Case Study 1 — The Same Register, Twice: Northgate at GMP and at 60% Complete

Kestrel Construction Group, the Northgate Outpatient Pavilion, and Meridian Health System are illustrative composites, as are all the people in them.

Setup

There are two honest ways to judge a risk register. You can judge it the day it is written, by whether it is well constructed. Or you can wait until the job is more than half built, lay the register next to the contingency-use log, and ask a much harder question: did it point at the right things?

This is that second exercise, run on Northgate at the 60% complete cost review — month fifteen, roughly, with the building enclosed, MEP rough-in under way, and Wei Chen's cost report on the table.

Where we started. At GMP conversion, Kestrel carried a $1,320,000 construction contingency, or 3.01% of the $43,800,000 base. The register behind it had fifteen managed rows with a total expected value of $1,201,550, of which:

Whose money Total EMV
Meridian (owner) — R-01, R-03, R-10, R-12 $326,400
Escalation allowance — R-02 $139,500
Kestrel's construction contingency — the other ten rows $735,650

Where we are. At 60% complete the contingency-use log shows $709,000 drawn, leaving $611,000. The burn ratio is 0.90 — inside the watch band, outside the comfortable one.

Nadia Haddad's question at the review was not "are we over?" It was better than that: "Of the seven hundred and nine thousand you've spent, how much of it did we see coming?"

What happens

Wei Chen had built the answer as a three-way sort. Every row on the original register falls into one of three buckets, and there is a fourth bucket for money that went out against nothing on the list at all.

Bucket 1 — the rows that materialized

ID Risk Estimated at GMP Drawn by 60% What happened
R-03 Imaging equipment change $185,000 @ 60% (owner-funded) $43,650 to Kestrel Became CO #14. Meridian paid $142,750 of a $186,400 cost. Kestrel's contingency absorbed the unsubstantiated remainder
R-05 Adjacent clinic production loss $145,000 @ 40% $75,000 Real, ongoing, and about half the estimate. The negotiated work-hour windows did their job
R-07 MEP clashes above ceiling $230,000 @ 50% $99,000 Clash detection cut it roughly in half. Two rework tranches: $47,000 at 30% complete, $52,000 at 50%
R-09 Weather beyond 12 anticipated days $128,000 @ 45% $71,000 Nineteen adverse days through 60%. Second winter still ahead
R-13 Interior finishes not fully documented $295,000 @ 70% $54,000 The bulk of this is still in front of us; the finish packages issued late but issued
R-15 Safety event on the north elevation $150,000 @ 15% $19,000 The week-34 scaffold near-miss. One day of stopped work on one elevation. No injury
R-01 Undocumented fill, south half $420,000 @ 35% (owner-funded) $31,000 Encountered — but a small pocket, and inside the allowance Kestrel carried. Meridian paid nothing
R-04 Curtain wall fabrication slip $260,000 @ 30% $28,000 Did not slip. The $28,000 is two shop-drawing revision cycles — the cost of the mitigation working
$420,650

Bucket 2 — the rows that never happened

ID Risk Why it did not happen
R-02 Steel escalation Ironbridge Steel was bought out 51 days after NTP, inside the 60-day trigger. Mill pricing moved 5.8% afterward. The $575,200 escalation allowance is untouched
R-06 Permitting cycle beyond 45 days Pre-application meeting with Frank Petrosyan; complete submission; plan-review comments came back in 24 days
R-10 Unknown utility in the north drive Private locate and potholing found the conflict before excavation. Resolved on paper for the cost of the locate, which was scope, not contingency
R-12 Asbestos in the utility vault Pre-demolition survey came back clean

Bucket 3 — the rows still live at 60%

R-08 (electrical craft at peak manning, week 61 — still ahead), R-11 (drywall subcontractor default — drywall is starting now), R-13 residual, R-14 (commissioning), R-15 (always live), plus the second winter on R-09.

Bucket 4 — the money that went out against nothing on the register

Item Amount Was it foreseeable?
Temporary power revision at 10% complete $21,000 Yes, by anyone who had built next to a live clinic before
Underslab utility resequence at 20% complete $41,000 Probably
Structural steel acceleration at 40% complete $168,000 This is the whole case study
Assorted small items, none over $12,000 | $58,350 Individually, no. Collectively, yes
Total off-register $288,350

$288,350 of $709,000 is 40.7% of everything drawn. Kestrel's historical rate for off-register draw is a fifth to a third. This job was worse, and one line explains it.

Analysis

The register was good at the world and blind to the office.

