Case Study 30-1 — Four Numbers on One Sheet of Paper: Northgate at Month Eleven

The projects, companies, and people in this case study are illustrative composites created for teaching. The arithmetic is real; the names are not.

Setup

Project: Northgate Outpatient Pavilion — 132,000 SF, four stories, structural steel frame, outpatient clinics, imaging suite, and an ambulatory surgery suite. Owner: Meridian Health System. Owner's representative: Priyanka "Pri" Sethi. Contractor: Kestrel Construction Group. Delivery: CM at Risk, guaranteed maximum price $47,500,000, of which the cost of the work is $40,000,000. Contract time 565 calendar days. Liquidated damages $5,500 per calendar day; extended general conditions $5,150 per calendar day; total daily exposure $10,650. Kestrel team: Ray Alvarez (project manager), Wei Chen (project controls), Lorena Vasquez (project accountant), Marguerite "Margo" Deacon (general superintendent), Dani Okonkwo (field engineer), Nadia Haddad (VP of Operations).

Data date: January 31, Year 2 — day 334 of 565, 59.1 percent of contract time elapsed.

Two documents were on Nadia's desk. Lorena's cost report said the job was $118,000 under budget. Wei's schedule update forecast substantial completion on September 27, Year 2 — nine calendar days past the contract date. Nadia asked whether we were ahead or behind, and I could not answer her, because the two documents were denominated in different units and neither one knew how much building we had actually built.

Wei had a third sheet of paper.

What Happened

The four numbers

CV = EV − AC = $24,970,000 − $25,642,000 = −$672,000 SV = EV − PV = $24,970,000 − $25,760,000 = −$790,000 CPI = $24,970,000 ÷ $25,642,000 = 0.974 SPI = $24,970,000 ÷ $25,760,000 = 0.969

"Read the three percentages in order," Wei said. "Time elapsed 59.1. Work earned 62.4. Money spent 64.1. You are producing slightly faster than the calendar and spending faster than you are producing, and the whole problem lives in the gap between the second number and the third."

Lorena's $118,000 was PV − AC. It compared what we spent to what we planned to spend, and it never asked how much work we got. We were not under budget. We were behind, and being behind looks exactly like being under budget until the month you catch up.

Three packages the aggregate was hiding

The project CPI of 0.974 is an average, and averages are where problems go to hide. At the package level the report exposed three things, and only one of them was a surprise.

Curtain wall — CPI 0.857, 54 percent complete. EV $2,662,000 against AC $3,105,000 on a $4,930,000 budget. Extrapolated at current performance the package forecasts $5,750,426 — an $820,426 overrun in one line item, which is 122 percent of the entire $1,320,000 construction contingency. The cause was not mysterious. The deck-edge geometry moved during the steel acceleration; the unitized panels are meeting embeds that are not where the model said they were; every unit is being field-shimmed and, on the north elevation, field-modified. This is the delayed bill for the acceleration, arriving five months later in a different cost code.

Framing and drywall — CPI 0.920, 15 percent complete. A cost variance of only $45,000. Everybody in the room wanted to ignore it because it was early. It is the most actionable line on the page, because it is the only trouble we found early enough to fix cheaply:

TCPI to budget = (3,450,000 − 517,000) ÷ (3,450,000 − 562,000) = 2,933,000 ÷ 2,888,000 = 1.016

The remaining 85 percent must run 1.6 percent better than budget. That is achievable. The curtain wall's equivalent is not.

Mechanical — CPI 0.997, SPI 1.000. Perfect, and worthless. EV was Cardinal Mechanical's billed percentage on their schedule of values; AC was what we approved for payment. Two numbers from the same document produce a CPI of 1.000 by construction. The 0.997 is a rounding artifact from a small backcharge. That line was not a measurement. It was a restatement of Sofia Marchetti's pay application.

The spread, and the test that closed the argument

Wei worked all four estimates at completion on the same four inputs.

$892,812 of spread, on the same day, on the same four inputs, with nobody lying.

I wanted #2. I want you to understand that I did not choose it cynically. I looked at $40,672,000 and it felt like the reasonable number, and then I went looking for the reason — which is the wrong order and feels identical from the inside.

