Case Study 34-1 — Four Quarters of Kestrel's WIP, and the Reporting Change That Made the Numbers Look Worse

Kestrel Construction Group, its jobs, its people, its surety, and every figure below are Tier-3 illustrative composites. The arithmetic is real; the company is not.

Setup

The date. Late March, Year 2. Kestrel's audited Year 1 financial statements were issued nine days ago. The annual surety review is next Tuesday.

The room. Owen Baptiste's office. Owen, chief financial officer. Nadia Haddad, vice president of operations. Me, because Owen had shown me the WIP schedule in February and Nadia decided that made me the person who had to explain it to the other project managers.

What is on the table. Five printed work-in-progress schedules — the four quarter-ends of Year 1 and the January close of Year 2 — laid out side by side, which is not how anybody normally reads them. A WIP schedule is usually consumed one page at a time, monthly, and then filed. Owen had never laid five of them on a table before either.

"We are going to do something the auditors just did to us," he said. "We are going to read across."

What happens

The company, five closes in a row

Owen's summary page. Six columns, eight rows, and it took him under a minute to build from the underlying schedules.

Look at row four. Nothing in any individual quarter is alarming: 5.20, then 5.05, then 4.88, then 4.83, then 4.75. The largest single step is seventeen hundredths of a percentage point. Any one of those movements is explainable, and each one was explained, in a meeting, by a person with a reason.

Read across and it is a slope. Kestrel's contract portfolio grew by $50,434,000 over five closes and got less profitable every single time it was measured.

Now row eight. Working capital fell $1,847,920 across the same window while contract value rose. That is Chapter 32's threshold concept — cash flow is not profit — arriving at the company level, and it is what §34.8 Step 1 looks like from inside a company that is not failing.

Which rows did it

Owen's second page: estimated gross profit, by job, by quarter.

Northgate's small gain is not skill. It is arithmetic: the change orders approved through January added $600,000 of cost and $24,000 of fee at the contract's 4.0 percent rate, which is why the margin percentage never moved off 3.80 in any period. A GMP job that stays at its fee percentage is a job whose forecast is holding.

Fair Oaks moved twice, and both moves have names — which is the test §34.5 applies to every gain. The Q4 improvement of $70,000 is the release of the remaining construction contingency once the tie-in shutdown passed without the risk it was holding. The January improvement of $80,000 is a change order that settled favorably: the contract amount rose $310,000 against $230,000 of added cost. Neither is "the job is going well." Both are events, on dates, with documents.

Rivermont Elementary #12 you already know from §34.4 — five consecutive periods of decline, cumulative fade of $565,000 measured from its first quarter, $385,000 of it inside the window on this page.

The row nobody had been looking at is Brenner Street.

The second finding

Brenner Street Parking Structure has been reported at 98 percent complete for three consecutive quarter-ends and again at the January close. Here is what was actually happening underneath that number.

Read the second and third columns together. The project manager raised his estimated cost at completion by exactly enough, each period, to keep the ratio at 98 percent. He was not lying about anything. Each individual increase was small, real, and defensible: a punch crew, a warranty callback, a light-pole base that had to be redone, extended general conditions for a month.

The consequences are exact:

• Cost incurred between Q2, Year 1 and January 31, Year 2: $12,026,560 − $11,809,000 = $217,560
• Revenue recognized in the same span: $12,544,000 − $12,544,000 = $0
• Gross profit earned, same span: $735,000 − $517,440 = $(217,560)

Every dollar Brenner Street spent over seven months went straight to gross profit reduction, because the percent-complete column never moved. And the remaining cost to complete — the thing the forecast is supposed to be about — went up: $12,050,000 − $11,809,000 = $241,000` in Q2, against `$12,272,000 − $12,026,560 = $245,440 in January. Seven months of work and the job had more cost left than when it started.

That is the §34.5 diagnostic — a job at 98 percent complete for three consecutive quarters — and it is invisible in the percent-complete column by construction. It is only visible in the estimated gross profit column, read across.

The question Nadia asked

Nadia put her finger on Rivermont's Q1, Year 1 cell: $920,000, down $180,000.

