Chapter 29 — Schedule Control: Updates, Delays, Acceleration, and Getting Back on Track

The Hook: Thirteen Thousand Dollars Is Not a Decision

Hank Duffy called at ten past four on a Wednesday afternoon, the twenty-second of April, Year 1. I was in the trailer with a curtain-wall scope sheet in front of me and I answered without looking at the phone, which is a habit I have never broken and never will.

"Ray. The slot's gone."

Hank runs projects for Ironbridge Steel. He is sixty-one years old, he has been putting steel in the air since he was nineteen, and he does not soften things.

"The mill wanted the release by the eighteenth. Your engineer's stamp came back on the twenty-first. I called them this morning. Next opening they'll give me for those shapes is May twenty-third."

I did the subtraction in my head and got a number I did not like. "That's five weeks."

"Thirty-five days. I can claw some of it back. My detailer's already through the shop drawings, so that time isn't lost. I can roll the secondary framing out of stock instead of waiting on the mill for it, and I can deliver by erection zone instead of by piece mark so you can start the north bays before the last truck comes." A pause. "Twelve days. That's the best I've got and I'm not going to pretend it's more."

Thirty-five minus twelve is twenty-three. Steel erection was supposed to start August 4. It was going to start August 27.

I hung up and walked twelve feet to Wei Chen's desk. Wei runs project controls for Kestrel — the schedule, the updates, the impact analyses. I said the words out loud, which is the first honest thing you do in a situation like this, because a delay you have only thought about is a delay you are still negotiating with yourself about.

"Twenty-three days on the steel start. I need to know what it does to September eighteenth."

Wei did not say anything encouraging. Wei said, "I'll have it by seven."

They were there at six. When I came in at 6:40 there was one page on the table, and I want you to notice what was on it, because the whole of this chapter is on that page.

Margo Deacon came in at seven with mud on her boots. Nadia Haddad joined on speaker from the office downtown. Three people, one page.

Nadia read it twice and said, "Thirteen thousand dollars. On a forty-seven-and-a-half-million-dollar job. That's not a decision, Ray, that's noise. Two of those three numbers are estimates and the error bars on both of them are bigger than thirteen thousand dollars. So tell me what actually decides it."

And that is the moment I want you to sit in, because I did not have an answer for about ten seconds, and the answer, when it came, was not on the page at all.

Meridian Health System had leased interim clinic space three miles from the site to keep seeing patients while we built. That lease expires October 1, Year 2. Nobody had put it in a schedule. Nobody had put it in a contract. Pri Sethi had mentioned it once, in a meeting, in month one, and Margo had written it in the margin of a set of minutes.

Six days late, Meridian has a hard week and a lot of apologies. Twenty-three days late, Meridian has clinics scheduled into a building that does not exist.

Nadia said, "Then it isn't a money decision. Accelerate."

That is what schedule control is. Not the software. Not the bar chart. It is the discipline of turning what happened into a defensible forecast, early enough that somebody with authority can still make a choice — and then being honest about everything that choice costs, including the parts that do not have a dollar sign on them. The part with the dollar sign came to $168,000. The rest of the bill showed up in weeks 34 through 36, and section 29.9 is where I pay it.

🏃 Fast Track: If you already run monthly updates, skim 29.1 and 29.3, and go straight to 29.4 (float erosion as the earliest warning a job gives you), 29.6 (the time impact analysis worked end to end), 29.8 (crashing with a critical path that moves mid-problem), and 29.9 (the safety impact review — the section this book exists to write).

🔬 Deep Dive: The forward pass, the backward pass, float, and the network you are about to update all come from Chapter 14, with more worked problems in Appendix B. The dollar side of every number here is Chapter 28; the index version is Chapter 30. When a delay becomes a dispute rather than a management problem, it moves to Chapter 33.

29.1 What a Schedule Update Actually Is — and the Data Date

Let me kill the most common misconception in the trade first, because it is doing real damage in real trailers right now.

A schedule update is not "moving the bars." It is recording what happened and then recalculating what is left.

An update has exactly four inputs and one output.

Output: you press calculate, and the software runs the forward pass, the backward pass, and the float computation exactly the way you did by hand in Chapter 14. It produces a new projected completion date, a new critical path, and a new float value for every activity in the job.

You do not type the completion date. The completion date is a result. If anybody on your team is adjusting bars until the finish milestone lands on a date they like, they are not updating a schedule. They are drawing a picture of one, and the picture will not survive ten minutes of scrutiny in a dispute — or, far more expensively, ten weeks of production.

The data date, and why it is the most misunderstood concept in scheduling

The data date is the instant the update describes. It is the boundary between two different kinds of information, and the boundary is absolute:

                          THE DATA DATE
                                │
        ACTUAL                  │                  PLANNED
        ─────────────────────►  │  ────────────────────────►
                                │
  What happened. Fixed.         │   What we forecast. Movable.
  Recorded, not forecast.       │   Recalculated every update.
  Actual starts, actual         │   Remaining durations, early
  finishes, quantities in       │   and late dates, total float,
  place, days consumed.         │   the critical path.
                                │
   Nothing to the LEFT of       │    Nothing to the RIGHT of
   this line may be forecast.   │    this line may be reported
                                │    as fact.

Three rules fall straight out of that picture, and violating any one of them produces a schedule that lies:

1. No activity may show remaining work to the left of the data date. If the drywall activity still has ten days remaining and those ten days are drawn in a week that has already happened, the schedule is claiming work will be performed in the past. Software will do this without complaining if you feed it bad progress. It is the single fastest tell that an update was not actually run.
2. No activity may show actual progress to the right of the data date. You cannot record an actual start next Tuesday.
3. Every report, every float number, every projected date is "as of" the data date and is meaningless without it. A schedule with no data date is not a schedule. It is a drawing.

I have watched a well-paid consultant present a "current schedule" in an owner meeting and be unable to answer the question what is the data date? The meeting ended right there, and it should have.

Update frequency, and what the contract requires

Kestrel updates Northgate monthly, with a data date on the last calendar day of the month, and issues the update and its narrative to Meridian by the fifth working day. Update 01 has a data date of March 31, Year 1; Update 11 has a data date of January 31, Year 2 — the same data date as the month-eleven cost report in Chapter 28 and the earned value report in Chapter 30. That alignment is not a nicety. If your cost report and your schedule update describe different weeks, you can never put them on the same page, and the whole of theme 2 — the schedule and the budget are the same conversation — dies quietly in a formatting problem.

Monthly is the industry norm for the formal, submitted, contractual update. It is not the tempo the job runs at. The field runs on the weekly work plan and the six-week look-ahead from Chapter 27. Those are the instrument that tells you what is happening this week. The monthly update is the instrument that tells you what it means for September of next year.

⚖️ What the contract says. Most commercial contracts of any size contain a scheduling specification, and it usually contains more obligations than people read. Typical requirements include: submission of a baseline within a stated number of days after notice to proceed; a maximum activity duration (often 15 or 20 work days); a prohibition on open-ended activities and on constraints that override logic; monthly updates on a fixed data date; a written narrative with each update; a requirement that the update be submitted as a condition of payment; a requirement that any request for a time extension be supported by a time impact analysis in a specified form; and sometimes a statement of who owns float (see §29.5).