Wei ran the credibility test.

TCPI to EAC #2 = (BAC − EV) ÷ (EAC − AC) = $15,030,000 ÷ ($40,672,000 − $25,642,000) = $15,030,000 ÷ $15,030,000 = 1.000

"Your forecast requires the remaining fifteen million dollars of work to run at exactly 1.000," Wei said. "In eleven months this job has posted a cumulative CPI of 0.982 at its best, in month four, and a period CPI of 0.987 at its best, also in month four. Eleven months, eleven attempts, never once at 1.000. What changed?"

I did not have an answer, and there is only one correct thing to do when you do not have an answer, which is to say so out loud.

Nadia's rule came out of that meeting and it is now Kestrel's rule: report #1 as the baseline forecast and #4 as the checked forecast; use #2 only when you can name the event, show it is closed, and prove the remaining work is unaffected — in one written sentence. If you cannot write the sentence, you do not get to use the formula.

The reason the rule exists is on this table, which Wei put in the appendix of the month-11 report and which I have never forgotten. Kestrel's fee is $1,804,800 and unused contingency splits 75 percent owner / 25 percent Kestrel, so the buffer available against a cost-of-work overrun is $1,320,000 contingency + $265,200 remaining escalation = $1,585,200.

The choice of formula was worth $223,203 to Kestrel — top row to bottom row. Which is exactly why the person whose job performance depends on the answer should not be the person who picks the formula, and why Nadia asks Wei for the EAC and not me.

What got done about it

Four things, and only one of them was a report.

1. Wei went and counted the mechanical package. In the second week of February, Wei and Dani Okonkwo spent a day and a half in the building with Cardinal's own submittal log and a tape. Not to catch Sofia Marchetti at anything — Cardinal is a good subcontractor — but because a CPI of 1.000 by construction is not a measurement and the only fix is to go look.

$275,400 of overstated earned value. Not fraud — the ordinary optimism of a percentage estimated by the person who benefits from it, spread across six lines, none of which is outrageous on its own. The controls line is the instructive one: $7,000 billed against a scope with nothing installed.

The correction was booked in month 12, not backdated, and it cost real money on both sides of the ledger. Kestrel had approved and paid for that work less retention; at 5 percent retention on current billings, roughly $261,630 of it was already out the door. And carried into the indices, the $275,400 moves the month-11 CPI from 0.974 to 0.963, the CPI-based EAC from $41,076,492 to $41,534,587, and the TCPI to hold budget from 1.047 to 1.066.

2. Curtain wall was contained, not recovered. At 54 percent complete with the cause still present, the package cannot be fixed; it can only be stopped from getting worse. Grace Lindqvist ran a laser scan of the remaining deck edge so the shims could be fabricated in the shop instead of improvised in the air, and Caldwell Structural issued a revised embed detail for the north elevation. The arithmetic on that effort:

Remaining budgeted work = $4,930,000 − $2,662,000 = $2,268,000 At the current 0.857: remaining cost $2,645,426, EAC $5,750,426 At 0.920: $2,268,000 ÷ 0.920 = $2,465,217, EAC $5,570,217 Value of containment: $180,209

3. Framing and drywall got the attention nobody thought it deserved. Margo changed the layout method to model-based layout off Grace's coordinated model, moved material staging from the ground to floor-level drops, and added a second lead carpenter. Lifting the remaining $2,933,000 from 0.920 to 0.980 pulls the EAC from $3,750,290 to `$562,000 + ($2,933,000 ÷ 0.980) = $3,554,857` — $195,433 of forecast improvement on the cheapest problem on the page.

4. The report changed shape. From month 12, every subcontract package carries a physically verified quantity before its billed percentage is accepted as EV, and every package under 20 percent complete is carried at budget or bottom-up rather than extrapolated. That second rule is why the disciplined sum of package EACs ($40,999,679) and the project rollup ($41,076,492) landed $76,813 apart on a $41 million forecast. Two methods computed different ways converging is evidence. One method by itself is an opinion.

Analysis

The instrument did not find anything new. It found things earlier. Every problem in that report existed before Wei wrote it. What the report did was convert them from opinions into numbers with dates on them, and put them in front of somebody who could spend money.

