Case Study 4-2 — The Percentage That Ate the Budget

Alder & Vine, Trestle Builders, and everyone named here are Tier-3 illustrative composites. The arithmetic is internally consistent and the failure pattern is common; the project is not a record of any single real job.

Setup

Nathan Poole runs a small restaurant group. Three locations, all successful, all leased spaces somebody else built out. For his fourth, Alder & Vine, he bought a 1920s brick warehouse on the east side of Rivermont — 9,800 square feet, twenty-two-foot ceilings, exposed timber trusses, and a loading dock he wanted to turn into a patio. Restaurant on the ground floor, private event space in the mezzanine.

He hired Trestle Builders, a good ten-person general contractor with real restaurant experience. Project manager Kayla Dubois had done four build-outs Nathan admired. He liked her, he trusted her, and — this matters — she was competent. Nothing that follows is a story about a bad contractor.

The design was maybe 40% done when Nathan wanted to start. His lease on the current commissary kitchen was ending, his investors wanted an opening date, and the architect was three months from a permit set. So Trestle proposed the honest structure for that situation: cost-plus.

The contract, in its entirety, was four pages. Its material terms:

Nathan read it, thought that seems simple, and signed. He is a smart person who has negotiated commercial leases. He had simply never seen the failure mode.

How I got involved. Nathan is a friend of a friend. He called me in month five and asked a question I have heard before in exactly that tone of voice: "Is this normal?"

What happens

Here is what Nathan sent me. It is his own spreadsheet, built from nine monthly invoices.

Nathan expected about $1,600,000. He paid $2,496,480 — an overrun of $896,480, or 56%. The job also ran nine months instead of seven.

Notice the shape of the curve. Through month four, cumulative spend was $939,680, which felt fine against a $1.6 million budget — 59% billed on a job Nathan believed was more than halfway done. Month five is where it broke, and Nathan did not see it until month five's invoice landed, because nothing in his contract required anyone to tell him where the job was going, only where it had been.

Where the $799,000 of cost growth came from

I spent two days with Kayla's job cost report. The original scope, honestly estimated, was about $1,430,000 of cost — which at 12% is $1,601,600, exactly the number in Nathan's cover email. The estimate was not wrong. Here is what happened to it:

Two of those five categories were genuinely nobody's fault. A 1920s warehouse has surprises, and $214,000 of them on a building like that is not scandalous. The other three — $501,000 — were decision failures, and every one of them was invisible to Nathan until after the money was spent.

The fee, which is the part that made him angry

Fee on the original scope: $1,430,000 × 12% = $171,600 Fee actually paid: $2,229,000 × 12% = $267,480 Difference: $95,880

Trestle Builders earned $95,880 more because the job went badly.

I want to be careful here, because the obvious conclusion is the wrong one. I looked at Kayla's records. There is no evidence anyone at Trestle inflated anything, padded anything, or slow-walked anything. The mechanism is not corruption. It is the absence of friction.

Every job produces dozens of small forks. The exposed brick can be repointed properly for $14,000 or patched for $6,000. The kitchen hood can go in this week with premium-time labor or in three weeks on straight time. The mezzanine railing detail the architect sketched can be built as drawn for $21,000 or value-engineered for $12,000 with a phone call and an argument. In every one of those forks, the cheaper path costs the contractor a hard conversation and pays it nothing. The expensive path is easy and pays it twelve cents on the dollar.

Nobody decides to waste money. The system just never pushes back.

💰 What Nathan actually bought with that fee structure

$95,880 of additional fee ÷ 9 months = $10,653 per month paid to the contractor as a consequence of the job going 56% over.

Nathan's contract paid a bonus for the outcome he most wanted to avoid.

The four controls that would have prevented most of it

Nathan asked me the only useful question: "What should I have done?" Here is what I told him, ranked by what each one would have saved on this job.

1. A fixed fee in dollars, not a percentage — saves $95,880

The single easiest fix. Negotiate the fee as a stated dollar amount at execution: $171,600. It does not move when cost moves. Trestle still earns exactly what it should for a seven-month job, and the moment cost starts growing, the contractor's interests align with Nathan's instead of against them. If the scope genuinely expands, the fee is adjusted by agreement in the change — visibly, on purpose, with a conversation.

2. A written definition of the cost of the work, plus audit rights — saves ~$46,000

"Cost of the work" appeared in Nathan's contract three times and was never defined. In two days of review I found roughly $46,000 of billed items that a normal cost-of-the-work schedule would have excluded or capped:

• A shared pickup truck billed at full rate while serving two jobs.
• Trestle's office manager's bookkeeping time, billed as project staff.
• A $4,800 tool purchase that stayed with Trestle at the end.
• Home-office overhead items that belong in the fee.

None of this was theft. Kayla billed what she thought was reimbursable because nobody had written down what wasn't. What you need: a schedule of reimbursable and non-reimbursable items, an agreed labor burden rate, an agreed equipment rate schedule, and the right to audit for a stated period after final payment.

