Chapter 29 — Key Takeaways

One page. Enough to re-ground you a year from now, in a trailer, with a number you do not like.

Key Takeaways

  • A schedule update is not moving bars. It is recording actual starts, actual finishes, remaining durations, and disclosed logic changes — then recalculating. You never type the completion date. It is a result.
  • The data date is the boundary between fact and forecast. Nothing to its left may be forecast; nothing to its right may be reported as fact. A schedule with no data date is a drawing.
  • Ask for remaining duration, not percent complete. Percent complete is a money question. Days remaining is a schedule question, and the gap between the two answers is the signal.
  • Float erosion is the earliest reliable warning a project gives you. A chain consumes its float before it consumes the completion date. Track total float by path across six updates and read the rate, not the level. Northgate's curtain wall went 34 → 32 → 26 → 19 → 11 → 3 while the projected completion never moved.
  • Consuming float is not extending the project. A delay to a non-critical activity is not a project delay — and this error runs in both directions with equal frequency. Write down that the float was spent anyway.
  • Excusable answers "do I get time." Compensable answers "do I get paid." Two independent questions, four boxes, one of which is always empty. True concurrency generally converts a compensable delay into an excusable-but-non-compensable one — but concurrency treatment varies by jurisdiction and contract form.
  • Notice runs from when you knew or should have known — not from when you priced it. You give notice of the event, not of the number.
  • Delay is not proportional to lateness. Three days of negative float against a discrete mill slot became thirty-five. Lead-time chains do not stretch; they snap. Find the thresholds.
  • A TIA starts from the accepted update in effect when the event occurred, inserts a fragnet with correct logic, changes nothing else, and recalculates.
  • Crash cheapest-first, on critical activities only, recalculating after every step — the path moves. Stop when the marginal cost of a day exceeds the value of a day, then ask separately whether a non-financial constraint justifies buying more.
  • The cost slope prices the compression, not the consequence. On Northgate the second-cheapest day on the crash table produced roughly $494,000 of downstream rework.
  • ⚠️ An acceleration decision is a safety decision and a quality decision, and both belong on the cost sheet. Require a safety impact review before the decision — with a stop condition in it.
  • You are entitled to be wrong. You are not entitled to publish a projection you do not believe.

Action Items — this week, on your job

  1. Find your data date. Ask your scheduler what it is and when the last update was actually run from field data rather than from a spreadsheet.
  2. Print the float report and sort it by total float. Then get last month's and the month before's and build a three-column trend. Any path that lost more than a week of float in a 30-day month goes on your list today.
  3. Ask one foreman for a remaining duration on an in-progress activity you care about, and compare it to what the schedule carries. Do this before you do anything else in this list.
  4. Read your contract's scheduling specification. Update frequency, data date, narrative requirements, prescribed delay-analysis method, and who owns float. Highlight the notice period and write it on the wall.
  5. Check the retained-logic / progress-override setting and whether it changed between the last two updates. If nobody knows, that is your answer.
  6. Open your ball-in-court report sorted by days aging, with a column for what each open item gates. Eleven minutes. Do it Monday.
  7. For every long-lead package, name the discrete external constraint at the end of its chain — the mill slot, the die run, the utility cut-in, the seasonal restriction — and the last date it can be hit. Then manage that path as critical whatever its float says.
  8. If you are accelerating anything, write the stop condition down today and give the person who owns safety the authority to invoke it without asking you.

Common Mistakes — and the fix

Mistake What it costs The fix
Updating the schedule from a spreadsheet without walking the job A schedule that reports the plan back to you with a new date on it, right up until the month it stops mattering Remaining durations come from the person doing the work, verified by a walk. Two mornings a week
Computing remaining duration as original duration − days elapsed A model arithmetically incapable of reporting a problem — the exact schedule twin of forecasting cost as budget × (1 − % complete) Ask "how many more days do you need." Never derive it
Filtering reports on zero float The near-critical paths, where the risk actually lives, are invisible by design Manage everything inside a stated threshold — Kestrel uses 10 work days — plus every chain ending at a constraint you cannot buy
Reading float level instead of float rate Missing a 34-day path that is dead in five months A six-update float trend table with a change-per-month column
Changing logic without disclosing it in the narrative Every number you produce afterward carries an asterisk Section 6 of the narrative, itemized, every change, every month
Claiming delay for slips that only consumed float A rejected notice and a reputation your future notices will inherit Check the float at the start of the period before you write anything
Waiting to give notice until you can price the impact An extinguished entitlement in many jurisdictions and contract forms Notice the event within the period; supplement the number later
Crashing without recalculating Dollars spent on a path that is no longer critical — the purest waste in acceleration Recalculate after every single step
Compressing detailing, layout, or interface coordination because the slope is cheap Rework that appears five months later in a different cost code Compress installation instead, even at a worse slope. Flag interface activities as non-crashable without a quality review
Submitting a recovery schedule with no reservation An admission you did not intend to make Cooperate and reserve, in the same transmittal, in two sentences
Submitting a recovery schedule you know is unachievable Your credibility, at the exact moment you will need it for the real recovery plan Resource-load it. If the crews do not exist, say so and propose what does
Surprising an owner with a delay in month sixteen Something you cannot buy back: their willingness to believe your other numbers Disclose in the narrative, in the ordinary course, every month, early

Decision Framework

Reading any schedule update in six moves

  1. Find the data date. No data date, no report. Check that nothing has remaining work to its left and no actual dates to its right.
  2. Build the float trend by path, six updates deep. Sort by days of float lost per month, not by float. That sort is the agenda.
  3. Name today's controlling path as a chain of work, and ask whether it moved. A path that moved has changed who can fix it — a superintendent can add a crew; nobody on site can accelerate a mill.
  4. Compare the forecast completion to the contract date, and to any hard business date behind it that has no clause attached.
  5. Read the narrative's change log. Every logic change, every duration change on unstarted work, with a reason. Unexplained changes are the finding.
  6. Put the cost report next to it. Same data date, or you have two documents about two different months.

When something has slipped

   Did the COMPLETION DATE move?
        │
        ├── NO ──► Float was consumed. Not a delay. WRITE IT DOWN anyway:
        │           which path, how much, and who spent it.
        │
        └── YES ─► 1. Run a TIA from the accepted update in effect at the time.
                   2. Classify: excusable? compensable? concurrent?
                   3. NOTICE — on the contract's clock, of the event, not the number.
                   4. Price recovery by cost slope. Crash cheapest-first, on the
                      critical path only, recalculating after every step.
                   5. Stop where marginal cost exceeds $/day exposure.
                   6. SAFETY IMPACT REVIEW before the decision. Stop condition in it.
                   7. Decide. If the arithmetic is inside its own error bars, say so
                      and go find the constraint with people behind it.
                   8. Write the reason next to the number, the same day.

The acceleration decision page — the columns it must have

| Days recovered | Direct crash cost | Enabling cost | Residual exposure at $/CD | Peak trades stacked | Interfaces compressed | Weeks of overtime and what ends it | Total |

The first four columns are the ones everybody builds. The middle three are the ones that decide whether the decision was any good, and if they are discussed only in prose they will lose to a column of dollars every time, in every trailer in this country.