"Which quarter did we first know?"

Owen said the honest thing, which is that "know" is doing a lot of work in that sentence, and that the audit had asked him the same question in a less friendly tone. Then he answered it.

Q1, Year 1 — March 31. Not the year-end audit, not Q3, not the January close. At 34 percent complete, a $22,400,000 hard-bid public school gave back $180,000 and its margin fell from 4.91 to 4.11 percent. A hard-bid school's margin does not fall eight tenths of a point at one-third complete because of one masonry crew. A move of that size, that early, is a re-based estimate wearing the costume of a variance. The information was on the page, in the first quarter, in the first period the number moved.

Curtis Boone said it was a masonry productivity issue he would recover in the next phase. He was not lying either. He believed it. Nobody asked him to write down what "recover" would consist of, or by when, or what would have to be true.

The annual review

Tuesday. Ellis Tobin, Kestrel's surety bond producer, and Marta Ibáñez, the underwriter for Cordillera Surety, in Kestrel's conference room with three years of WIP schedules and the audited statements. Nadia was in the room, which was new; historically Owen went alone.

Marta spent nine minutes on the balance sheet — working capital at $20,332,080, current ratio 1.70, debt to net worth 1.24, aggregate program at 7.4 times working capital — and said it was fine, which it is.

Then she spent fifty minutes on one page.

Marta Ibáñez: "Your ratios are not my problem. Your program is 7.4 times working capital. I could write more. What I want to talk about is that I have five of your WIP schedules in front of me and your company margin goes down every time you measure it. And on Rivermont, I have four consecutive quarters where somebody tells me the same story and the number gets worse anyway."

Owen Baptiste: "The cumulative fade on that job is $565,000."

Marta Ibáñez: "I am not worried about the $565,000. On a $410,000,000 contractor, $565,000 is a rounding error and you can absorb it. I am worried that it took five quarters, and that the explanation was the same each time, and that your revenue-to-working-capital is 20.2 — which means when you finally are wrong about something large, you will not have the liquidity to be wrong slowly. Show me one quarter where somebody in your company tells me something I do not want to hear before I find it, and we will talk about the single-project limit."

Cordillera held the program at $150,000,000 aggregate and $60,000,000 single project — no reduction — with four conditions:

1. Quarterly WIP schedules with a same-job fade column and a written explanation for every movement over $100,000.
2. Cost-to-complete detail on every contract over $10,000,000.
3. A mid-year meeting with the vice president of operations present, not only the CFO.
4. No increase to the single-project limit until two consecutive quarters of stable forecasting.

Condition four is the expensive one. Owen has been working toward a $75,000,000 single-project limit for two years, because the difference between $60,000,000 and $75,000,000 is an entire class of work Kestrel currently watches go by. That conversation is now nine months away at best, and it was postponed by a $22,400,000 school on the other side of the metro.

What Nadia changed

Nadia did not change the forecasting method. She changed how project managers report, effective with the February close, and she wrote it on one page.

Then the numbers got worse

The first full quarter under the new rules was Q1, Year 2. Here is what Kestrel reported, against the four quarters before it.

Kestrel reported more fade in one quarter than in the previous four combined — `$206,000 + $238,000 + $261,000 + $94,000 = $799,000`, against $1,142,000.

Two project managers thought they were being set up. One of them said so.

They were not. Of the $1,142,000, only about a third was news that arose during the quarter. The rest was old — held estimates on seven different jobs that people had been carrying, in good faith, waiting to see whether they could absorb it. Owen wrote exactly that in the transmittal to Cordillera, before Marta could ask:

"Q1 fade of $1,142,000 exceeds the full prior year. Approximately two thirds of it is the release of previously unreported estimates following the February change to our reporting requirements, not new deterioration. We expect elevated fade for one to two more quarters as the backlog of held estimates clears, and we expect the fade run rate to fall below the Year 1 average thereafter. We would rather explain this now than have you find it in the Year 2 audit."

That paragraph is the whole case study. A surety that reads it understands what happened. A surety that discovers the same $1,142,000 for itself, unannounced, in an audit, does not.

