staffed and funded to consume it. Selling a live twin to an owner whose facilities department is four people and a work-order spreadsheet is selling a monument.

What "asset data" actually means

For every piece of equipment that a facility manager will ever have to maintain, replace, or warranty:

Look at the right-hand column. Almost every one of those fields already exists somewhere on your project, in the submittal register, the O&M package, or the subcontract. The work is not creating the data. The work is capturing it in a structured form as it arrives, instead of reconstructing it in the last six weeks from equipment that is already above a ceiling.

COBie — described by function, because that is what matters — is a standardized, spreadsheet-shaped way of carrying exactly that non-geometric facility data: spaces, zones, systems, equipment types, individual components, warranties, spares, and maintenance jobs, structured so a facility-management system can import it without anybody re-typing it. The point of it is the data, not the geometry. An owner who asks for "BIM at turnover" and means a pretty model has asked for the wrong thing; an owner who asks for structured asset data has asked for the thing that saves their facilities group real hours every week for thirty years.

🏗️ From the field. I got this wrong once and it still bothers me. On a job before Northgate, we treated the asset data as a closeout task. Six weeks before substantial completion, we put a temporary employee with a clipboard and a ladder on it. He was diligent and he was doomed: by then a third of the equipment was above hard ceilings, another chunk had nameplates facing a wall, and the mechanical sub's foreman who knew where everything was had demobilized. We delivered a spreadsheet with roughly 40 percent of the serial numbers wrong. The owner did not catch it. Their facilities manager caught it eleven months later, on a warranty claim, on a unit whose serial number did not exist. That is not a technology failure. That is a process failure, and the fix costs almost nothing: capture the data at the submittal, verify the nameplate at installation, and make it a line on the daily report — which is exactly where Chapter 40 puts it.

35.9 The Economics: What It Costs, What It Returns, and Who Pays for the Detailing

This is the section that makes the chapter credible or makes it a brochure. So: real numbers, from Northgate, with the uncomfortable ones left in.

What it cost

Two observations before we go on.

First: most of the cost is invisible. Only $46,000 of that $643,500 sits on a line that says anything about BIM. The rest is buried inside general conditions, inside subcontract sums, and inside corporate overhead. This is precisely why people say BIM is free, and precisely why it fails on projects where nobody added it up.

Second, and this is the point of the section: $369,000 of it is money paid to trade contractors to model.

A BIM requirement with no budget produces a bad model.

Write that on the wall. A trade contractor who is told to model, and not paid to model, will do one of three things, all of which you will pay for later: they will decline and force a fight; they will do it badly with whoever is cheapest; or — the most common and the most corrosive — they will detail on paper the way they always have, build the job from the paper, and then produce a model afterward to satisfy the requirement. That last one is worse than no model at all, because it looks like compliance. The coordination never happened. You just bought a very expensive record of a process that did not occur.

What it returned

Every line below was individually identified, logged at the time, and priced by the estimating group when it was found — not reconstructed afterward to make a case.

The ratio, honestly

$861,750 ÷ $643,500 = 1.34 — call it about 1.3 to 1. Net gain: $218,250 on a $47,500,000 project.

That is not the number on the conference slide. You will hear ratios of five to one and ten to one at industry events, usually sourced to a study nobody in the room has read, usually counting savings against a counterfactual project that does not exist. I am not going to give you a multiplier, because I do not believe anybody's, including my own. What I will give you is the shape, which every experienced coordinator recognizes: resolving a conflict in the model costs on the order of a detailer's afternoon; resolving it in the field after fabrication costs on the order of tens of thousands; resolving it after the ceiling is closed costs on the order of tens of thousands again, plus the schedule, plus doing it at night next to an operating clinic. Illustrative magnitudes, not a statistic.

And a warning about my own table: the RFI line is the softest number on it. It compares Northgate against Kestrel's own prior jobs, which had different owners, different designers, and different scopes. It is an estimate dressed as an accounting entry, and I have left it in only because leaving it out would be its own kind of dishonesty. If somebody hands you a BIM return-on-investment case, find the RFI line and ask what it is compared against. That question tells you how seriously to take the rest of the page.

The benefits that are not in the table

Because the honest accounting cuts both ways:

• Safety exposure moved off the ceiling. Roughly 2,240 linear feet of corridor systems were assembled at waist height in a shop instead of overhead on a lift among four other trades. Kestrel has no way to price a shoulder injury that did not happen, and no way to prove it would have.
• Eight calendar days of enclosure float, produced by the 4D area-based enclosure study — which the steel delay then ate. It was worth exactly $0 in the ledger and it was worth a great deal in September of Year 1, because the acceleration described in Chapter 14 started from a better position than it would have.
• The conflicts that never existed. The value-engineering session in Chapter 11 where Sofia Marchetti killed VE-06 in four minutes with Grace's model on the screen prevented a coordination catastrophe that would have surfaced as RFIs eleven months later. It is unpriceable precisely because it worked.
• The owner's turnover data. Meridian's facilities group received an as-built model with a structured asset register. Real value to them. Zero dollars in Kestrel's column.
• Fewer disruption and trade-stacking claims from subcontractors — which shows up as an absence, and absences do not have invoices.