Read that specification at buyout, not in month nine. Two of its provisions decide things that matter enormously later: whether float is a shared project resource or is assigned to the owner, and whether the contract prescribes a delay-analysis method. Submitting the wrong kind of analysis can get an otherwise valid extension request rejected on procedure without anyone reading the substance. And note that these requirements, the enforceability of float-ownership clauses, and the treatment of schedule submissions as contract documents vary by jurisdiction and by contract form. Bring your actual scheduling specification to your attorney and your scheduler in the same meeting, before the first update.

🔄 Check your understanding. Your superintendent tells you the concrete crew is "about 70 percent done" with the elevated deck. Your scheduler needs an update input. What do you ask him for instead, and why does it matter?

29.2 Progress Override, Retained Logic, and Out-of-Sequence Work

Here is a setting inside every scheduling package that changes your projected completion date by weeks and that most project managers have never heard of. I am giving it its own section because I have seen a job lose an argument over it.

Out-of-sequence progress is work that started before its predecessor finished. It is not a software error. It is normal, because crews are pragmatic and a foreman with an available bay will work in that bay whether or not the schedule says the predecessor is done.

The question is what the software should do about it, and there are two answers.

Retained logic. The relationship is honored for the remaining work. The activity's remaining duration cannot begin until its predecessor finishes, even though some of the activity is already done.

Progress override. The relationship is ignored for the portion already progressed. The activity's remaining duration is allowed to proceed from the data date, independent of the unfinished predecessor.

Work an example, because the abstraction is useless and the arithmetic is brutal.

Data date: contract day 100. Activity A — underground plumbing rough-in, gym slab area. 20 calendar days remaining. Finishes day 120. Activity B — slab on grade, gym. 18 CD total, a finish-to-start successor of A. But the crew placed 8 days' worth of slab in the two bays where the plumbing was already in. 10 CD remaining.

Twenty calendar days on one activity, from one radio button, with identical field data. On Northgate that is $213,000 of exposure. On Willow Street it is $56,000.

Now the professional answer, which is neither setting.

Neither one is "correct." The setting is a question about physical reality, and if you cannot answer the physical question, your logic is wrong and no setting will fix it.

Ask it plainly: can the remaining 10 days of slab physically be placed before the remaining 20 days of plumbing is finished? If the remaining slab is in the bays where the plumbing is already in and the remaining plumbing is somewhere else — then yes, and progress override happens to produce the right answer for the wrong reason. If the remaining slab sits directly over the remaining plumbing — then no, and retained logic is right.

The honest fix is to split the activities by area so the logic describes what actually has to happen: plumbing bay 1 → slab bay 1, plumbing bay 2 → slab bay 2. Then out-of-sequence progress mostly stops occurring, because the schedule finally matches the way the crews are working.

My rules, and they have never let me down:

• Default to retained logic, because it is conservative and because it forces the out-of-sequence work to be visible rather than silently absorbed.
• Run the update both ways once a quarter. If the two answers are within a few days, your logic is healthy. If they are twenty days apart, you have a schedule whose logic no longer describes the job.
• State the setting in the narrative, every single time. Changing it between updates without saying so moves the completion date for reasons no reader can see, and that is a very short walk from §29.10.
• Persistent out-of-sequence progress on the same activities is not a scheduling problem. It is information. It means the field has found a better sequence than the one you gave them. Go find out what it is and put it in the model.

🔍 Why this works. Out-of-sequence progress is the schedule's version of a smoke detector. A CPM network is a model of physical constraint — every relationship is a claim that this cannot happen until that happens. When work routinely violates a relationship and nothing bad occurs, the model has made a false claim, and a model with false constraints in it produces false float everywhere downstream. That is why the fix is in the logic and not in the settings: the setting changes the answer, the logic changes whether the question was right.

🔄 Check your understanding. Your update is run under retained logic and shows substantial completion on June 14. Your subcontractor's own schedule, run under progress override on the same field data, shows May 22. Neither of you has made an arithmetic error. What is the actual disagreement, and how do you settle it?

29.3 Where the Data Comes From — and the Narrative That Has to Go With It

The scheduler who never leaves the trailer produces fiction

Update data has exactly four legitimate sources, and three of them are people.

That last row is the one this section exists for. Wei Chen walks Northgate every Tuesday and Thursday morning before doing anything else. Not to catch anybody. To make the numbers mean something.

🏗️ From the field. I inherited a job in my eighth year where the schedule was updated by a corporate scheduler who visited the site twice in fourteen months. Every update was arithmetically perfect. Every activity had an actual start, an actual finish, a remaining duration, and a footnote. The projected completion date moved by three days over five consecutive updates, which everybody read as stability. It was not stability. It was a scheduler asking a project engineer for remaining durations by email, and the project engineer answering by subtracting elapsed days from the original duration, because he had never been told that was not the question. Every in-progress activity in that schedule reported exactly the remaining duration the baseline implied, forever. The job finished seventy-one days late and the first update that said so was issued eleven days before substantial completion was contractually due. Nobody lied. Nobody even guessed. A schedule updated from a spreadsheet reports the plan back to you with a new date on it, and it will do that with total confidence right up until the day it stops mattering.

The schedule narrative

An update without a narrative is a data file. The schedule narrative is the document that makes it readable, and on a well-run job it is one to three pages, issued with every update, in the same structure every month so a reader can find things.

Here is the structure Kestrel uses. Steal it.

Section 6 deserves one more sentence. A logic change that is not disclosed in the narrative is the most common form of schedule dishonesty in this industry, and it is usually not malicious. Somebody re-sequences three activities to fix a problem, the completion date improves by nine days, and nobody writes it down. Four months later an owner's consultant compares Update 07 to Update 08, finds eighteen changed relationships and no explanation, and every number you produce for the rest of the job carries an asterisk.

🔄 Check your understanding. Your projected substantial completion has been September 18 for four consecutive updates. Is that good news?

29.4 Variance Analysis: Float Erosion Is the Earliest Warning a Job Gives You

Variance analysis on a schedule is three comparisons, and only one of them is worth much.

The float trend table

Here is the single most valuable page Wei Chen produced on Northgate. Six updates, six paths, total float in calendar days. Positive is room; negative is lateness.

Northgate — total float by path, Updates 01 through 06 (calendar days)

Read it as a picture rather than a table.

   Total float, calendar days                 U-01  U-02  U-03  U-04  U-05  U-06
   ────────────────────────────────────────────────────────────────────────────
   P-5 elevators        +48 ████████████████  +48   +47   +45   +42   +40   +37
   P-3 CURTAIN WALL     +34 ███████████       +34   +32   +26   +19   +11    +3   ◄── falling off a cliff
   P-6 precast          +29 █████████         +29   +28   +26   +23   +21   +19
   P-4 MEP ROUGH-IN     +21 ███████           +21   +21   +18   +14    +9    +2   ◄── falling off a cliff
   P-2 steel             +7 ██                 +7   −23    −6    −6    −6    −6   ◄── everyone watched this one
   P-1 substructure       0 |                   0    +3   +20   +20    ✔     ✔
                            ──────────────────────────────────────────────────
                            0 = critical.  Below 0 = the job is late by that much.

Everybody on that job spent five months staring at P-2. It was the crisis, it had a name, it had a $168,000 price tag, and after Update 03 it never moved again. It was also, by Update 04, the least interesting line on the page.

