Chapter 35 — Key Takeaways

One-page reference card. Print it. Read it before you write a BIM execution plan, price a coordination program, or sign a subcontract that mentions a model.

Key Takeaways

• 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 is a byproduct.

• A building information model is a database with a geometric interface. When somebody says "send me the model," the useful question back is which model, and what are you going to do with it?

• Three models, three purposes, three liabilities. The design model is a wish (not for coordination or quantity). The construction model is a plan (not for fabrication). The fabrication model is an instruction (not for anything but that trade's work). Confusing them is the most expensive mistake in this domain.

• Level of development describes how much you may rely on an element — not how detailed it looks. Detail is a property of the picture. Level of development is a property of the promise.

The hard rule: you may not extract a quantity from, or coordinate a clearance from, an element whose level of development does not support it.

• The LOD matrix, not the slogan. "We will use LOD 350" means nothing. Every system, every phase, one number, one owner. Northgate pushed duct mains, hydronic pipe, gravity waste, medical gas, fire protection, feeders, and owner equipment to 400 for coordination and left ceilings at 350 and partitions at 300 — deliberately.

• The execution plan decides whether any of it pays. Most failed BIM efforts failed here, before anybody modeled anything. Nineteen pages on Northgate. Incorporate it by reference into the subcontracts — a BxP no subcontract names is a recommendation.

• Agree the coordinate origin in week one. A stated system, a project base point with real-world coordinates, a survey point, units, north conventions, and a shared origin-test file. One hour. A project without one wastes weeks — and the dangerous failure is the detailer who "fixes" alignment by hand, after which every extracted coordinate is wrong by the amount of the drag.

• Raw clash counts are noise. 1,840 became 61 through tolerance rules (−611), a self-clash filter (−525), grouping (−486), and already-assigned items (−157). Read the clash matrix, not the total — it converts a meaningless number into four specific phone calls, three about process and one about the building.

• Routing priority follows degrees of freedom: gravity drainage → large duct mains → sprinkler mains → large pressure pipe → branch duct and small pipe → conduit and cable tray. Above all six: the structure, which is not a routable system at all.

• 4D finds spatial and sequence conflicts that CPM logic cannot express — and nothing else. It calculates no float, no critical path, no resource level, and it will animate a fiction beautifully. Its highest-value use is sequence validation: the crane trap, found in an afternoon in June, cost $33,900 and would have cost $295,150 in October.

• 5D works where quantity is geometry and fails everywhere else. Steel from an LOD 400 fabrication model: 984.6 tons against a canonical 985. Gypsum board from an LOD 300 coordination model: 338,800 SF against a 412,000 SF takeoff — 17.8 percent low, $306,700, with not a single modeling error in it. A precise wrong number is more dangerous than a rough right one.

• The owner's real deliverable is data, not geometry. Rungs 3 and 4 pay today. Capture the data as it arrives; do not reconstruct it in the last six weeks from equipment already above a hard ceiling.

• The economics, honestly. Northgate's program cost $643,500 (1.35% of the GMP, $4.88/SF) — $369,000 of it paid to trade contractors to model. Net identified savings $861,750. Ratio about 1.3 to 1. Not five to one. And the value is in the distribution, not the average: one finding was 30 percent of the gross return.

• A BIM requirement with no budget produces a bad model — and the most common response is the worst one: the trade details on paper, builds from the paper, and produces a model afterward. A reverse-engineered model is worse than no model, because it looks like compliance.

• Coordinate the model. Verify the building. Two different verbs; you need both. And on the evidence of Case Study 1, a third: verify the process — ask which trade is not actually modeling, and how you would know.

• On a $2,000,000 tenant fit-out, full coordination will very often not pay for itself. Scope the model uses to the actual congestion. Sometimes the answer is a marker and a ladder.

Action Items — this week, on your job

1. Find your execution plan and read section 6. If there is no stated coordinate system, project base point, survey point, units, and north convention — write them today and issue an origin-test file. One hour. Nothing else on this list works until this is done.

2. Build the LOD matrix row for your three most congested systems. Not a slogan, a table: system, phase, number, owner. Then run the one query that matters — is anything in the model below its matrix cell? Make that report a condition of sign-off.

3. Open your MEP subcontracts and look for the detailing allowance. If there is not a stated dollar figure for modeling in each of them, you do not have a coordination program. You have a requirement. Price it and take it to somebody with authority this week.

4. Ask, by RFI, what the design models are represented to be — LOD by system, and whether they are suitable for construction coordination. Ten minutes, and it ends an argument you have not had yet.

5. Put coordination sign-off on the CPM as a milestone with predecessors and successors, level by level, with fabrication release tied to it in the subcontract. Without that clause, a sub under schedule pressure will fabricate ahead of sign-off and hand you the bill.

6. Run one 4D week. Pick the single most spatially constrained week on your job — the crane pick, the equipment set, the enclosure hand-off — and build only that. Cheapest 4D you will ever produce, and it is where the money is.

7. Ask your owner which rung they want, and write the asset-data field list before the first equipment submittal is approved.

8. Ask one uncomfortable question in Thursday's meeting: whose foreman has actually opened the model this month? The answer tells you which trade reverse-engineered.

Common Mistakes — and the fix

Decision Framework

Before you commit to a program — five questions

When a clash report lands — work it in this order

1. Filter before you look. Tolerance rules, self-clash, grouping, already-assigned. Never open a raw report in a meeting.
2. Read the matrix, not the total. Separate process problems (a trade on a stale file, lighting laid out without the coordinated overhead) from building problems (genuine congestion). Fix the process problems with phone calls before Thursday.
3. Triage into three buckets: must resolve · resolve if convenient or route elsewhere · false positive. Record the rule that suppressed each false positive; do not delete it.
4. Check the LOD before you rule. An item depending on an element below its matrix cell cannot be decided — only flagged.
5. Apply the routing-priority convention, in writing, the same day. Say it once at the first meeting and never re-litigate it.
6. Escalate anything the structure cannot yield to. That is an RFI, never a field decision.
7. Sign off per level, in writing, all parties — after which a deviation without written approval belongs to the deviating party.

The one test that separates a 4D conflict from a geometric one

Does the conflict disappear if you change only dates and move nothing in space? If yes, it is a 4D conflict and no clash detector will ever find it. A geometric test asks where. A 4D test asks where and when.

The economics, in one line each

• Price every option twice — direct dollars and days at your daily exposure rate. Northgate: $10,650/CD. Willow Street: $2,800/CD. Know yours by heart.
• Price the options you intend to reject. That is what turns a number into a decision.
• Do not quote a multiplier. Give the shape: a detailer's afternoon in the model, tens of thousands in the field after fabrication, tens of thousands again plus schedule after the ceiling closes. Illustrative magnitudes, not a statistic.
• Buy it like insurance. The return lives in the tail, not the average — which is exactly how risk R-07 was priced on Northgate: 50% × $230,000 = $115,000 expected value, and that expected value is what bought the budget.

The one sentence to carry out of this chapter: A model is worth exactly what somebody was paid to put into it, and worth nothing at all if nobody wrote down what you are allowed to rely on.