🔍 Why this works. Here is the argument that actually justifies the spend, and it is not the ratio. It is the distribution.

Look at the returns table again. One line — the crane trap — is 30 percent of the gross savings, and it came from a single afternoon in June. Two lines together are 51 percent. BIM's return is not a steady trickle of small efficiencies. It is a small, certain, budgeted cost that buys down a low-probability, high-consequence risk. That is the same instrument as an insurance premium, and you evaluate it the same way: not by the expected value alone, but by what the tail looks like.

Chapter 6 put this in the risk register explicitly. Risk R-07 — uncoordinated MEP driving field rework — carried a $230,000 exposure at 50 percent probability, and the register's own note says the resulting $115,000 expected value is the number that buys Grace Lindqvist's clash-detection budget. That was the right way to think about it then and it is the right way now. The question is never "does BIM pay?" The question is "what does the tail of my rework distribution look like on this job, and what does it cost to cut it off?"

Which is exactly why the answer changes on a smaller job.

35.10 Where BIM Fails — and Who Owns the Model

Every honest chapter about a tool needs this section. Here is where I have watched it go wrong.

The small-project reality, said out loud

On a $2,000,000 tenant fit-out, full multi-trade coordination will very often not pay for itself.

I am going to say that plainly because the industry mostly will not. The reason is arithmetic, not ideology. The cost of a coordination program is roughly proportional to the number of systems, the number of parties, and the congestion — and it has a floor, because you need a coordinator and a process regardless of size. The return is proportional to the rework you avoid, and on a fit-out with an existing ceiling grid, two systems, and one general contractor who has done forty of them, there is not much rework to avoid.

What does pay on small jobs, almost always:

• A scan of the existing conditions above the ceiling, if you are renovating. Cheap, fast, and it will find the abandoned duct that is not on any drawing.
• Coordination of the two or three genuinely congested spots — the mechanical room, the kitchen hood and grease duct, the one corridor where everything crosses. Not the whole job.
• Model-based layout if you already own the total station.
• A 4D sequence for one hard week — the week the equipment comes in through a hole in the roof.

The skill is not "do BIM" or "don't do BIM." The skill is scoping the model uses to the actual congestion of the actual job. That is the whole content of §35.3's second table, and it is why Willow Street's column has so many empty boxes.

Contract and liability

⚖️ What the contract says. This is genuinely unsettled ground, it varies by jurisdiction, and it is evolving. What follows is the framework and the questions to bring to your attorney — not legal advice, and not a substitute for reading your own contract.

1. Ownership and intellectual property. Who owns the design model, the federated model, and each trade's derivative model? What licence does each party get, for what purposes, for how long? What happens to those licences on termination or default? The common structure grants the contractor a limited licence to use the design model for construction purposes only, and grants the owner a licence to the as-built deliverable for operations — but "common" is not "universal," and the default in the absence of a clause is not something you want to discover in a dispute.

2. Reliance. May the contractor rely on the dimensions in the designer's model? The prevailing practice is that designers disclaim reliance and the contract documents — drawings and specifications — govern, with the model provided for information only. That practice exists because a designer's professional liability insurance is written around the documents they seal. Understand it, plan around it, and do not build a workflow that quietly assumes otherwise.

3. The standard-of-care question. When a contractor relies on a designer's model and the model is wrong, the questions are whether the designer's conduct fell below the professional standard of care, whether the contractor's reliance was reasonable given what the model was represented to be, and what the contract said about all of it. Reasonable people litigate this. The best defence is boring: know the represented LOD, put the represented LOD in writing, and do not rely past it.

4. Which documents govern — and this one you can and must settle. Chapter 7 taught you the order of precedence among contract documents. The model needs a place in that order, and it needs it before the first coordination meeting, in both the BxP and the subcontracts. Kestrel's convention on Northgate, stated in one sentence in every MEP subcontract:

The contract documents govern quality, product, performance, and quantity. After coordination sign-off for a given area, the coordinated model governs the three-dimensional location of systems within that area. Deviation from the coordinated model without written approval is at the deviating party's cost.

That sentence is the reason "but the mechanical drawing shows it here" stops being an argument in February. The mechanical drawing showed a single line in plan. The model showed a 46-inch insulated envelope at a specific elevation, and eight companies signed it.

5. What a signature means. Decide, in advance and in writing, what sign-off obligates. Does it mean "I have reviewed this and my work is correctly represented"? Does it release the trade to fabricate? Does it start a clock on change pricing? Ambiguity here is not neutral — it defaults to whoever has the better lawyer.

Industry standard-form contract families publish digital-data and BIM exhibits or addenda for exactly these questions. Use one rather than drafting from scratch, read the version your project actually adopted, and have your counsel confirm how it interacts with the rest of your agreement in your jurisdiction.

Spaced Review

Answer these before you read the responses. Writing the answer down is the point; recognizing it is not.

