Case Study 12-2 — Lost by 0.4 Percent, Won by Eleven

Two bid-day post-mortems, and the spread as a diagnostic instrument

Setup

Two public hard-bid openings, four months apart, both attended by Kestrel Construction Group.

On the first, Kestrel came second by fifty thousand dollars and the estimating department spent a week being unhappy about it.

On the second, Curtis Boone's team was low by $2.46 million and the office bought lunch.

One of those was a good day. It was not the second one.

(All companies, people, projects, and figures here are illustrative composites, internally consistent with the rest of this book.)

Job A — Ridgeline Aquatic Center: lost by 0.40%

The project. A municipal aquatics and recreation facility, design-bid-build, lump sum, public owner, prevailing wage, 100% payment and performance bonds. Kestrel bid it out of the Rivermont office with Tomás Reyes leading the estimate.

The bid tab, read out loud in a public opening at 2:00 p.m.:

Total spread, low to high: 3.10%.

Tomás's post-mortem memo was four paragraphs long and its conclusion was: no corrective action.

Here is his reasoning.

Seven bidders inside a 3.1% band means the documents were clear. When a set of drawings and specifications is ambiguous, bidders guess differently, and the guesses scatter. A tight cluster is evidence that everyone read the same scope and priced the same building. That is a compliment to the design team and a useful piece of market information.

Kestrel's number was not wrong. It sat in the middle of a tight cluster, which is the strongest available evidence that an estimate is correctly built. Six independent estimating departments, working from the same documents with different subcontractors and different overhead structures, converged. That convergence is the closest thing to external validation an estimator ever gets.

The $50,000 gap had two identifiable sources, and both were legitimate:

Neither is a mistake. The first is Kestrel's mason being more conservative about a February masonry start than Northline's mason — and if the winter is bad, Northline's mason will be the one having a difficult conversation. The second is timing, and timing is not a skill.

The uncomfortable question Tomás asked anyway. Should Kestrel have shaved the winter-protection allowance to win? The memo's answer:

We can price risk out of a bid or we can price it in. Pricing it out does not remove the risk; it transfers it from a line item into next spring's cost report. If we had cut the $31,000 and won, we would now be the low bidder on a job with a February masonry start and no protection money. That is not a win. It is a bet, placed without telling anyone we placed it.

Diagnosis: a good loss. Kestrel was outbid, not out-estimated. The estimating department's job on a loss like this is to record the market intelligence — Northline's steel advantage, the mason spread — and move on.

Job B — Rivermont Elementary School #12: won by 11.0%

The project. $22.4M elementary school, hard-bid design-bid-build lump sum, public school district owner, prevailing wage, 100% payment and performance bonds, 20-month schedule. Curtis Boone's job.

The bid tab:

Read that tab the way Tomás read it, which took him about ninety seconds.

   Gap from #1 to #2:      $22,400,000 → $24,864,000  =  +11.0%
   Gap from #2 to #7:      $24,864,000 → $26,905,000  =   +8.2%

The gap between first and second is larger than the gap across the entire rest of the field.

That is the signature. It is not a signature of superior buying, superior productivity, or a lean overhead structure — none of those things produce an 11% step function at the top of a tab while six other contractors agree with each other within 8%. Six independent estimating departments priced the same building and arrived in the same neighborhood. One did not. The probability that the one is right is very low.

Tomás wrote a one-paragraph note to Nadia Haddad, Kestrel's VP of Operations, on bid day: "Recommend we ask the district for a bid-verification review before award. The spread says we are missing scope."

The district declined. Public procurement rules in that jurisdiction gave the low bidder a narrow window to claim a clerical error with proof, and "our spread looks wrong" is not proof. Boone was confident. The award went through.

The post-mortem, six weeks into the job

The spread was fully explained by errors and thin markup. None of it was efficiency.

What happened next

The gym joist package became a claim, and it was a weak one. Addendum 3 was part of the bid documents. Boone had signed the bid form acknowledging receipt of all addenda — which is precisely what a bid form's addendum-acknowledgment block exists to establish. The district's position was simple and correct: the scope was issued before bid, the bidder acknowledged the addendum, and the lump sum covers the work shown.

Kestrel absorbed the great majority of the $1,180,000. The OFCI installation was absorbed entirely. Escalation was absorbed as it accrued, invisibly, a few thousand dollars at a time, across twenty months of purchase orders — which is the most insidious way to lose money because no single invoice looks wrong.

