Case Study 28-2 — Seven Months at Budget, Then $1,900,000: The Ashcombe Ridge Fade

All companies, people, and projects in this case study are illustrative composites.

Setup

The project. The Ashcombe Ridge Hotel and Conference Center — 214 keys, seven stories, plus 18,000 square feet of meeting and banquet space, for a regional hospitality developer. $36,000,000, negotiated lump sum, 18-month schedule (548 calendar days). Kestrel Construction Group, three years before Northgate started. Kestrel self-performed the cast-in-place concrete frame and the rough carpentry.

The numbers at award.

The project manager. Dominic Fiore. Twelve years at Kestrel at that point, well regarded, and — I want this on the record before anything else — not dishonest. He is still at Kestrel. He runs our largest jobs. He is, today, the best forecaster in the company, and the reason why is this case study.

What the reports said. For seven consecutive monthly reporting periods, Ashcombe Ridge forecast a total estimated cost of $33,840,000 and an estimated gross profit of $2,160,000. Not approximately. Exactly. The same two numbers, seven times.

What happened in month eight. A $1,900,000 fade. Estimated gross profit went from $2,160,000 to $260,000 — from a 6.00 percent margin to 0.72 percent — in one reporting period, on a job that was 47 percent complete.

What happens

The march, reconstructed

After it landed, Owen Baptiste asked for something nobody had asked for at Kestrel before: a reconstruction. Not "why did it fade," but "what would an honest report have said in each of the eight months, given only what was knowable at the time?" Two of us spent a week on it with the daily reports, the quantity logs, the buyout file, and the subcontract correspondence.

Here is the answer. Cost to date is a fact and is identical in both columns; the only thing that moves is the forecast.

Look at the honest column and then at the reported one. The fade was not an event. It was a continuous, monotonic, month-after-month decline that a working forecast would have printed every single period. Nothing surprising happened in month eight. What happened in month eight is that an auditor asked for the cost to complete on every code and would not accept the budget as an answer.

Where the $1,900,000 actually came from

Four components, and every one of them is a failure mode this chapter names by title.

Note what is not on that list: fraud, theft, a catastrophe, a differing site condition, an owner who would not pay. Four ordinary things, each individually small enough to be explained away, and none of them written down.

The four decisions that were lost

This is the part that matters, and it is the part nobody computes.

Decision 1 — month 3: change the concrete production system. At month 3 the package was 27 percent placed with three consecutive periods pointing the same way. By the chapter's own test, the trend was believable. The forecast overrun was $310,000 and 73 percent of the quantity was still ahead of the crew — roughly $540,000 of the eventual $740,000 still preventable. There is a stronger point: the rate did not hold, it degraded — 8 percent, then 10.5, then 16.1, then 19.2. A production system nobody is measuring gets worse, because nothing is pushing back on it. On Northgate, two specific changes by a superintendent recovered about a quarter of the still-manageable exposure. Applying that ratio conservatively, month 3 was worth $135,000; if the intervention had also stopped the degradation, closer to $400,000.

Decision 2 — month 4: change the buyout posture. At month 4, budget minus committed on the two largest unbought packages — finishes at $5,240,000 and HVAC at $3,780,000 — was $9,020,000 of scope with no price behind it. They were bought in months 5 and 7, at $290,000 and $330,000 over budget. Some of that gap was scope and would have appeared whenever it was bought. But on interior packages moving at roughly half a point a month, the extra three months on HVAC alone was worth about $57,000 and the extra month on finishes about $26,000 — call it $83,000 of pure escalation. The larger loss is not money. A project manager who knew his job was already down $550,000 would have bought differently: not hunting the last two percent, but buying certainty and buying it now.

Decision 3 — month 5: convert the directed work while the record was contemporaneous. By month 5, $145,000 of directed work had been performed with no signed change. The owner's representative was still on the job. The crews that did the work were still on site. The daily reports were three months old, not nine. Kestrel finally submitted in month 9 and recovered $96,000 of $355,000 — twenty-seven cents on the dollar, for work it unquestionably performed. A contemporaneous submission on directed work with clean records recovers seventy to eighty-five percent. At 75 percent that is $266,250, so the lost decision is worth roughly $170,000. This is Chapter 31's entire thesis arriving through the cost report instead of the change log: the price of a change is set by what you can document, not by what it cost you.