The two lines that took the job apart in Year 2 are P-3 and P-4, and look at how they announce themselves. Curtain wall started with thirty-four days of float — an enormous cushion, more than a month, the kind of number that gets a path taken off every report. It then lost float at an accelerating rate: 2, then 6, then 7, then 8, then 8. At Update 06 it has three days left and is losing eight a month. It will be controlling before the end of September. It is not a warning. It is an appointment.

MEP rough-in tells the same story more quietly: 21 days, then flat, then −3, −4, −5, −7. Two days left.

The level of float tells you where you are. The rate of change tells you where you are going. Only one of those is actionable, and it is not the one on the report.

Kestrel's house rule from Chapter 14 is that any path with 10 work days or less of total float is managed as critical. In calendar days that trips at about 14. P-3 crossed it at Update 04 and P-4 at Update 05. Both were flagged. Neither was fixed, because at Update 04 the entire management attention of that job was pointed at a mill slot in another state.

That is the real lesson of the float trend, and it is a management lesson rather than a technical one: a crisis consumes the attention that would have prevented the next one. The float trend table exists precisely so that the quiet paths keep a voice in the room while the loud one is talking.

The three variance questions worth asking every month

1. Did the completion date move, and by how much? (Lagging, but it is the number the owner asks.)
2. Did the controlling path move — to what, and why? A critical path that jumps from the concrete chain to a procurement chain has changed who can fix the problem. Margo can add a crew. Nobody on site can accelerate a steel mill.
3. Which paths lost float this period, and at what rate? Compute float this update − float last update. Any path that lost more than about a week of float in a 30-day month is being eaten, and the cause is almost always visible in the constraint log if you go look.

💰 Money check — what one month of float erosion is worth. At Update 04, curtain wall had 19 days of float and was losing 7 a month. If someone had spent that month recovering four days on the fabrication chain — one expedited mock-up review, one earlier release of the corner-condition detail — the path would have entered Year 2 with 7 days of float instead of −4. In Chapter 33 you will see that the curtain wall drove eight calendar days of critical delay in the February window of Year 2. Eight days is 8 × $10,650 = $85,200. The intervention that would have prevented it was two phone calls and a walk of a mock-up, in June of Year 1, on a path with nineteen days of float that nobody was worried about.

🔄 Check your understanding. Path A has 2 days of float and has had 2 days of float for six months. Path B has 26 days of float and had 40 last month. Which one gets your Monday morning?

29.5 The Delay Taxonomy — and the Error That Runs Both Ways

Every delay on every project gets sorted by two independent questions. Learn them as two questions, not one, because half the arguments in this industry come from people collapsing them into one.

Question 1 — Is it EXCUSABLE? Does the contractor get a time extension? Question 2 — Is it COMPENSABLE? Does the contractor get money for the extended time?

Two yes/no questions make four boxes, and one of the four does not exist.

Northgate's steel delay lands in the bottom-left box and it is worth saying out loud: the submittal sat in Kestrel's office for eleven days. That is non-excusable. It is ours. No entitlement, no extension, no money — which is exactly why the $168,000 shows up in Chapter 28 as a draw against Kestrel's own construction contingency and in Chapter 30 as a cost variance on the steel package, rather than as a change order to Meridian. Caldwell Structural took its full contractual fourteen days and did nothing wrong. A party using the time the contract gives it is not causing a delay. If fourteen days did not work for your schedule, the place to fix that was the subcontract and the review-period negotiation, months earlier.

Concurrent delay, stated carefully

Concurrent delay is two or more delays, caused by different parties, affecting the critical path during the same period.

The general principle — and I am going to hedge this deliberately, because it deserves hedging — is that true concurrency typically converts a compensable delay into an excusable-but-non-compensable one. The contractor gets time but not money; the owner gets no liquidated damages. The logic is symmetrical: neither party can show that the other's conduct was the operative cause of the loss, because each party's own conduct was independently sufficient to produce the same delay.

The treatment of concurrency varies substantially by jurisdiction and by contract form. Some contracts define it expressly. Some allocate it by clause. Courts and tribunals in different places apply materially different tests for what counts as "concurrent" at all. Two contractors on two sides of a state line, with identical facts, can get opposite answers. Chapter 33 works this in depth with the tests, the apportionment question, and a worked windows analysis. For now, the operational point is the one that sits in your control:

You cannot prove or disprove concurrency without contemporaneous schedule updates showing what was on the critical path in each period. If you are not updating, you have already lost every concurrency argument you will ever have.

🧩 Productive struggle — the error that runs in both directions

Take four minutes on this before reading on. Two scenarios, same job.

(a) Your masonry subcontractor finishes the exterior CMU nine days late. The masonry chain carried 15 days of total float when the period started. The subcontractor's PM sends you a letter asserting a nine-day delay caused by your late layout. He wants nine days of extended general conditions.

(b) The owner is eleven days late releasing an approved finish schedule. Interior finishes carried 20 days of float when the period started. You are about to send a notice claiming eleven days of compensable delay.

In each case: how many days of project delay occurred? What is each party actually entitled to?

Notice: the shortest fuse on the job

⚖️ What the contract says. Nearly every construction contract conditions a contractor's right to additional time — and to money — on written notice given within a stated period after the event. Four things about that provision are worth carrying permanently:

1. The period is short. Commonly somewhere in the range of a few days to a few weeks. It varies by contract form, by owner, and by jurisdiction, and public contracts frequently impose shorter or more formal requirements than private ones. Look it up on your job. Do not assume the last job's clause.
2. The clock starts when you knew or should have known — not when you quantified the impact. This is the misunderstanding that costs the most money. Hank Duffy's call reached me at 4:10 p.m. on April 22. That is when Kestrel's clock started, whatever we did or did not yet know about the downstream effect.
3. You give notice of the event, not of the number. "We do not yet know the cost or time impact and will supplement within X days" is a complete, professional, entirely proper notice. Waiting until you can price it is the single most common way a valid entitlement dies.
4. Notice may be a condition precedent. In many jurisdictions and under many contract forms, a missed notice can extinguish an otherwise perfectly good claim regardless of the merits. In others, courts apply prejudice or waiver doctrines. This variation is enormous. Ask your attorney what it is where your project is.

The practical habit: the notice log lives next to the schedule update. Every delay event identified in §7 of the narrative gets a line — event, date known, notice required by, notice sent, supplement due. On Northgate, section 7 of the narrative and the notice log were reconciled by Dani Okonkwo on the fifth working day of every month, and that half-hour is among the cheapest insurance a project buys. See Chapter 5 for the legal framework and Chapter 31 for the change-order side of the same clock.

29.6 The Time Impact Analysis, Worked

A time impact analysis (TIA) answers exactly one question: how far does this event, by itself, move the completion date? It is the standard prospective method — you run it when the event happens, not two years later in a conference room — and many contracts require it as the form of a time-extension request.

It is five steps. It is not complicated. It is just disciplined.