The CPI never got better and never was going to. Look at the S-curve in §30.7: month 3, 0.978; month 11, 0.974. Between those two readings sat an acceleration, a change of superintendent on the enclosure, and eight months of effort. Nothing about reaching the 60 percent mark improves your drawings, your subcontract prices, or your site access. The honest reading of an early bad CPI is not "it is early, it will recover." It is: this is probably the number, and I found out early enough to do something.

The three packages are three different management problems wearing the same clothes. Curtain wall is a discovery — a consequence of a decision made in a different cost code five months earlier, now unavoidable, worth $180,209 of containment. Framing and drywall is an opportunity — small, early, and fixable, worth $195,433. Mechanical is a measurement failure — the number was never a measurement at all, and finding that out was worth $275,400 of correction plus the process change that keeps it from recurring. Sorting your report into those three categories is more useful than sorting it by variance.

Steel is the only line on the page that is a decision. Its $168,000 cost variance is exactly what the acceleration cost: a second erection crew, premium time, and a resequenced enclosure, bought deliberately in Nadia's office with the arithmetic on the table. Everything else in the report is something we found out. That distinction matters, because a decision has an owner and a rationale and a date, and a discovery has none of those until you give it some.

And a number that appears nowhere on any of these tables: the nine-day slip carries 9 CD × $10,650/CD = $95,850 of extended general conditions and liquidated damages. BAC is the budget for the work; it has no line for being late. The earned-value report answered Nadia's cost question completely and answered her schedule question not at all. The forecast completion date of September 27, Year 2 came from Wei's CPM update — a forward pass, a backward pass, and total float on the controlling path — and it is the only instrument on the job that produces the number nine. Read them side by side or do not read either.

Discussion Questions

1. Ray wanted EAC formula #2 and went looking for a justification afterward. Wei's TCPI test caught it. Design a second control that would catch the same behavior when nobody like Wei is in the room — and state what your control costs, because every real control costs something.
2. The mechanical verification found $275,400 of overstated EV without finding a single dishonest person. Explain the mechanism that produces that result reliably, and say why Kestrel is partly responsible for it under Chapter 32's pay application process and Chapter 16's buyout.
3. Containing the curtain wall was worth $180,209 and fixing framing and drywall was worth $195,433. You have one superintendent's attention and one month. Which do you spend it on, and what additional information would change your answer?
4. Nadia's rule requires a written sentence before formula #2 may be used. Write the sentence for a case where formula #2 genuinely is correct — invent the closed, non-recurring event and make the sentence good enough that you would sign it.
5. The report was correct, complete, and $672,000 of bad news. §30.6 carries a ⚠️ warning that a bad index communicated as a demand rather than as a problem to be solved is how schedule pressure gets installed as a hazard — the third finding of the week-34 scaffold investigation (Chapter 24). Describe how you would communicate a CPI of 0.857 to the curtain wall crew without producing that outcome.

Your Turn

You are Wei Chen, and Nadia wants the one-page month-11 executive summary on her desk before the report goes to Pri Sethi at Meridian. Meridian's contract gives it audit rights over the cost of the work and a 75 percent share of unused contingency, so this page will be read by someone with a financial interest in it and a right to accurate cost information.

Produce it. One page, and it must contain:

(a) The four measures and the four derived numbers, with the data date stated at the top. (b) The three percentages — time elapsed, EV ÷ BAC, AC ÷ BAC — and one sentence saying what the gap between the second and third means. (c) The three packages that matter, each with its CPI, its percent complete, the cause in one clause, and whether it is a discovery, an opportunity, or a measurement failure. (d) The EAC you are reporting, the formula that produced it, the TCPI it requires, and the one-sentence justification. Include the spread across all four formulas — you do not get to publish only the number you like. (e) One sentence on the contingency position: buffer available, forecast draw, buffer remaining. (f) The last two sentences, which are the hardest ones to write and the reason this chapter exists: state plainly that SV and SPI are measured in dollars of work volume and cannot produce a completion date, then give Meridian the actual forecast substantial completion date from the CPM and the total float on the controlling path.

If you cannot fit it on one page, cut (c) to two packages. Do not cut (f).