Look at bucket 2. Four risks did not happen, and in three of them it was not luck — it was the response working. The extra borings, the pre-application meeting, the potholing, the 60-day buyout trigger: each of those was a mitigation with a name and an owner, and each one retired a risk. Tomás's register earned its keep before the job started.

Now look at bucket 4. The single largest draw on the entire job — $168,000 of steel acceleration — appears nowhere in the original forty-one rows. Kestrel had a row for steel escalation, a row for the fabricator's capacity, and a row for the erector's manning. It had no row for the eleven days the anchor-bolt and embed submittal sat in Kestrel's own office before it went to Caldwell Structural, after which Ruth Caldwell took her full contractual fourteen days, Ironbridge Steel missed its mill rolling slot, and the next opening was five weeks out. Steel erection start moved from August 4 to August 27 — 23 calendar days, on the critical path.

That is not a market risk, a site risk, or a subcontractor risk. It is a process risk, generated inside Kestrel's own document control. And the register never looked there, because registers are almost always written by people describing what the world might do to them.

The economics of the decision are worth restating, because they were not obvious:

Option Cost Days recovered
Do nothing — absorb 23 days 23 CD × $10,650/CD = $244,950 0
Accelerate — second erection crew, premium time, resequence enclosure by area $168,000 17
Accelerate and absorb the residual 6 days $168,000 + (6 × $10,650) = $231,900 17

Acceleration saved $13,050 — essentially a wash. The decision was not made on that number. It was made because Meridian's leased interim clinic space expires October 1, Year 2: six days late is survivable, twenty-three days late is not. And then the second bill arrived, unpriced — trade stacking, a rework event on deck-edge detailing, and a spike in near-misses in weeks 34 through 36 that produced Milo Serrano's scaffold incident.

The register's blind spot had a signature, and the drawdown curve showed it. Look at the burn ratio across the job: 0.39, 0.66, 0.65, then 0.91 in a single reporting period at 40% complete. That step is the steel acceleration. It did not tell Margo Deacon anything she did not know from standing on the deck. What it did was make an internal process failure visible, in dollars, to people who had never been on the site — about four weeks before a conventional cost report would have shown the same thing.

And the forward look is the uncomfortable part. Wei re-priced the live rows at 60% using today's information rather than GMP-day information:

Row Re-priced at 60% Forward EMV
R-08 electrical craft at peak manning 35% × $175,000 | $61,250
R-11 drywall default — probability raised from 12% to 18% after two late payments to a second-tier supplier 18% × $640,000 | $115,200
R-13 finishes, $186,000 of estimated exposure remaining | 80% × $186,000 $149,000
R-14 commissioning and medical-gas certification 30% × $120,000 | $36,000
R-15 safety — probability raised from 15% to 20% after the near-miss 20% × $150,000 | $30,000
R-05 residual clinic impact 40% × $70,000 | $28,000
R-09 second winter 45% × $57,000 | $25,650
Identified forward exposure $445,100
Unknown-unknown loading @ 35% of identified $155,800
Total forward exposure ~$601,000

The contingency balance is $611,000. The forward exposure is about $601,000.

Ray's line at the review: "We are not comfortable. We are exactly funded, and those are not the same thing."

Discussion questions

  1. Forty percent of contingency draw went to items that were never on the register — above Kestrel's historical range. Design a specific change to the identification process in §6.4 that would have caught the submittal-turnaround risk. Who would have had to be in the room, and what question would they have had to be asked?
  2. R-04 (curtain wall) drew $28,000 and never materialized as a risk. Is that $28,000 a contingency draw or an estimating error? Argue both sides, then say which column you would put it in and why.
  3. Wei raised R-11's probability from 12% to 18% and R-15's from 15% to 20% at the 60% review. What is the professional argument for revising probabilities mid-job, and what is the argument that doing so makes the original register meaningless?
  4. The acceleration arithmetic said the two options were within $13,050 of each other. The decision was made on a lease expiration date instead. What does that tell you about the limits of expected monetary value as a decision tool?
  5. At 60% complete the balance is $611,000 and the forward exposure is roughly $601,000. Write the three sentences you would say to Pri Sethi at the next owner-architect-contractor meeting. Then write the three sentences you would say to Nadia Haddad. Explain any differences.

Your turn

Take the four items in bucket 4 — temporary power revision, underslab utility resequence, steel acceleration, and the $58,350 of assorted small draws. For each one, write the register row that should have existed at GMP: description as an event, category, probability, cost impact, schedule impact, expected value, a named owner from the Northgate cast, a response, and a trigger.

Then total the expected values of your four new rows and add them to the original $735,650 of Kestrel-owned exposure. Compare the new total against the $1,320,000 that was carried. Write two sentences on whether the contingency was sized correctly for the wrong reasons.