Case Study 35-2 — Stonebridge: A BIM Requirement With No Mechanism

All people, companies, and projects in this case are Tier-3 illustrative composites. The failure described here is common; the numbers are one project's arithmetic, not a statistic.

Setup

I did not build this job. I read it.

Two years before Northgate's notice to proceed, Meridian Health System built the Stonebridge Ambulatory Care Center — $28,400,000, 62,000 gross square feet, three stories: imaging on level 1, infusion and procedure rooms on level 2, physical therapy and medical office on level 3. Design-bid-build, lump sum, 480 calendar days, liquidated damages $3,500 per calendar day.

During Northgate's preconstruction, Pri Sethi handed me Meridian's Stonebridge lessons-learned file — about forty pages, most of it invoices. Everything below comes out of that file. It is the reason Northgate's BIM execution plan exists in the form it does.

The requirement. Stonebridge's project manual contained an eleven-page Division 01 section titled Building Information Modeling. It required the contractor to "utilize Building Information Modeling for the coordination of all building systems," to submit a BIM Execution Plan within thirty days of notice to proceed, and to furnish an as-built BIM model at closeout. Design models "may be made available to the Contractor for information only, without representation or warranty as to accuracy or completeness."

What it did not contain. No level-of-development matrix. No coordinate system or shared origin. No model exchange schedule. No clash rounds, tolerances, or rule sets. No statement of who federates. No sign-off authority and no statement of what a signature obligates. No statement of which document governs three-dimensional location. No requirement that trade contractors carry a detailing allowance, and no allowance in the bid form for one. No asset-data field list for the closeout model.

Vantry's estimator read the section, called two subcontractors, received no adders, and carried zero dollars for coordination modeling. The subcontracts passed the section down by reference. In month one, every party on the job was contractually obligated to do something that nobody had priced.

What Happens

Month 1 — the execution plan that wasn't

Arjun Belliappa, a second-year project engineer at twenty percent allocation, is assigned to write the execution plan. He downloads a template. The document submitted on day 34 is four pages and restates the specification in different words. Perrin + Vale stamp it Reviewed. Nobody notices it contains no LOD matrix, no origin, and no dates, because nobody reviewing it knows those are the parts that matter.

Months 2–3 — the origin, and three weeks

Design models are transmitted under a disclaimer. Two of the three MEP subcontractors state plainly that they do not model; they detail. Teddy's answer — the only one available to her — is "the specification requires it."

The first federation attempt is month 3. The mechanical model lands 1,140 feet northeast of the building and rotated 21 degrees. The plumbing model sits eleven feet below the slab. Every model is internally correct and no two agree, because nothing told them what point to build on.

Three weeks disappear, and one worse thing happens quietly: the fire-protection detailer fixes his own alignment by hand. It now looks right, and every coordinate it will ever produce is wrong by the amount of the drag.

A paragraph in month one — stated coordinate system, project base point with real-world coordinates, survey point, units, project north and true north, and a shared origin-test file with three cubes at known coordinates — would have taken an hour.

Months 3–4 — the number in the room

The first clash run is reported unfiltered, in an owner–architect–contractor meeting, on a screen: 31,600 clashes.

Pri Sethi asks the question any owner would ask: is the design defective?

Perrin + Vale's principal answers correctly and uselessly. The models were furnished for information only; the MEP models are design intent at roughly LOD 200 to 300 — no insulation, no hangers, no access zones, sprinkler at performance level because the layout is delegated design. All true, and all exactly what Trellis was paid to produce. It lands on the table as an accusation of somebody, because nobody had written down what those models were before they were shipped.

Three more weeks go to that argument.

Month 5 — the decision that ends it

Vantry engages Northbound Digital for $96,000 to "produce the coordination model."

This is the fatal move, and it is worth being precise about why. Northbound is competent, and does exactly what it is hired to do: it models the design, from the 2D contract drawings, on the architect's origin. What it delivers in month 6 is a tidy LOD 300 representation of a set of drawings — no insulation, no trapeze hardware, no service clearances, no fittings anybody's shop actually builds — produced by a party that will never install anything.

Nobody who was going to build the building was in the model. That is not a coordination model. That is an expensive redraw.

Meanwhile the mechanical subcontractor has been detailing on paper since month 2 and released level-1 sheet metal in month 5. Fabrication has outrun the model. From here the model cannot change anything, because the thing it would have changed is already cut.

Six months of activity. Zero coordination.

Months 7–16 — the field, and the RFI curve

At a fully burdened $340 per RFI, processing the 258 coordination RFIs alone cost about $87,700 — before anybody answered one, and before anybody built anything because of one.

The field work those RFIs represent:

The medical gas line is the one people underestimate. Cutting into an installed medical gas system does not just cost pipe and brazing; under the health care facilities code framework it triggers re-verification and re-certification of that zone by an independent verifier — a test, a document, and a schedule you do not control. (Requirements depend on the code edition your jurisdiction has adopted and on the AHJ. Confirm before you price a field fix, never after.)