3. A control budget by cost code with a monthly cost-to-complete forecast — saves ~$143,000

This is the big structural one. Nathan received nine invoices telling him what had been spent. He never once received a document telling him what the job was going to cost.

A control budget — the estimate broken into cost codes — plus a monthly forecast of cost to complete, not cost to date, would have shown the $1.6 million becoming $2.1 million somewhere in month three, while there was still scope left to trade. Of the $286,000 in verbal scope additions, my honest estimate is Nathan would have declined about half — roughly $143,000 — if he had seen each one priced before it was built.

He did not decline them because he never got the chance. He learned about them as line items on an invoice for work already installed. (Chapter 28 builds this document.)

4. A written authorization threshold — saves ~$79,000

One sentence: no cost above $10,000 outside the control budget is committed without the owner's written authorization.

That sentence catches the $118,000 of decide-build-reverse rework. Every one of those reversals started with somebody saying "let's just try it." My estimate is that two-thirds — about $79,000 — evaporates the moment a decision requires an email and a number.

The counterfactual

Now the honest part, which Nathan did not want to hear. Even with perfect controls, this was never a $1,600,000 job. A 1920s warehouse with unmapped structure, an undersized electrical service, and no sprinkler main was always going to cost more than $1.6 million once anybody opened the walls.

The controls would not have made the job cheap. They would have made it known — month by month, decision by decision, while Nathan still had the ability to trade the patio for the truss repair, or delay the mezzanine, or push the opening three weeks and save $97,000 in premium time.

Cost control is not primarily about spending less. It is about deciding on purpose.

Analysis

Cost-plus was the right structure and the wrong contract. At 40% design with a hard deadline, cost-plus was genuinely the honest answer — a lump sum would have been priced on assumptions nobody could defend, and a GMP at 40% design would have carried a contingency so large Nathan would have rejected it. Trestle did not steer Nathan into a bad structure. The failure was that a cost-plus contract requires four pages of controls, and Nathan signed four pages of nothing. The structure that transfers the most risk to the owner is the structure that demands the most owner discipline, and Nathan bought the risk without building the discipline.

The percentage fee is the part that is indefensible in hindsight. Every other failure here is a management failure that a diligent owner might have caught. The percentage fee is a design flaw written into the document — it guaranteed that the contractor's economics would never push back on cost. There is almost no situation where a percentage fee is the right answer for an owner, and if a contractor insists on one, that is information.

Where cost-plus should have ended. The permit set was complete in month three. That was the moment to convert to a GMP, and neither party raised it because neither party had a contractual obligation to. A well-drafted cost-plus agreement contains its own exit: the parties shall negotiate a guaranteed maximum price upon completion of construction documents, and the cost-plus arrangement shall govern only until such amendment is executed. Cost-plus is a bridge. Write the far end of the bridge into the contract, or you will still be standing on it in month nine.

What this looks like from Trestle's side. Kayla got a well-paid job and a client who will never call her again, will tell the story at dinner parties, and whose three restaurant-owner friends now know it. Trestle earned $267,480 and destroyed a referral source in a market where restaurant build-outs move almost entirely by referral. That is not a good trade, and the tragedy is that Kayla could have proposed every one of the four controls herself and been the hero of the story.

Discussion questions

1. Kayla Dubois did nothing dishonest, and the job still went 56% over with the contractor earning $95,880 more for it. What, specifically, was her professional obligation to Nathan under this contract — and is "the contract didn't require it" an adequate answer for a licensed contractor?

2. Of the $799,000 in cost growth, $214,000 came from genuine existing-conditions discoveries in a hundred-year-old building. Under a lump-sum contract with a standard differing site conditions clause, who would have paid that $214,000 — and would the total have been higher or lower? Show your reasoning.

3. Rank the four controls by ease of negotiation rather than dollars saved. Which one would a contractor resist hardest, and why? Which would a contractor actually welcome?

4. Nathan's budget lived in a cover email, not the contract. Draft the sentence you would have put in the agreement to give that $1,600,000 contractual force without turning the contract into a GMP. What would the contractor's objection be?

5. The conversion trigger — "negotiate a GMP upon completion of construction documents" — sounds obviously correct. What goes wrong if the parties cannot agree on the GMP number in month three? What does the contract need to say about that?

Your turn

You are advising a nonprofit that must renovate a fire-damaged community kitchen in eight weeks. Scope is genuinely unknown until demolition. Cost-plus is unavoidable. The board has $400,000 and cannot spend a dollar more.

Draft the six contract provisions you would insist on, in plain English, each in one or two sentences. At least one must address the fee structure, one must address the definition of cost, one must give the board visibility before money is spent rather than after, and one must define what happens when the running forecast crosses $400,000.

Then answer the hard question: what does your contract do when demolition reveals $520,000 of work? Cost-plus has no cap, the board has no more money, and the kitchen has to open. Write the two sentences that govern that moment. If you cannot, you have found the reason cost-plus without a conversion mechanism is dangerous — and you have found the exact conversation the nonprofit's board needs to have before it signs, not in week six.