Decision 4 — month 6: intervene on the failing subcontractor. The waterproofing sub's crew went from nine to four. Two of its suppliers called Kestrel about unpaid invoices. Its month-6 pay application asked for materials stored off site. At month 6 the available moves cost almost nothing: joint checks, a funded recovery plan, or a consensual takeover with the crew intact. At month 8 the sub defaulted, the balance was bought on an emergency basis, and the enclosure lost 14 calendar days. Half the $185,000 was avoidable — call it $90,000 to $185,000, plus the days.

The four together: roughly $478,000 to $838,000 of the $1,900,000 — a quarter to a half.

And now the honest half of the sentence, because I do not want to sell you a fantasy: the rest of it was not recoverable. The concrete frame was probably mis-estimated at bid, and no amount of early knowledge un-bids a job. But there is a category difference between unrecoverable and undisclosable. The company could not have saved every dollar. It could have known about every dollar, seven months earlier, and it did not.

What it did upstream

Under percentage-of-completion accounting — the standard for construction — the number a project manager writes in the cost-to-complete column is the number the company reports as profit. Watch it.

At month 7, cost to date was $14,551,200. Under the reported forecast:

Percent complete = $14,551,200 ÷ $33,840,000 = 43.00%
Revenue earned   = $36,000,000 × 43.00%      = $15,480,000
Gross profit earned = $15,480,000 − $14,551,200 = $928,800

Under the honest forecast, with the same cost to date:

Percent complete = $14,551,200 ÷ $35,570,000 = 40.91%
Revenue earned   = $36,000,000 × 40.91%      = $14,727,107
Gross profit earned = $14,727,107 − $14,551,200 = $175,907

Kestrel's books carried $752,893 of gross profit that did not exist — and, because cost to date is a fact and identical in both, exactly $752,893 of revenue that did not exist either. That is not a metaphor. It was on a financial statement, and it had been growing for seven months.

When the correction landed in month 8, cumulative gross profit earned on the job fell from $928,800 to $123,089. In a month during which Ashcombe Ridge performed $2,368,800 of work, it contributed negative $805,711 of gross profit to the company's income statement. A profitable job produced a loss on a quarter's results — not because anything went wrong that month, but because seven months of reality arrived at once.

For scale, at Kestrel's present size — roughly $410,000,000 of revenue and about $9,430,000 of net income — $1,900,000 of fade is 20 percent of a full year's net income, and the $752,893 that had to come back is roughly a third of a normal quarter's. The company was smaller then. The proportion was worse.

Then the surety. Chapter 34 works the underwriting arithmetic properly; here is the short version and it is not primarily arithmetic. The capacity formulas moved a little, because working capital and net worth moved a little. What moved a lot was character — one of the three things a surety underwrites, and the one you cannot rebuild in a quarter. The underwriter now knew that Kestrel's forecasts had been wrong, in the same direction, for seven consecutive periods, on one of the company's three largest contracts. A work-in-progress schedule that reports 6.00 percent for seven months and then 0.72 percent is not a document anyone can price risk from.

The surety held the aggregate program. It froze the single-project limit for two renewal cycles, required quarterly rather than annual work-in-progress schedules with cost-to-complete detail on every contract over $10,000,000, and asked in writing who reviewed project forecasts and with what authority. Kestrel passed on two pursuits inside that window because the limit could not be raised in time.

The fade did not cost Kestrel $1,900,000. It cost Kestrel $1,900,000 and two years of not being allowed to compete for the work it wanted.

Analysis: be fair to Dominic

It would be comfortable to end here with a paragraph about a project manager who should have known better. That reading is wrong, and worse, it is useless — it teaches you to hire differently instead of to manage differently.

Four things the company built, deliberately, that produced this behavior:

1. The Monday operations board — as it was then. The board displayed every job's current margin against its bid margin, and any project manager whose number dropped explained it standing up in front of every other project manager. The forecast had been converted from a management tool into a public performance. People do not tell the truth in performances; they tell a story. (The board still exists at Kestrel. What it measures does not.)
2. The bonus math. Project-manager incentive compensation was computed on final margin against bid margin. An honest fade in month 3 reduced Dominic's own compensation nine months before anybody could know whether the recovery would work. The company had priced honesty and then charged the project manager for it.
3. No controls function. Ashcombe Ridge had one project accountant split across three jobs and no project controls person at all. Dominic personally coded accounts payable, produced the pay application, forecast four hundred cost codes, and ran a hotel. Compare Northgate: Lorena Vasquez full time on the job and Wei Chen on controls. That is not a coincidence. It is the fix, paid for.
4. The staffing plan was cut at buyout. The job was priced with an assistant project manager and a field engineer in general conditions. Both were deleted in the final review to help hit the number — $186,000 removed from a $2,192,000 general-conditions estimate. The person who would have gone back and re-forecast four concrete codes is the person who was cut to win the job. $186,000 saved; $1,900,000 found eight months later.