⚖️ What the contract says. Two clauses did all the work here. First, the addendum acknowledgment on the bid form: it converts "I did not see it" into "I certified that I did." Second, the lump-sum scope clause — the contractor's obligation is the work shown in the contract documents, and an addendum issued before bid is a contract document. There is no doctrine that rescues a bidder from a scope item they were given and did not price. The remedy for a bid mistake is withdrawal before award, under whatever narrow statutory window the jurisdiction provides, and those windows are short, evidence-hungry, and vary substantially by state. Boone's team did not have the evidence, because they did not know yet that they had made an error. The only mechanism that would have saved them was the one Tomás recommended and nobody used: reading the spread on bid day.

Analysis: the spread as an instrument

Bid results are usually treated as a scoreboard. They are better used as a measurement of your own estimate, taken with an instrument you cannot buy: six competitors' independent opinions of the same documents, delivered on the same afternoon, for free.

The mathematical intuition behind the third row. Estimating errors are roughly independent across firms. Six firms clustering within 8% of each other and 11% away from you is a strong statistical signal, in the same way that six independent measurements agreeing with each other and disagreeing with yours tells you which instrument to distrust. You do not need a formal test; you need the willingness to believe the tab.

Why Boone is not a villain

Curtis Boone is a skilled builder, a good closer, and well liked by owners. He is also wrong about the model, and that is the more useful lesson.

His estimating process was fast, and fast is a real competitive asset — he bids more work than Tomás does. What his process lacked was a reconciliation step: no independent in-house number for the steel package, no addendum-to-quantity re-tabulation checklist, no escalation policy, and no bid-day spread review. Every one of those is a check, and checks cost time. Boone had traded them for volume, deliberately, and for several years the trade worked.

It worked until the day the tab told him it hadn't, and he did not read the tab.

The chapter's ninety-minute pre-bid review checklist (§12.10.2) would have caught two of the three errors — item 4 (addenda incorporated and drawing log reconciled) catches the gym joists, and item 17 (escalation applied to the correct scope and window) catches the escalation. The third, OFCI installation, is a specification-reading error that item 11 and a scope sheet would likely have surfaced during buyout, when it would still have been survivable.

The two post-mortems, side by side

The uncomfortable conclusion: on the day, Job B looked like a win and Job A looked like a disappointment. The estimating department knew otherwise within ninety seconds of each opening, and in one of the two cases nobody acted on what they knew.

Discussion Questions

1. Tomás recommended asking for a bid-verification review on Job B and the district declined. Given that public procurement rules are jurisdiction-specific and generally require proof of a clerical error rather than a suspicion of one, what could Kestrel realistically have done between the 2:00 p.m. opening and the award? What should the company's standing policy be?

2. Job A's $31,000 masonry winter-protection allowance is the difference between a bid that carries a risk and one that does not. Argue both sides. Then decide: does the answer change if the bid form permits qualifications, and if so, how would you word the qualification?

3. The "markup differential" line explains roughly $660,000 of the Job B spread. Boone carried 6.5% where the market carried 9.5%. Is a thin markup an error? Under what circumstances is it a defensible strategic choice, and what has to be true about the rest of the estimate for that choice to be survivable?

4. Design the bid-day spread review as a formal fifteen-minute procedure: who attends, what gets computed, what thresholds trigger what actions, and who has authority to act. Be specific about the arithmetic — what exactly do you calculate from a bid tab?

5. Both post-mortems produced information. Only one produced a change. What makes an organization capable of learning from a win that was actually a loss? Connect this to the argument in Chapter 28 that the cost report is what generates next year's estimate.

Your Turn

You are the chief estimator. Your firm has just bid a $9,200,000 lump-sum project. The tab reads:

(a) Compute the percentage gap from #1 to #2, and the percentage gap from #2 to #5. State which row of the diagnostic table you are in.

(b) Write the three-sentence note you send to your VP of Operations at 2:15 p.m. on bid day.

(c) List the six specific things you would check first, in order, and say why that order. (Hint: order them by dollars-at-risk per minute of checking, not alphabetically or by CSI division.)

(d) Your estimator swears the number is right and can defend every line. Does that change your recommendation? Write one paragraph explaining why an estimator's confidence is not evidence — and one sentence describing the single piece of evidence that would change your mind.