  ┌────────────────────────────────────────────────────────────────────────┐
  │ 1. START FROM THE ACCEPTED UPDATE IN EFFECT WHEN THE EVENT OCCURRED    │
  │    Not the baseline. Not today's schedule. That one.                   │
  ├────────────────────────────────────────────────────────────────────────┤
  │ 2. BUILD A FRAGNET                                                     │
  │    A small network of activities representing the delay event itself   │
  ├────────────────────────────────────────────────────────────────────────┤
  │ 3. INSERT IT WITH CORRECT LOGIC                                        │
  │    Tie it to the real predecessors and successors. Change nothing else.│
  ├────────────────────────────────────────────────────────────────────────┤
  │ 4. RECALCULATE                                                         │
  │    Forward pass, backward pass, float.                                 │
  ├────────────────────────────────────────────────────────────────────────┤
  │ 5. MEASURE THE CHANGE IN PROJECTED COMPLETION                          │
  │    New projected finish minus old projected finish. That is the impact.│
  └────────────────────────────────────────────────────────────────────────┘

Step 2 needs a definition. A fragnet — fragmentary network — is a small set of activities and relationships that models the delay event as work: not "23 days of delay," but the actual sequence of things that now has to happen. Modeling a delay as a bar labeled "delay" is what amateurs do and it proves nothing. Modeling it as "await next mill rolling opening → roll → fabricate → galvanize → deliver by zone" is a network, and a network can be checked.

Step 3 carries the discipline that makes a TIA credible. You insert the fragnet and change nothing else. Not a duration you have been meaning to fix. Not a sequence you wish you had drawn differently. If you change three things and the completion date moves 23 days, you have not analyzed anything — you have produced an opinion with a network attached.

The before condition: Update 01, data date March 31, Year 1

Here is the procurement chain as Wei's accepted update showed it the month before the event. Northgate reports total float in calendar days, because the contract's remedies — liquidated damages at $5,500/CD and extended general conditions at $5,150/CD — are denominated in calendar days, and a float number in work days invites a conversion error in exactly the meeting where you cannot afford one.

  BEFORE — Update 01, data date March 31, Year 1

  ...substructure chain...──► [ I  slab on grade ]────────────┐   TF = 0  ◄── controlling
                                        EF Aug 1              │
                                                              ▼
  [ C transmit ]──►[ D review ]──►[ F  mill / fab / deliver ]──►[ J  ERECT STEEL ]──► ...
     EF Apr 1        EF Apr 11        EF Jul 25   TF = +7        ES Aug 4, TF = 0
                                                     ▲
                                    seven calendar days of room between
                                    planned approval (Apr 11) and the
                                    required mill release (Apr 18)

Seven days. That is the entire cushion protecting a chain that terminates at a reserved mill rolling slot — a discrete, external, unbuyable event. Margo said it in a meeting in month one, and I wrote it on the back of an agenda, and then buyout happened and nobody looked at it again. The question that would have surfaced it is a single sentence, asked at buyout (Chapter 16): "Hank, what is the last date you can release to the mill, and what happens if you miss it?" Nobody asked, so April 18 lived in exactly two places — Ironbridge's production schedule and Hank Duffy's head — and in none of Kestrel's documents.

🧩 Productive struggle — before you read the answer

Steel delivery is going to slip. Before you look at Wei's numbers, write down your answers:

1. If steel delivery slips 30 calendar days, how many calendar days does the project slip?
2. What single number do you need in order to answer question 1?
3. Kestrel held the submittal 6 days beyond plan. Caldwell used its full contractual review, 4 days beyond the baseline allowance. Whose 6 days? Whose 4? And does the order matter?

The float absorption analysis — where the 23 days actually came from

This is the table I want you to be able to build. It is not "steel is late so the job is late." It tracks a single pool of days being spent.

Line 3 is the whole education. Three days of negative float became thirty-five, because the thing those days ran into was discrete. A crew is continuous — being three days late to a crew costs three days. A reserved rolling slot at a mill is a reservation: you either make it or you wait for the next one. Lead-time chains do not stretch. They snap.

Now the counterfactual, which is the most expensive table in this chapter.

Look at the middle row. If the submittal had left Kestrel's office on the day the plan said, the engineer could have taken every single day the contract allowed her and the job would have lost nothing. Six calendar days of a document sitting in a stack — not eleven, six, the six beyond plan — produced a 23-day project delay and every dollar in this chapter. And they did it because they were the last six days, spent against a threshold nobody had written down.

That is theme 1 in one table: construction management is the management of risk, and the risk here was never "steel might be late." It was "a chain with seven days of float ends at something we cannot buy."

The fragnet, inserted

  AFTER — the fragnet inserted into Update 01, recalculated

  ...substructure chain...──► [ I  slab on grade ]───────────────────────┐  TF now +3
                                        EF Aug 1                         │
                                                                         ▼
  [C]──►[D 14CD]──►┌───────── FRAGNET ─────────┐──►[ J  ERECT STEEL ]──► ...
   Apr 7  Apr 21   │ F1 await next mill opening│      ES Aug 27
                   │    Apr 21 → May 23  32 CD │      TF = −23   ◄── now controlling
                   │ F2 roll & fabricate  46 CD│
                   │ F3 galvanize & load  14 CD│
                   │ F4 deliver by zone    9 CD│
                   │ (F2–F4 compressed 12 CD by│
                   │  Ironbridge recovery)     │
                   └───────────────────────────┘

  RESULT:  projected substantial completion  Sept 18, Y2  ──►  Oct 11, Y2

All 23 days carried straight through, and it is worth saying why, because the reason is not automatic. Downstream of erection, the controlling path — deck and topping, deck-edge and enclosure release, curtain wall, dry-in, finishes, commissioning — carried zero float in the accepted update. There was nothing left in the network to absorb anything. Had there been ten days of float in the enclosure chain, the project impact would have been 13 days, not 23, and the entire decision below would have been different.

⚖️ What the contract says. Two formal references, named for what they do rather than quoted. AACE International publishes a recommended practice on forensic schedule analysis that catalogues and classifies the delay-analysis methods — prospective and retrospective, observational and modeled — and describes the data each one requires and the conditions under which it is appropriate. It is the reference practitioners and experts on both sides of a dispute most commonly work from in North America. The Society of Construction Law Delay and Disruption Protocol performs a similar function internationally: it sets out guidance on delay analysis, concurrency, float ownership, and the recovery of prolongation and disruption costs. Neither is law. Neither is binding unless your contract incorporates it. Both are worth knowing by name, because a consultant who cites one and a consultant who cites the other are often arguing about the same facts from two different frameworks — and because concurrency treatment and the doctrine of constructive acceleration (§29.7) vary by jurisdiction, which both documents say plainly and which you should say plainly too.

🔄 Check your understanding. Wei ran the TIA on the accepted Update 01, not on the current schedule and not on the baseline. Why does it matter which one?

29.7 Recovery and Acceleration: The Four Levers

First, a distinction that matters in both management and law.

• Recovery is getting back to your current obligation after you fell behind. You caused it; you own it; you are restoring the plan.
• Acceleration is finishing earlier than your current entitled completion date. If the owner directs it, the owner generally pays for it.

The techniques are identical. The entitlement is completely different, and the difference is decided by one question: what is your entitled completion date today? Which is decided by the delay taxonomy in §29.5, which is decided by whether you gave notice, which is decided by whether you were updating your schedule. Everything in this book connects to everything else, and this is one of the tightest knots.

There are exactly four ways to compress a schedule. Every acceleration proposal you will ever see is a mix of them.