Substantial completion slipped 34 calendar days. Vantry claimed all 34 as compensable, on the theory that the architect's 16.2-day mean response exceeded the contractual review period. The settlement gave 21 calendar days of excusable, non-compensable time — an extension, no money — and assessed liquidated damages on the remaining 13. The causation problem was the one that sank Curtis Boone at Rivermont Elementary: you cannot recover for a bottleneck you manufactured, and Vantry could not separate its coordination RFIs from genuine design RFIs, because nobody categorized the log while it was being written.

The Accounting

What was spent to comply with the requirement:

What it returned:

The total:

Set that next to Northgate. Northgate's entire VDC program — two staff positions, software, scanning, layout robotics, training, and $369,000 of paid subcontractor detailing — cost $643,500 and had a mechanism. Stonebridge spent $799,400 and had none.

⚖️ The uncomfortable finding. The requirement itself created cost with no mechanism to create value — and worse than neutral, it displaced the method the team would otherwise have used. Vantry's subcontractors had coordinated four previous Meridian buildings with layered 2D overlays. Nobody ran overlays on Stonebridge, because the specification said model, everyone assumed somebody was modeling, and for six months everyone was right that somebody was and wrong about what.

The Four Decisions That Would Have Changed It

None of these is a technology decision. All four are made before anybody opens a file.

Read that total carefully. It is not "$358,000 to save $799,400" — some of those decisions carry their own returns, and some of that rework would have happened anyway. It is $358,000 that buys a mechanism, a way for effort to turn into avoided rework. The $242,000 Stonebridge spent bought no mechanism at all, which is why it bought nothing.

Decision 2 is the load-bearing one, and notice what kind of decision it is: procurement. The most important thing you will ever do about BIM happens in a scope sheet.

Analysis

The requirement is not the mechanism. A specification can compel activity. Only a plan, a budget, and a clause can compel value. Stonebridge had all of the requirement and none of the mechanism, and it got what that combination always produces: motion.

The disclaimer was correct, and correct was not enough. Trellis Engineering's models were precisely what Trellis was paid to make. Two years later the same firm's models fed Northgate's coordination without an argument — not because Trellis changed, but because Northgate's execution plan said, in advance and in writing, what those models were, who would add the missing layer, and who would pay for it. The same input produced a disaster and a success. The difference was a document.

Reverse-engineering is the predictable output of an unfunded requirement, not a moral failing. Vantry's subs were told to model and not paid to model, so they built the job the way they knew how and produced a model afterward for the file. Failure mode #5 in §35.10 is not about bad people. It is about an incentive a general contractor set and was then surprised by.

One hour against three weeks. The origin paragraph is the highest-return hour on the job, and the one nobody schedules because it looks like housekeeping.

And the owner got the wrong rung. Meridian paid $34,000 for a closeout model with geometry and no asset data — rung 2 on the §35.8 spectrum, where rungs 3 and 4 are the ones that pay. Hollis Wrenn's line in the lessons-learned file is the best sentence in the document:

"I have a model of a building I can already walk through. What I needed was the serial number of the thing on the roof."

Which is why Northgate looks the way it does. Pri Sethi wrote that file. Meridian's next project manual required the execution plan to carry an LOD matrix and an exchange schedule, required the contractor to carry a stated detailing allowance and disclose it, and specified the asset-data fields for turnover. Kestrel's $369,000 detailing line and Grace Lindqvist's nineteen-page BxP are the direct descendants of a job that went badly for somebody else.

Discussion Questions

1. Vantry complied with the letter of the specification at every step. Separate two questions and answer both: did Vantry breach the contract, and did Vantry manage badly? Then a third — who should have caught this, and at what moment?
2. The largest line in the "spent complying" table is Northbound's $96,000. Build the strongest case for hiring Northbound as Teddy understood the situation in month 5. Then say what she should have done instead, given that fabrication had already been released.
3. You are Vantry's PM in month 8 and can see the RFI curve turning. The model is useless, fabrication is out, the money is spent. Name three actions available to you at that moment and rank them.
4. Decision 2 costs roughly $220,000 out of a lump-sum bid, and carrying $220,000 nobody else carries is how you lose the job. Resolve that: how does a contractor price a real coordination program in a competitive lump-sum environment? (Compare how Northgate's CM-at-Risk structure changed the answer.)
5. Hollis Wrenn wanted a serial number and received a model. Whose job was it to know that, and when was it still cheap to fix?

Your Turn

You have been handed the project manual for a $19,000,000 public library and archive. Division 01 contains a nine-page BIM section that reads almost exactly like Stonebridge's. Bids are due in eighteen days. Your estimate carries $0 for coordination.

Produce two one-page documents:

1. A pre-bid RFI to the design team. Five questions, no more, each written so a one-line answer materially changes your price. At minimum resolve: the represented level of development of the design models by system; whether they are represented as suitable for construction coordination; whether the execution plan is required before the first pay application; whether trade detailing is intended to be a paid scope; and what asset data the closeout model must carry. Write them so that "coordinate per the contract documents" is not a possible answer.

2. An internal memo to your chief estimator recommending what to carry: a dollar figure with a build-up, the two or three model uses you would pursue and the ones you are declining, and one sentence on what you will do if the RFI comes back silent — plus the number you would carry in that case, which is not the same number.

Then write the single sentence for your bid qualification page. Under thirty words, and acceptable to a public owner without a protest.