Chapter 27 — Key Takeaways

A one-page reference card. Self-contained: if you come back to this page in three years having forgotten the chapter, everything you need to run the system is here.

Key Takeaways

• 🚪 The threshold concept: the schedule tells you what should happen; the constraint log tells you what can. A CPM network assumes that when a predecessor finishes, its successor is ready to go. On a real job that assumption is false more often than it is true. Work is only ready when every constraint on it has been removed — and making work ready is a separate discipline from scheduling it.
• This is not an attack on CPM. The critical path method is excellent at four things: proving the work fits in the time, calculating the longest path, analyzing delay after the fact, and communicating with an owner and a lender. It is poor at exactly one thing — telling a foreman whether this work is ready for this crew on Tuesday. Different questions, different tools, and you need both on the wall.
• The system in four words: SHOULD → CAN → WILL → DID. Master schedule and pull plan say what should happen. The look-ahead screen says what can. The weekly work plan says what will. PPC and reason codes say what did — and feed it back to the top.
• The "last planner" is the person who makes the final assignment before work is executed — the trade foreman who tells a crew where to be at 6:30 tomorrow. Everyone above them plans; only the last planner commits.
• Variability, not slowness, is what destroys construction productivity. Five trades in a chain, each 90% reliable, produce an on-time fifth handoff about 59% of the time (0.9⁵). At Northgate's measured 54% per trade it is 4.6% (0.54⁵). To get a 90% reliable chain of five you would need 97.9% per trade, which is unattainable — so either raise reliability a great deal or shorten the chain.
• Every trade protects itself, and the buffers stack. Longer quoted durations, a second job kept warm, a deliberate late start, hidden float nobody declares. Each response is locally intelligent and globally catastrophic, and nobody in it is behaving badly.
• Trades will trade their hidden float for reliability. They will not trade it for pressure. Renny Kovach's curtain-wall duration went from 26 work days to 21 in a pull-planning session — not because anyone pushed, but because he was guaranteed a continuous elevation with nobody else on it. A GC who demands the five days gets a foreman who says 26 and means 30.
• Half the disagreements on a job site are not about sequence. They are about what "done" means, discovered three weeks after everyone assumed they agreed. Twenty minutes writing the definition of "dried-in" was the most valuable twenty minutes of the Northgate pull-planning session.
• PPC is binary. Ninety percent of a handoff is zero handoffs. The score does not measure production — the cost report and earned value already do that. It measures whether the next trade can rely on the plan, and reliability is binary from the receiver's point of view.
• The reason codes matter more than the score. PPC is the thermometer; the reasons are the diagnosis. A team that collects the score without the reasons has bought a thermometer and no medicine.
• Track commitments made, not just PPC. The denominator is under the team's control, and the cheapest way to raise the number is to promise less. 90% on nine commitments is much worse than 72% on forty. 100% is a warning sign, not a triumph — it means the team is planning below its capacity.
• On Northgate, 55% of the misses were caused by the general contractor, not by a subcontractor: prerequisite work, material, and information. That is the usual finding, and it is why the score has to be safe to report. The moment a GC uses PPC as a stick, it loses access to the data showing the GC is the problem.
• A project that adopts the Last Planner System without giving someone protected time to remove constraints has adopted a meeting, not a system. Northgate's six-week look-ahead carried 35 to 60 open constraints at any moment, each needing a phone call, an escalation, a booked inspection, a confirmed delivery, or somebody physically walking a floor to verify that a prerequisite was actually finished rather than reported finished. That is not something a PM does "as well as" running a $47.5 million job.
• Screening protects the credibility of the plan, and credibility is the asset the whole system runs on. Put unready work in a plan and you get a demobilized crew, a displaced crew stacked into somebody else's area, and — worst — a foreman who has learned the plan is fiction. A short plan that is true beats a long plan that is aspirational, every week.
• A screen without a workable backlog just idles people. Keep a reserve of pre-screened, fully ready work somewhere else in the building for the crews the screen displaces.
• Six weeks is a default, not a law. The right horizon is the longest lead time among your typical constraints. If your dominant constraint is a fabricated item with an eight-week lead, a six-week look-ahead finds the problem two weeks after it became unfixable.
• A weekly work plan is a contemporaneous record, and it cuts both ways in a claim. A subcontractor reluctant to write down a commitment is being appropriately cautious about creating evidence against themselves, not obstructive. And a planning artifact is never contractual notice — a logged constraint is not notice of delay, and a missed commitment is not a written claim. Run both systems.
• Variability produces firefighting, and firefighting is the condition in which people get hurt. Unready work forces improvisation: the scaffold modified at night by a trade that did not erect it, the crew working in a congested area it was displaced into, the stale inspection tag nobody wanted to stop for. The most underrated safety intervention in this book is a constraint log that runs six weeks ahead and is actually worked.

The Five Levels

Four nested levels of planning, plus a scoreboard. Levels 2 and 4 are the meeting. Levels 3 and 5 are the system — and they are the two that get skipped, because they are boring and require somebody's actual time.