And Dominic was not lying. Every forecast he wrote he half-believed, and he had a specific, plausible story for each one — the rain in month two, the crew split in month three, the rhythm they were about to find in month four. Each story was partly true. Collectively they cost the job, and they were exactly the trap in §28.6: the most expensive psychological error in cost control is assuming the overrun was a one-time event.

What the company changed

Every one of these is now standing practice at Kestrel, and you have already met most of them in this chapter:

1. Staff the controls function. One project accountant per major job; a project controls person on anything over $25,000,000. On Ashcombe Ridge that was the $186,000 that got cut.
2. A method column on every line, and a count of the lines forecast at budget. If most lines are at budget, the report is unforecast and the review does not happen that month.
3. An accrual log with Type 2 items listed separately and initialed, and a cut-off memo to every subcontractor and supplier by the 20th.
4. Separate the disclosure meeting from the forensics. The first question about a fade is "what are we changing, and when do we know if it worked?" — never "how did this happen?" The forensics belong at closeout (Chapter 40), where they teach something, instead of at the disclosure, where they only teach people not to disclose.
5. Change what the operations board measures, and pay on the same thing. Fade and gain against the month-3 forecast, not against the bid margin, on the board and in the bonus formula. A project manager who forecasts four percent in month three and delivers four percent has done excellent work, and should be paid and displayed as if. The board is not the problem; the metric was.
6. A quarterly re-forecast reviewed by somebody who is not the project manager, with the superintendent in the room. A cost review without the superintendent is a fiction review.

Nadia Haddad was a project executive on Ashcombe Ridge. If you have wondered why she answers a bad forecast with "what are we changing and when do we know if it worked" instead of "how did you let this happen" — that is where it came from, and it cost the company two years of bonding capacity to learn.

The job finished at $356,000 of gross profit, a 0.99 percent margin, after Kestrel recovered $96,000 of the directed work. It never lost money. That is the most dangerous fact in this case study, because a job that never loses money is a job nobody investigates — and the honesty of a company's cost reports is a property of its culture, not of its people.

Discussion questions

1. Ashcombe Ridge reported exactly $2,160,000 of estimated gross profit for seven consecutive months. Design a single automated exception that a controller could run on a WIP schedule to catch that pattern across a whole company. What is it, what threshold would you set, and what false positives would it generate?
2. Of the four lost decisions, only one (the concrete production system) is a classic "cost control" move. The other three are a purchasing decision, a documentation decision, and a subcontract decision. What does that tell you about who needs to read a cost report?
3. The company saved $186,000 by deleting an assistant project manager and a field engineer at buyout. Build the argument you would make to an executive who wants to make that cut on your next job — in dollars, not in principle. Then build the strongest honest argument for the cut, because sometimes it is right.
4. The bonus was paid on margin against bid margin. Write the replacement formula. Be specific about the baseline month, what counts as fade, and how you prevent a project manager from simply sandbagging the month-3 forecast to guarantee a gain.
5. At month 7 the company's books carried $752,893 of gross profit that did not exist, and the same amount of revenue that did not exist. Who relied on that number, in what order, and what did each of them do differently because of it?

Your turn

Reconstruct the month-4 report Dominic should have written, and the conversation that goes with it.

Using the table above, produce three things:

1. The forecast. Total estimated cost $34,390,000, estimated gross profit $1,610,000, 4.47 percent — with a one-line variance narrative for each of the three components that had become knowable by month 4 (concrete labor, the two unaccrued directives, and the $9,020,000 of unbought scope). Each line names the method used and, where the method is judgment, its assumption.
2. The plan. One specific, mechanical, dated change to the concrete production system, and one to the buyout sequence. Not "improve productivity." A named change, with a date, that next month's report can be tested against.
3. The three sentences you say out loud to a project executive who has been told for four months that this job makes six percent. Write them so that the executive leaves the room knowing the true number, believing you have a plan, and understanding what she will be able to check in thirty days.

Then do the harder version: write the same three sentences for a company whose Monday operations board works the way Ashcombe Ridge's did. If you cannot make them work — if every honest version of the sentence gets you punished — you have located the actual problem, and it is not on the cost report.