Chapter 11 — Key Takeaways

One page. Self-contained. Come back to this before your next budget meeting.

Key Takeaways

  • The project is built twice. Once on paper — estimates, schedules, coordination, submittals — and once in the field. The first build determines the second, and the cheapest place to fix anything is in a model, an estimate, or a schedule.
  • Accuracy is a function of design completeness, not of effort. Class 5 (program only) is roughly −30/+50%. Class 1 (bid documents) is roughly −3/+10%. You cannot out-work missing information.
  • The band is usually wider than the argument. Northgate's SD estimate of $52,300,591 carried a Class 4 band of roughly $44.5M to $65.4M. The gap everybody spent nine days on was $4.8M. Say the band out loud.
  • Escalate to the midpoint of construction, never to NTP. On Northgate that difference was $1,049,957 — 22% of the gap — and it would have looked like a better number, not a mistake.
  • Escalate only un-bought scope. Northgate's escalation line fell from $2,801,338 at SD to $575,200 at GMP because 84% of the work was bought at fixed prices.
  • Three adjustments to historical cost, in order: time (from the comparable's construction midpoint), location (city cost index ratio), size (roughly 1.5% unit-cost reduction per 10% area increase).
  • Compute the burden multiplier once, before the meeting. Northgate SD: 1.2005. Closing a $4,800,591 gap meant finding $3,998,920 of actual work, not $4.8 million of it.
  • Value = Function ÷ Cost. Same function for less, or more function for the same. Less function for less money is a scope cut — legitimate if the owner chooses it knowingly, dishonest if it wears VE's coat.
  • Contingency is not money. It is an admission. Deleting the reserve does not delete the risk; it deletes the funding and transfers the exposure. A CM who lets an owner "VE" the contingency has harmed that owner.
  • Construction is about 78% of the owner's project budget. If you do not know the other 22%, you will give confident, expensive, wrong advice.
  • The qualifications and assumptions page is the most valuable page in a GMP. The drawings define the building; that page defines the price.
  • Most of what closes a budget gap was never in the building. Between SD and GMP, Northgate's cost of work went up $233,310 while the price came down $4,800,591. Design contingency retiring and escalation dropping did the heavy lifting — and only the VE was something you could promise in advance.

Action Items — this week, on your job

  1. Divide. Take your current estimate and divide it by gross square feet, by functional unit, and by enclosure area. If any ratio surprises you, find out why before anyone else does.
  2. Compute your burden multiplier and write it on the front of the estimate. Then tell your team what a $100,000 raw saving is actually worth.
  3. Check where your escalation is priced to. If it is NTP, fix it today and tell the owner what changed and by how much.
  4. Open your contingency line and ask what each dollar is for. If the answer is a percentage rather than a list of named risks, you cannot defend it and you will lose it in the first budget meeting.
  5. Write four qualification sentences for the four most obviously open decisions on your current documents. Four sentences on Bellhaven were worth $480,000.
  6. Run one constructability lens across your current set — pick access — and ask of the three largest pieces of equipment: how does it get to where it goes, and is the path still open when it arrives?
  7. Ask the owner for their whole project budget. Not the construction line. All of it. Then read the FF&E and owner-furnished equipment lines carefully.

Common Mistakes and the Fix

Mistake What it costs The fix
Presenting a conceptual estimate as a single number with no range The owner makes irreversible decisions on a number that was never that precise State the estimate class and the band, in the same sentence as the number
Escalating to NTP instead of the midpoint of construction $1,049,957 on Northgate; the estimate is quietly low and nobody sees why Compute months to midpoint = months to NTP + half the contract duration. Show the arithmetic on the page
Escalating from a comparable's completion date instead of its construction midpoint ~2.4% low; roughly $1.2M on a $50M project Midpoint = completion date + half that project's duration, counted backward from today
Using a $/SF number without knowing what is in it Anything from 20% to 100% error, invisibly Ask: what area, what included, what excluded, when, where, what market
Setting contingency as a percentage nobody can trace to a risk You lose it the first time an owner asks, because it really is just a number somebody picked Build it bottom-up from a named risk register, with an owner and a response per risk
Letting the owner cut contingency to close the gap Northgate: the drawings later added $2,755,010. Bellhaven: 76% of contingency gone in 31% of the schedule Decline in the meeting, explain what it funds, and if they insist, document the decision and the exposure in a letter
Calling a scope cut "value engineering" The owner discovers what they gave up at year eleven, in an occupied building Run the four-question test. Label program decisions as program decisions
Deleting the mockup or the testing that verifies a change you made to save money A $508,400 facade saving becomes a seven-figure water-infiltration remediation The verification is part of the option, not an add-on to it
A one-paragraph assumptions page on a GMP Bellhaven: ~$480,000 absorbed across four change orders — about $120,000 per missing sentence Twelve to twenty numbered, enforceable items. Every "TBD" on the drawings becomes a line
Never revisiting the VE log after it is signed VE-04's $71,300 giveback never got recorded; the estimating database still shows a $267,700 win Keep a VE risk register and reconcile it at each milestone and at closeout
Optimizing for fee, or for closing the gap, without saying so Trust erodes over eighteen months in ways nobody can point to Name your incentive out loud in the room, then give your recommendation anyway

Decision Framework

Is this real value engineering?

1. Does the FUNCTION change?
      YES → it is a SCOPE CUT. Say so. Legitimate only if the owner
             chooses it knowingly and it is written down as a program decision.
      NO  → continue.

2. Does LIFE-CYCLE cost stay level or improve?
      NO  → it is a TRANSFER from capital budget to operating budget,
             not a saving. Price the whole life or reject it.
      YES → continue.

3. What SCHEDULE and QUALITY risk does it carry?
      Any → it can still be genuine VE, but the risk goes in the log,
             in writing, with a name attached, before anyone accepts it.

4. Is the VERIFICATION that proves it works still in the budget?
      NO  → you have not saved money. You have bought an unpriced liability.
      YES → this is value engineering. Log it, price it burdened, take it
             to the owner with all four columns filled in.

Can I set a GMP yet?

Answer all seven honestly. A "no" anywhere is not a veto — it is a line item you must price, name, or qualify.

Check
1 Are the documents complete enough, or have I carried a named allowance for every incomplete system?
2 Is there a qualifications page — numbered, enforceable, one line per open decision?
3 Does every "TBD" on the drawings appear either as an allowance with a written definition or as an exclusion?
4 Is the contingency bottom-up against a risk register, and is it sized for the documents I actually have?
5 Is the escalation allowance limited to un-bought scope, with a stated buyout assumption and date?
6 Is there a CPM schedule the price is built against, with the milestone dates named in the contract?
7 Do I have real subcontractor pricing on the majority of the work, and do I know the bidder count per trade?

The three numbers to put on the front of every conceptual estimate

  1. The estimate class and its honest band — "Class 4, roughly −15% to +25%."
  2. The effective date the price is built to, and the escalation rate and midpoint assumed.
  3. The burden multiplier, so that everyone in the next meeting is doing the same arithmetic.