Lever 1 — overtime, and why the premium is the smaller half

Everyone thinks overtime costs the premium. Work it.

A 12-person crew, burdened straight-time rate $58.00/MH, overtime hours at $81.00/MH. Assume the efficiency figures below, which are in the middle of the range long-standing industry guidance suggests for sustained overtime — the schedule-related productivity factors published by mechanical and electrical contractor associations, and the classic overtime studies practitioners still cite. Treat any published curve as a starting point, not a measurement; the loss varies by trade, season, crew, task, and above all by how many consecutive weeks it runs.

Read the last two columns together, because that is the whole point:

Going from a 40-hour week to a 60-hour week costs you 70 percent more and produces 27.5 percent more. You are buying schedule at about two and a half times the normal price of the work.

Now take the extra $19,440 a week apart, because the split surprises people.

  WHAT YOU PAID EXTRA
    Straight-time-and-burden cost of 240 additional
      attendance hours          240 MH × $58.00              = $13,920
    Overtime premium on those hours
                                240 MH × ($81.00 − $58.00)   = $ 5,520
                                                               ────────
                                                               $19,440

  WHAT YOU DID NOT GET
    Attendance hours purchased                                = 720 MH
    Productive hours delivered at 0.85 efficiency             = 612 MH
    Hours paid for that produced nothing                      = 108 MH
    Average paid rate  $47,280 ÷ 720                          = $65.67/MH
    VALUE OF THE FATIGUE LOSS  108 MH × $65.67               = $ 7,092

The fatigue loss ($7,092) is larger than the entire overtime premium ($5,520) — and unlike the premium, which is a fixed contractual multiplier, the fatigue loss grows the longer the overtime runs. Everybody in the industry prices the premium. Almost nobody prices the 108 hours. Which produces the operational rule:

Overtime is a sprint instrument. Two to four weeks it works. Eight weeks in, you are paying roughly a 70 percent premium for something close to a 15 percent gain, and you have a tired crew on a compressed site.

⚠️ Safety alert. Fatigue is not a productivity variable with a safety footnote. It is a hazard. Sustained long weeks degrade reaction time, judgment, and situational awareness, and they do it on a site where the exposures are falls, struck-by, caught-in-between, and electrocution — OSHA's Focus Four. Every acceleration plan that includes sustained overtime must state how long it runs and what ends it, and it must be reviewed by whoever owns safety on your job before it is priced, not after. See Chapter 24, and §29.9 below, which is the reason this chapter exists.

Lever 2 — added crews and shifts

Adding people is not linear, for reasons that are physical rather than managerial.

Levers 3 and 4 — the ones people skip

Resequencing has no direct cost, which makes it the first thing to try and the last thing anybody tries. Northgate's acceleration included re-releasing the enclosure by elevation instead of by level, so the north and east faces could start while the deck-edge work on the south was still running. That change cost nothing on the crash table. It cost something else, and §29.9 is where that bill arrives.

Changed means and methods — prefabricating MEP racks off site, switching from stick-built to panelized framing, changing a formwork system — can genuinely reduce both cost and duration. Its constraint is lead time. It is available in month four and gone by month fourteen, which is a good argument for running your acceleration options early, before the only lever left is the expensive one.

💡 Aha moment. Notice the shape of that table. The four levers are ordered by cost in almost exactly the reverse of the order in which people reach for them. The first instinct in every trailer in America is overtime, which is the most expensive schedule you can buy and the one that degrades with every week it runs. The last thing anybody tries is resequencing, which is free. And the lever that is genuinely cheapest of all — changing the means and methods — is usually not even considered, because by the time somebody is desperate enough to compress a schedule, the lead time that made prefabrication possible has already gone by. The cost of an acceleration is set months before the acceleration is decided, by how much of the menu is still on the table.

🔄 Check your understanding. You need 12 days on a 9-week activity. Option 1: six-day weeks for the whole activity. Option 2: a second crew for the middle four weeks. Both price out at roughly the same number. Which do you take, and what single piece of information would flip your answer?

29.8 Crashing, the Cost Slope, and the Path That Moves

Crashing is the systematic version of all of this: shorten the project by the cheapest available day, one day at a time, until you have enough days or run out of cheap ones.

The instrument is the cost slope — the cost of buying one calendar day from a given activity.

Cost slope = (crash cost − normal cost) ÷ (normal duration − crash duration)

Two rules govern the procedure, and violating either produces a number that is confidently wrong:

1. You may only crash activities on the critical path. Money spent compressing an activity with float buys nothing. Nothing at all. It is the most common waste in acceleration planning.
2. Crashing changes the critical path, so you must recalculate after every step. A path you compress stops being the longest one at some point, and every dollar after that point is spent on the wrong chain.

The network: Northgate after the TIA

Nine activities, from steel erection start to substantial completion. Durations in calendar days. Day 0 is August 27, Year 1 — the delayed erection start.

                     ┌─►[ L 44 ]──►[ M 90 ]──►[ N 26 ]──┐   ENCLOSURE, EF 284
  [ J 84 ]──►[ K 40 ]┤                                  ├─►[ T 84 ]──►[ U 42 ]──► SC
                     └─►[ P 134 ]──────────►[ S 22 ]────┘   MEP, EF 280

Project duration: 410 CD. August 27, Year 1 + 410 CD = October 11, Year 2 — the TIA answer, reproduced by a different route, which is exactly the cross-check you should always run.

Backward pass, with the finish at 410:

Critical path: J → K → L → M → N → T → U. The MEP branch has 4 days of float. Remember that number; it is about to decide everything.

The crash candidates

Wei and Margo priced five activities with Ironbridge, Cardinal Mechanical, Halcyon Electric, and Kestrel's own crews. Every price is a real quotation, not an allowance.

Check one slope by hand so you know the formula is doing what you think: L: ($1,024,000 − $982,000) ÷ (44 − 38) = $42,000 ÷ 6 = $7,000 per calendar day.

Crashing it, one step at a time

Step 1 — cheapest critical activity is J at $6,000/CD. J sits ahead of the split, so every day bought on J shortens both branches. Crash it to its limit, 7 CD, for $42,000.

Recalculate: J 0→77, K 77→117, L 117→161, M 161→251, N 251→277, P 117→251, S 251→273, T 277→361, U 361→403. MEP float still +4.

Project 410 → 403. Bought 7 CD for $42,000.

Step 2 — next cheapest critical activity is L at $7,000/CD, with 6 CD available. But L is on the enclosure branch only, and the enclosure branch leads the MEP branch by just 4 days. Compressing L by more than 4 buys nothing, because at 4 the MEP branch takes over. Crash L by 4 CD for $28,000.

Recalculate: L 117→157, M 157→247, N 247→273, P 117→251, S 251→273, T 273→357, U 357→399.

Project 403 → 399. Bought 4 CD for $28,000. Running total: 11 CD, $70,000.

And here is the moment. Both branches now finish on day 273. There are two critical paths. The $7,000/CD slope on L still has 2 days of capacity in it, and those two days are now worth exactly nothing on their own, because the MEP branch would govern.