1. From Chapter 10. Why does MEP coordination, rather than the structure, set the interior schedule — and what is the "ceiling cavity budget"?

2. From Chapter 25. Why is fabrication release the long pole in the submittal chain, and what does that have to do with a coordination sign-off date?

3. Deep callback — Chapter 12. What is a quantity actually a fact about? Now apply it: why was the model's 338,800 SF of gypsum board correct and unusable at the same time?

Project Checkpoint: The Willow Street BIM Execution Plan

In Chapter 34 you looked at Willow Street from the company's balance sheet — the work-in-progress schedule entry, the over- and under-billing position, and what one job does to Kestrel's bonding capacity. Now come back down to the building and decide how you are going to build it the first time.

Your deliverable: a BIM execution plan for the Willow Street Community Center, sized honestly for a $6,800,000 job. Six parts.

Part 1 — Model uses pursued and declined. Use the §35.3 table format. For every use you decline, write the reason in one sentence — and be prepared to defend declining. On a 24,000 SF community center with packaged rooftop units, most of the systems are on the roof or in a mechanical room, and most of the building has generous plenums. That is genuinely different from Northgate. Your case for coordination should be narrow and strong: the gymnasium roof structure (long-span joists, oversized rooftop units on curbs, and the only truly congested plenum on the job), the commercial kitchen (hood, grease duct with its required clearance to combustibles and its slope, makeup air, gas, and the fire-suppression tie-in), and the mechanical and electrical rooms. Not the whole building.

Part 2 — The LOD matrix. By system and by phase, in the §35.2 format. Willow Street has a structural steel and CMU first floor, a wood-framed second floor, open-web joists over the gym, a TPO roof, and a standing-seam canopy. Wood framing is a genuinely interesting row: decide what level it needs and justify it. So is the gym roof, where the joists, the RTU curbs, and the roof drains all compete.

Part 3 — Coordinate origin and file structure. One paragraph, but write it as if it were binding, because it is. State the coordinate system, the project base point with its real-world coordinates, the survey point, units, project north versus true north, the file-naming convention, and the origin test file every party inserts before modeling anything. This is the paragraph that saves three weeks.

Part 4 — The clash-detection process. Rounds (how many, on what dates, tied to which submittal and fabrication milestones), tolerances by system pair, who federates, meeting cadence and required attendees, the issue-tracking method, and — the part people skip — who signs off, on what, and what their signature obligates. Write the governing-document sentence from §35.10 into your subcontract language.

Part 5 — The 4D sequence for the gym roof. This is the one that will actually earn its money. Sequence the joist erection, the roof deck, the RTU curbs, and the RTU setting. Then add the two constraints that make it hard: Willow Street's frontage is a two-lane street across from Danforth Elementary School, with a hard delivery and crane blackout during arrival and dismissal — so your crane has a working window of roughly 9:15 a.m. to 2:30 p.m. — and the RTUs must be set before the roofing crew can complete around the curbs. Show the sequence week by week and state the one conflict you found by building it.

Part 6 — The honest cost-benefit case. Price your program the way §35.9 prices Northgate's, and include the subcontractor detailing line, because if you do not pay for it you will not get it. Then state, in writing, whether it pays — and if part of it does not, say which part and recommend cutting it. A deliverable that recommends spending less is a better deliverable, not a weaker one.

Next chapter you take the same building and ask a different question: what the energy code actually requires of you, what a rating system adds on top, and which parts of "green" are the contractor's problem rather than the designer's.

Chapter Summary

The argument, in one line: BIM is a coordination and scheduling instrument. Its return comes from rework avoided and sequence problems found early, and it produces that return only when somebody plans it, budgets it, and pays the trades to participate.

The reference table:

The eight rules:

1. LOD is about reliance, not detail. You may not extract a quantity from, or coordinate a clearance from, an element whose level of development does not support it.
2. The design model, the construction model, and the fabrication model are three different things. Confusing them is the most expensive mistake in this domain.
3. The execution plan decides whether any of it pays — and most failures happened there, before anybody modeled anything.
4. Agree the coordinate origin in week one. A project without one will waste weeks.
5. Raw clash counts are noise. 1,840 became 61 by tolerance rules, a self-clash filter, grouping, and judgment — and the rules belong in the execution plan with a signature on them.
6. 4D finds spatial and sequence conflicts that CPM logic cannot express — and nothing else.
7. A BIM requirement with no budget produces a bad model, and a reverse-engineered model is worse than no model, because it looks like compliance.
8. Coordinate the model; verify the building. Two different verbs. You need both.

The decision framework — five questions before you commit to a BIM program:

What's Next

Chapter 36 takes the same posture toward sustainability: the energy code as the binding requirement, rating systems as the voluntary layer on top, and a clear accounting of what the contractor is actually responsible for — envelope commissioning and air-barrier continuity, construction waste management, indoor air quality during construction, material documentation, and the paperwork burden all of that puts on your field staff. Same question, asked the same way: what does it cost, what does it return, and which parts are genuinely yours?