The Eight Constraint Categories

Learn these as a list you can recite, because you will run through them dozens of times a week. They are also, deliberately, the same taxonomy you use to code a miss — so every failure tells you which part of your screen leaked.

The rule

An activity does not enter the weekly work plan until every constraint on it has been removed.

Not "is likely to be removed." Not "the architect said probably Thursday." Removed. The RFI is answered and in the field. The material is on the floor. The inspection is scheduled with a confirmation number.

The Measurement

PPC = commitments completed ÷ commitments made. Northgate's baseline: 101 ÷ 187 = 54%.

Three rules govern it, and all three matter.

1. Binary. No partial credit, ever. "We got 90% of it" scores zero.
2. The last planner scores their own commitment, and the score is not a performance review. The foreman says done or not done; nobody argues. Treat it as a disciplinary instrument and foremen will commit to less and less until the number means nothing.
3. Track commitments made alongside PPC, or the ratio will be gamed — usually without anybody intending to.

On benchmarks, honestly: practitioners commonly report first measurements in the 40s to 60s and sustained programs in the high 70s to high 80s, but PPC is not measured consistently enough across projects, companies, or trades to compare one job's number to another's. The only PPC number that means anything is your own, trended over time. Treat any specific published figure with suspicion, including anyone who quotes one at you in a sales meeting.

Then the part that creates improvement: every miss gets a reason in the foreman's own words on the spot Friday morning → every reason gets one of the eight categories (plus weather, plus a genuine "other") → categories are counted and trended, not just listed → the top two or three get a countermeasure with a name and a date, not a discussion.

Action Items — What to Do on Your Job This Week

1. Measure before you change anything. Six weeks, fourteen foremen, commitments made Monday and scored Friday, binary, reasons recorded. Change nothing during the baseline — if you start fixing things you will never know what the baseline was.
2. Count your constraints, right now. Ask whoever you think owns make-ready: how many constraints did you close last week? If nobody can answer, you have a meeting, not a system. This one question is the whole diagnostic.
3. Assign make-ready hours with a name and a source. Twelve to twenty hours a week from a specific person, taken from a specific other duty you reassign or drop. If you cannot name what you are taking it from, you have not done this step.
4. Re-cut your look-ahead by area, not by activity. "Interior rough-in Level 3" becomes nine zone-level lines. You can verify a room; you cannot verify an activity. Prerequisite-work misses fall immediately.
5. Back-calculate material need-by dates to the room, and put the vendor's name and phone number in the constraint log. Most material misses are things sitting in a warehouse forty minutes away that nobody asked for.
6. Put a general-contractor row on the board and score yourself first, before you ask anyone else for a commitment. Publish your own PPC even when — especially when — it is below the job average.
7. Say out loud, in the first meeting, that no recorded miss will be used against anyone. Then spend the next eighteen months proving it. There is no shortcut.
8. Build a workable backlog before you screen for the first time, so the crews the screen displaces have pre-screened ready work to go to.
9. Put the participation requirement in the subcontract, not in a meeting — attendance by a person with authority over the crew, participation in pull planning, look-ahead information. Price it during buyout and make sure the subcontractor knows it was priced. By month four, a request is a suggestion.
10. Run one pull-planning session on the next phase. Three to four hours, a blank wall, the foremen who will actually do the work, a facilitator with no authority to overrule anybody, and one rule above all the others: every note says what that trade needs from the trade before it.

Common Mistakes — and the Fix

Decision Framework

Screening one activity — run these eight, in this order, every time

  1. DESIGN INFO      Every RFI answered, answer usable, current revision in the field?
  2. SUBMITTALS       Approved — and is the approved version what's being installed?
  3. MATERIAL         Here, right, undamaged, and reachable by the crew?
  4. LABOR            This crew, this week, this job — not promised elsewhere?
  5. EQUIPMENT        Reserved for these people on these days?
  6. PREREQUISITE     Physically complete in THIS AREA. Walk it. Don't take a report.
  7. PERMITS/INSPECT  Permit issued; inspection booked with notice and prerequisites met?
  8. SPACE/ACCESS     Available, reachable, lit, and not full of somebody else's crew?

         ANY ONE OPEN  ──►  activity is REMOVED from the weekly work plan.
                            Log the constraint: what · which category · WHO (a person) ·
                            need-by date back-calculated from the start · status.
                            Then send that crew to the workable backlog.

Is this a real commitment, or a topic?

Diagnosing a Last Planner program you inherited

The four conditions — all four, or it decays inside six months

Numbers worth carrying

The honest reading of that last line: on extended general conditions alone this roughly pays for itself and not much more. The real return sits in the tail — it is insurance against the day the slip crosses the contract date and liquidated damages switch on. Sell it that way and you will be believed. And note who actually gained the most: every avoided crew-idle event is a subcontractor's labor productivity, and the subcontractors paid nothing toward the $75,184. That imbalance is the central commercial problem of lean construction, and pretending it does not exist is how a program loses its credibility in month five.

The sentence to carry out of Part V

The schedule tells you what should happen; the constraint log tells you what can — and making work ready is a separate discipline from scheduling it.