Analysis

The mechanism, not the moral. Rivermont Elementary #12 did not fade $565,000 because Curtis Boone is a bad project manager. He is a good one with a bad model. It faded in five increments because five increments was the size the organization would accept without a fight, and Curtis — like every project manager who has ever worked anywhere — released bad news at the rate the organization could tolerate rather than the rate he discovered it. That release rate is not a personality trait. It is a policy, set by whoever runs the operations meeting.

Why the change made the numbers worse first, and why that was the point. A company that has been under-reporting fade is carrying an inventory of it. When you change the reporting rule, the inventory clears in one or two periods and the reported number spikes. If management panics at the spike, the rule silently reverses and the inventory rebuilds — permanently, because the staff has now learned the lesson twice. Nadia's insulation against that was to predict the spike out loud, in advance, to Cordillera and to the project managers, so that when it arrived it confirmed the diagnosis instead of contradicting the policy.

The honest observation, stated plainly: a company that punishes the first honest fade report is manufacturing its own surprises. Punish a $200,000 fade at 30 percent complete and you have not prevented $200,000 of fade. You have moved it to 80 percent complete, where it will be four times larger, nothing can be done about it, and the surety will find it before you do.

What the surety was actually buying. Marta Ibáñez did not reduce the program over $565,000, and she would not have reduced it over $2,000,000. What she was pricing was whether Kestrel knows what is happening inside itself — the Character leg of the three C's in §34.7, which is not sentimentality and is not measurable from any of the ratios she reviewed in nine minutes. Condition four is the price of not knowing, denominated in single-project limit.

What it cost, in a number. One $22,400,000 school, running about 2.5 percentage points below the company's average margin, delayed a bonding-capacity increase that would have opened a class of work worth many multiples of the job itself. That is the sentence at the end of §34.9: your forecast is not a project document.

Discussion Questions

1. Owen's summary page shows company gross profit percentage falling in every one of five consecutive closes, and every individual step is under two tenths of a point. Design the specific report or trigger you would build so that this pattern raises a flag automatically, without requiring anyone to lay five schedules on a table. What is the threshold, who receives it, and what must happen when it fires?

2. Brenner Street's project manager raised his estimated cost at completion honestly, every period, and the percent-complete column never moved. He broke no rule that existed. Whose failure is that, and what is the minimum change to the reporting system that catches it? Compare your answer to Nadia's change #4.

3. Nadia's change #5 — early fade is a management item, late fade is a performance item — is easy to write and hard to run. Describe two ways it could be implemented badly, and the specific behavior each bad implementation would produce.

4. Marta Ibáñez said the $565,000 did not worry her. Was she right, given that Kestrel's Year 1 net income was $9,430,000 and one $22,400,000 job produced 6 percent of a year's profit in fade? Argue both sides, then say which position you would take as the underwriter and why.

5. Owen's transmittal disclosed the $1,142,000 spike before he was asked. Some CFOs would call that unnecessary — the number would have been in the quarterly package anyway. What does the volunteering buy that the disclosure alone does not, and when, if ever, is it the wrong move?

Your Turn

You have Kestrel's five-close summary page above and the estimated-gross-profit-by-job table. Produce three artifacts, on two pages total.

1. The one-page WIP cover memo that should have gone to Nadia at the Q2, Year 1 close — the second quarter, not the fifth. It must name the jobs that moved, the dollar amounts, the causes as you would have demanded them, and one recommendation that is an action rather than an observation. You are writing it with only the Q1 and Q2 columns available; do not use hindsight from the later columns.

2. The trigger specification for change #4 (the two-month rule). Define exactly what counts as "the same percent complete," who performs the verification, what they must produce, and what happens if the verification disagrees with the report. Fewer than 200 words, in language a project manager can follow without an interpretation.

3. The two sentences you would give Marta Ibáñez at next year's review, describing Kestrel's fade performance for Year 2. Write them as though the numbers came out well, then write the alternate pair for the case where they did not. Notice which pair is harder to write, and notice that the hard pair is the one that decides the single-project limit.