📊 Diagram (described): the path shift. Draw the network twice, side by side. In the left panel the enclosure chain is drawn heavy and the MEP chain light, with "+4" written beside the MEP chain. In the right panel — after L is compressed by four days — both chains are drawn heavy and the "+4" is replaced by "0." Nothing about the MEP work changed. No crew was added, no duration moved, no logic was touched. It became critical because the thing it was standing behind moved out of the way. This is what people mean when they say the critical path is calculated rather than chosen, and it is why you recalculate after every crash step instead of buying a package of days up front.

Step 3 — with two critical paths, a day now costs the sum of a day on each branch, unless you can buy it upstream.

Crash K to its limit, 6 CD, for $48,000.

Recalculate: K 77→111, L 111→151, M 151→241, N 241→267, P 111→245, S 245→267, T 267→351, U 351→393.

Project 399 → 393. Bought 6 CD for $48,000. Running total: 17 calendar days, $118,000 of direct crash cost.

Step 4 — the stopping rule. J and K are exhausted. The next day requires compressing both branches simultaneously, at a combined $26,000 per calendar day. Compare that to what a day is worth:

   Value of one recovered calendar day  =  $5,150 extended GC + $5,500 LDs  =  $10,650
   Marginal cost of calendar day #18                                        =  $26,000
                                                                              ────────
   Net value of buying day 18                                              =  −$15,350

Stop at 17. Not because 17 is a nice number and not because somebody negotiated it, but because day 18 destroys $15,350 of value. That is the rule, and it is the whole of crashing economics:

Crash until the marginal cost of a day exceeds the value of a day — and then, before you stop, ask the separate question of whether any day beyond that point is worth buying for a non-financial reason. Which, on this job, it was.

The direct cost is not the whole cost

The crash table prices the work. Acceleration also carries enabling costs that no cost slope captures, because they are not attached to any one activity:

Total acceleration package: $118,000 direct + $50,000 enabling = $168,000.

That is the canonical number, and now you can see where every dollar of it comes from. In Chapter 28 it appears as two contingency transfers — CT-03, $118,000 to Division 05 Metals, and CT-04, $50,000 to Division 01 General Requirements. In Chapter 30 it is the −$168,000 cost variance on the steel and deck package, the one line in that report that is a decision rather than a discovery. Same money, three documents, one conversation.

🔄 Check your understanding. In step 2, activity L had six calendar days of compression available at $7,000 a day, and Wei bought only four of them. Buying the remaining two would have cost $14,000 — cheap, by the standards of everything else on that page. Why would spending it have been pure waste?

29.9 The Decision — and the Other Bill

The options, priced

Study those totals before you read another sentence.

A is $244,950. B, C, and D are $234,400, $229,800, and $231,900. The three acceleration options are within $4,600 of each other on a $47,500,000 project — 0.0097 percent of the contract. The cheapest one on the page is C, at $229,800, and it beats the option Kestrel actually chose by $2,100.

💰 Money check — what $13,050 and $2,100 actually mean. The headline arithmetic says accelerating fully saves $13,050 against doing nothing (`$244,950 − $231,900). Cross-check it the other way: 17 recovered days are worth17 × $10,650 = $181,050`, and the package cost $168,000. Same $13,050. The arithmetic is right.

The arithmetic is also useless, and knowing why is the professional skill. Every one of those numbers rests on estimates: Ironbridge's crash quotation carries at least ±10 percent, which is ±$42,000 on the erection line alone. The efficiency loss assumed for the Saturday work is a mid-range industry factor, not a measurement. The $5,150 extended-general-conditions rate is a contractual convention, not a forecast of Kestrel's actual burn. A $13,050 spread — and certainly a $2,100 spread — is smaller than the error bars on its own inputs.

When your options differ by less than your own uncertainty, the arithmetic has told you everything it can: it does not matter financially. Stop asking it. Go find the constraint that is not denominated in dollars.

What actually decided it

The constraint was a lease. Meridian's interim clinic space expires October 1, Year 2, and behind that date are patients with appointments and a health system with a board.

Kestrel bought six additional days of margin for $2,100 — three hundred and fifty dollars a day — on a date that had no contract clause behind it at all.

Margo: "One day isn't margin. One day is a coin flip you're calling in September of next year." Ray: "C is two thousand cheaper." Margo: "Two thousand dollars is a Tuesday. Tell Pri you're going to be done on the thirtieth and then be done on the second and see what two thousand dollars bought you." Nadia: "Take D. And Ray — write down why. In a year and a half somebody is going to look at that page and see that we paid two thousand dollars more than the cheapest option, and I want the reason on the same piece of paper as the number."

That last instruction is the one to keep. A decision that is not the arithmetically cheapest option must have its reason recorded next to it, contemporaneously, or it will be read later as an error.

Directed versus constructive acceleration

Two more terms, and they decide who pays.

Directed acceleration is the owner telling you, in writing, to finish earlier than your entitled completion date. This is normally a change: it entitles you to compensation, and the price should be negotiated the way any change is priced (Chapter 31).

Constructive acceleration is the one that gets litigated. Its elements, as generally described, are:

1. An excusable delay occurred;
2. The contractor requested a time extension, properly and within the notice period;
3. The extension was denied, or unreasonably deferred;
4. The owner insisted on the original completion date — by direction, by threat of liquidated damages, or by conduct;
5. The contractor actually accelerated and incurred cost as a result.

Miss any element and it is not constructive acceleration; it is you spending money.

Two honest cautions. First, the doctrine and its elements vary by jurisdiction and by contract form, and some contracts attempt to disclaim it outright. Second, element 5 is where these claims die most often, because the contractor accelerated everything at once and cannot separate the acceleration cost from the cost of the job it was already running. Chapter 33 names this as a recurring failure: an acceleration claim where premium time was never segregated in payroll. If you are accelerating and you believe someone else should pay for it, open separate cost codes on the day you start (Chapter 28) and keep the acceleration hours in them. That is a thirty-minute setup task worth six figures.

The recovery schedule, and the trap in it

When you fall behind, an owner will ask for a recovery schedule. Sometimes the contract requires one when the projected completion slips more than a stated number of days.

What it must contain:

The trap. A recovery schedule submitted with no comment on entitlement can be read — fairly — as an admission that the delay was yours and that you are curing it at your own cost. I have watched an owner's consultant put a recovery schedule on a table and say "they submitted this without reservation; they owned it."

The professional answer is to do both things, separately, in the same package:

"Kestrel submits the attached recovery schedule in cooperation with the Owner's request and without prejudice to its position on entitlement. Kestrel's position regarding responsibility for the delay described in Update 07 §7 and in our notice of [date] is unchanged and is expressly reserved. Nothing in this submission constitutes agreement that the delay is non-excusable or that Kestrel bears responsibility for it, and Kestrel reserves its rights to a time extension and to the costs of acceleration."

Two sentences. Cooperate on the recovery; preserve the position in writing. They are separate acts and both are professional. The contractor who refuses to submit a recovery plan until entitlement is resolved is being obstructive and will pay for it in the relationship. The contractor who submits one silently is giving away an argument for free.

⚠️ The other bill for acceleration

Here is the section this chapter exists to write, and I want to be blunt because the industry is not.

Kestrel accelerated. It worked — the seventeen days came back, and the forecast moved from twenty-three days late to six, which put the certificate of occupancy comfortably ahead of Meridian's October 1 lease expiry. (The residual six days were closed the following September, in the closeout sprint you will read about in Chapter 40; substantial completion was ultimately achieved on the contract date, September 18, Year 2.) The $168,000 was money well spent by any reading of the arithmetic on that page.

And the page was incomplete.

Compression does not stay inside the schedule. It leaves and it goes into the building and into the people. On Northgate the second bill arrived in three installments.

Installment one — trade stacking on the enclosure. Releasing the enclosure by elevation instead of by level — the free lever, the one with "no direct cost" on the crash table — put the deck-edge detail crew, the precast erector, the curtain-wall crew, and the fireproofing crew on the same elevations in the same weeks. Four trades in a space planned for two.

Installment two — the deck-edge rework. The deck-edge geometry was detailed and installed under compression. The embeds did not end up where the model said they were. Every unitized curtain-wall panel on the north elevation was field-shimmed and a portion field-modified. In Chapter 28 that is CT-05, $64,000, drawn from contingency. In Chapter 30 it is a curtain-wall package running at a cost performance index of 0.857 and forecasting $820,426 over budget — which is 122 percent of the entire construction contingency, sitting in one package, five months later, in a different cost code, caused by a decision made in April.

Installment three — week 34. A Tuesday morning, north elevation, level 3. A frame scaffold had been partially modified overnight by a different trade to run conduit; a plank was lifted and not re-secured. Milo Serrano, a mason tender, stepped onto it at 7:20 a.m. The plank shifted. He went down onto the platform and caught himself on the top rail. No injury — entirely by luck. Bea Salgado stopped work on that elevation for the day and ran the investigation.

She found three failures, not one:

1. No competent-person inspection tag for that shift; the tag was two days stale.
2. A scaffold modified by a trade that did not erect it, with no re-inspection.
3. A crew running behind after the steel acceleration, with an unwritten "make it up" pressure.

Finding three is the one that belongs to this chapter, and it is the one nobody wanted to write down. The first two are conditions. The third is the reason the first two were tolerated. A crew under recovery pressure does not decide to be unsafe. It stops stopping. It does not re-tag a scaffold it is standing on, because that is twenty minutes it does not feel it has. Near-miss reports on Northgate spiked in weeks 34 through 36 — the compressed enclosure window — and then fell back. That is not a coincidence. That is a dose-response curve.

So let me state the position this book takes, plainly:

An acceleration decision is a safety decision and a quality decision, and both belong on the cost sheet.

Not as a caveat. Not as a paragraph at the end of the memo. On the sheet, with the other numbers. Because if trade stacking, rework, and elevated incident risk are discussed only in prose, they lose every argument to a column of dollars, every time, in every trailer in this country.

Here is what the Northgate decision page should have looked like:

The "+" is not decoration. On Northgate that plus sign turned out to be $64,000 of deck-edge rework and a curtain-wall package that ran $820,426 over. Nobody could have predicted those figures in April. Everybody could have predicted the sign.

The safety impact review

So here is the practice, and it is not complicated. Any acceleration decision requires a safety impact review, performed by the person who owns safety on your project, before the decision is made — not after. One page. Six questions.

Question 6 is the one that would have changed week 34, and it is the one almost nobody writes. Acceleration plans have start dates and no stop conditions. Build them with both, and give the person who owns safety the authority to invoke the stop condition without asking you first. If that sentence makes you uncomfortable, notice that it is the same authority you already give a crane operator to refuse a pick — and read Chapter 24 again.

Theme 4, stated as hard as I know how: safety is not a line item, and schedule pressure is a hazard exactly like an unguarded edge. The difference is that an unguarded edge is created by somebody in the field and schedule pressure is created by somebody in the office, in a meeting, with a sentence.

29.10 The Owner, the Record, and the Ethics of an Update

Reporting slippage early, and in writing

Three sentences that will serve you for a career:

A schedule discussion is a management conversation about a shared problem. A claim is an assertion of a contractual right. Do not confuse them, and never let a schedule discussion become a claim by accident — or a claim become a schedule discussion by neglect.

The practical translation:

• Tell the owner early, in the narrative, in writing, in the ordinary course. Section 3 and section 7 of the schedule narrative exist for this. A delay disclosed in Update 07 is a project condition. A delay disclosed in month sixteen is an ambush, and an ambush costs you something you cannot buy back: the owner's willingness to believe your other numbers.
• Give notice on the notice clock, separately, in the form the contract requires. The narrative is not notice unless your contract says it is. Do both.
• Bring a plan, not just a number. The three questions Nadia asks about a cost fade in Chapter 28 are the same three here: what is the real forecast, what are we changing, and when do we know whether it worked.

🏗️ From the field. In Year 2 on Northgate, a windows analysis identified 29 calendar days of slip across February, March, and April: 11 days owner-caused (a suspended imaging suite), 7 days genuinely concurrent, and 11 days that were ours — a curtain-wall supplier problem that was Kestrel's responsibility to Meridian. Meridian granted the 18 days it owed. Kestrel built a recovery plan for the 11 it owned, executed it, and hit September 18, Year 2 — the original contract date. The 18-day extension became genuine float, and we spent most of it later on two problems nobody had forecast. That is what a well-run job looks like: you request the extension you are entitled to, you fix the part that is yours, and you bank the difference against the next surprise. The full windows analysis is worked in Chapter 33.

Four ways to lie with a schedule update

Every one of these is available to you. Every one is trivially easy. Every one will be found by somebody with a copy of last month's file and an afternoon.

1. Manufacturing or hiding float. Lengthen a few unstarted durations and float appears on a path you want protected. Shorten them and float vanishes from a path you want to claim on. Nothing in the file is provably false; every number is somebody's estimate. The tell is a duration change on unstarted work with no field reason and no narrative entry.

2. Undisclosed logic changes. Re-sequence three relationships, recover nine days, say nothing. The completion date improves and no reader can see why. This is the most common form of schedule dishonesty in the industry and it is usually careless rather than corrupt — which does not help you at all when the comparison report is on the table.

3. Submitting a recovery schedule you know is unachievable. This is the one I have the least patience for, because it is a lie told to buy three weeks of quiet, and it has a specific victim: the field. A recovery schedule showing 60 electricians when you have hired none is a document that will be used, next month, to tell a superintendent he is behind against a plan that was never real. It also destroys your credibility at the exact moment you will need it, because when the genuine recovery plan arrives in month twelve, nobody believes it.

4. The schedule of convenience. A schedule built not to manage work but to support a position: a baseline padded with float you intend to claim, or an update whose logic is arranged so that a particular party's delay lands on the critical path. Recognize it by a simple test — does this schedule get used? A schedule that nobody works from, that the superintendent has never seen, that exists only to be submitted, is a claim document wearing a hard hat.

Here is the line, and it is the same one as in Chapter 28, because it is the same line:

You are entitled to be wrong. You are not entitled to publish a projection you do not believe.

A schedule is a model, and models are wrong constantly. Durations are estimates; logic is an opinion about physical constraint; the completion date is a calculation performed on both. Nobody in this industry will hold an honest miss against you. Stating a date you do not believe is different in kind, because of who relies on it: an owner planning a move-in, a subcontractor staffing a crew, a supplier reserving a mill slot, and a superintendent deciding how hard to push a crew on a scaffold in week 34.

And the management half, aimed at anyone who will ever receive an update rather than write one: the honesty of your schedules is a property of your culture, not of your people. If the reliable consequence of an early "we are six days behind and here is the plan" is a bad meeting and a note in somebody's file, you have taught your project managers to hold the date and hope. They are not being unethical. They are responding rationally to the incentive you built — and you have converted a manageable six days into an unmanageable thirty.

📋 Try it — run the update, find the delay, write the notice

The Willow Street Community Center. $6,800,000, 24,000 SF, two stories, design-bid-build lump sum, 425 calendar days, liquidated damages $1,200/CD, extended general conditions $1,600/CD — a total exposure of $2,800/CD.

Data date: contract day 150. Here is the accepted baseline for eight activities, and here is what actually happened. Milestone W8 (building dried-in) is a required interim milestone at day 201.

Baseline

Baseline float: W1, W2, W3, W6, W8 = 0 (critical). W5 = +6. W4 and W7 = +36.

As-built progress at data date 150

Your tasks.

• (a) Did the dried-in milestone move, and by how many calendar days? What is the exposure in dollars?
• (b) Was W2 on the critical path? Is the critical path in the same place now?
• (c) How much float was consumed on two other paths?
• (d) Does this require notice, a recovery plan, both, or neither? Write the first sentence of the notice.

Do it with a pencil. Fifteen minutes. The forward pass starts at the data date.

   3 CD   W1 finished day 104 against a baseline 101      (weather)
 + 9 CD   W2 ran 39 CD against a 30 CD duration           (6 RFI + 3 rework)
 + 3 CD   W3 is tracking 29 CD against a 26 CD duration   (crew + delivery)
  ─────
  15 CD

Spaced Review

Recall first. Cover the answers and try to produce them before you read.

1. From Chapter 28 — which column of a cost report actually tells you whether you are making money, and what is the schedule equivalent?

2. From Chapter 27 — when is an activity actually ready to start, and why does that matter to a schedule update?

3. Deep callback to Chapter 14 — what is the critical path, and who owns float?

Project Checkpoint: The Willow Street Month-6 Schedule Update and Recovery Plan

In Chapter 28 you built a cost-code structure and produced a month-6 cost report with a cost-to-complete forecast. This checkpoint is the other half of the same data date, and the two must agree. If your cost report said masonry is running 9 percent over and your schedule update says masonry is on plan, one of them is wrong and finding out which is this month's most valuable work.

Willow Street: $6,800,000, 24,000 SF, 425 calendar days, general conditions $1,600/CD, liquidated damages $1,200/CD, total exposure $2,800/CD. Full package in Appendix K.

Your deliverable has six parts.

1. The update itself. Take the CPM schedule you built in Chapter 14 and update it at a month-6 data date with as-built data: actual starts, actual finishes, remaining durations for every in-progress activity, and any logic changes. State your data date at the top. State whether you used retained logic or progress override, and say why. Then recalculate — forward pass, backward pass, float — and report the new projected substantial completion. Do not type the date. Calculate it.

2. Variance analysis. Baseline finish versus current forecast finish, in calendar days and in dollars at $2,800/CD. Then state today's controlling path as a named chain of activities, and say whether it moved from where it was in your baseline and why.

3. Float erosion on three paths. Build the trend table from §29.4 — three paths you chose in Chapter 14 (make one of them a procurement chain), with total float at each of six monthly update points, and a change-per-month column. Identify the path that is eroding fastest and the month it reaches zero at its current rate. This is the deliverable that will look most like real work to an employer, because almost nobody produces it.

4. The schedule narrative. One to two pages, using the ten-section structure in §29.3. Section 6 — changes to the model — is mandatory and must itemize every logic and duration change with a reason. Section 7 must state your entitlement position on each delay or expressly reserve it.

5. The recovery plan, with three priced options built by the cost-slope method. Identify at least four crashable activities on your controlling path. For each, give normal duration, crash duration, maximum compression, normal cost, crash cost, and cost slope. Then crash step by step, in cheapest-first order, recalculating after every step, and show where the critical path moves. Present three options as a table: days recovered, acceleration cost (direct plus enabling), residual exposure at $2,800/CD, total cost, and forecast substantial completion. State your stopping rule explicitly: the marginal cost of the next day against the $2,800 a day is worth.

6. The safety impact review for the option you recommend. One page, the six questions from §29.9, answered honestly for your plan — including question 3 (how many weeks of overtime and what ends it) and question 6 (the stop condition). Then write the one paragraph that goes on the decision page next to the dollars: what this acceleration does to trade density, to inspection and hold points, and to the two or three activities most exposed to rework.

Recommend one option in a single sentence that does not contain a dollar sign.

Next: Chapter 30 fuses this update and your Chapter 28 cost report into one arithmetic — and shows you why a schedule performance index of 1.02 can sit on top of a critical path that is thirty-four days late.

Chapter Summary

The update, in four inputs and one rule.

The variance hierarchy — read them in this order.

1. Float trend by path, six updates deep. Leading indicator. The rate matters more than the level.
2. Controlling path — where is it, and did it move? A path that moved has changed who can fix it.
3. Forecast completion versus contract. Lagging, and the number the owner asks for.

The delay taxonomy.

Consuming float is not extending the project. A delay to a non-critical activity is not a project delay — and the error runs in both directions with equal frequency.

The TIA in five steps: accepted update in effect at the time → fragnet → insert with correct logic, changing nothing else → recalculate → measure the change in projected completion.

The four compression levers, and what each one really costs.

Crashing, in five moves.

1. Cost slope = (crash cost − normal cost) ÷ (normal duration − crash duration)
2. Crash the cheapest critical activity. Money on a floated activity buys nothing.
3. Recalculate after every step. The path moves.
4. When two paths are critical, buy upstream of the split if you can; otherwise you pay both.
5. 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.

Six things to carry out of this chapter.

• Float erosion is the earliest reliable warning a project gives you. A path with 34 days losing 8 a month is not comfortable. It is scheduled to become your problem in four months, and no report that filters on zero float will ever show it to you.
• Delay is not proportional to lateness. Three days of negative float against a discrete mill slot became thirty-five. Find the thresholds in your procurement chains and manage every path that ends at one as critical, whatever its float says.
• Cooperate on recovery and preserve your entitlement position in writing. They are two separate acts, and doing only the first gives away an argument for free.
• When your options differ by less than your own error bars, the arithmetic is done talking. Go find the constraint that has people behind it — and write down why you chose what you chose.
• ⚠️ An acceleration decision is a safety decision and a quality decision, and both belong on the cost sheet. Northgate's $168,000 was followed by $64,000 of deck-edge rework, a curtain-wall package $820,426 over, and a near-miss spike in weeks 34 through 36 with a crew under an unwritten "make it up" pressure. 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 — and if you ever receive updates rather than write them, remember that the honesty of your schedules is a property of your culture, not of your people.

What's Next

You now have the two halves of project controls: the money's forward view from Chapter 28 and the time's forward view from this chapter. They are still two documents in two units, and an executive who puts them side by side can ask a question neither one can answer. Chapter 30 fuses them into a single arithmetic — three measures, two variances, four indices, and one honest warning about what earned value cannot see. Then Chapter 31 takes on the events that move both the budget and the schedule at once, including CO #14 — the change that got built on a Thursday verbal and cost Kestrel $43,650 it could not substantiate.