Every superintendent has lived this Tuesday. The electrical foreman said his rough-in in the east wing would be done Friday. You built the whole next week around it — drywall stocked, hangers scheduled, the inspection request already keyed in. Then Friday afternoon you walk the wing and find half the boxes still hanging by their pigtails because he was two men short all week and never said a word. Now drywall has nowhere to go Monday, you're eating their idle time, and the inspector shows up to a job that isn't ready.
That's not a scheduling failure. It's a promising failure. And it's the single thing that separates a look-ahead process that actually works from a wall full of sticky notes everyone quietly ignores.
A promise is not a prediction
The Last Planner System lives or dies on one idea that sounds obvious and almost nobody does well: the people who do the work commit to what they can actually finish, out loud, to each other. Not what the master schedule says should happen. What they can genuinely promise given the crew, the material, and the conditions in front of them.
The distinction matters more than it looks. A prediction is "electrical should be done Friday." A promise is "I will have rough-in complete in the east wing by Friday, and I'm telling drywall to plan on Monday." The first is a guess dressed up as a plan. The second is a commitment another trade is going to build their own week on top of. When you treat the master schedule's dates as promises, you're asking a foreman to be accountable for a date he never made and may have never believed.
Reliability is the whole ballgame because a jobsite is a chain of handoffs. If your weekly plan strings together ten activities and each one has a 70% chance of finishing as promised, the odds that the whole chain lands on time are around three percent. Get every trade to a genuine 90% and that same chain jumps to roughly a third. You feel that difference in your gut long before you see it on a report — it's the difference between a job that flows and a job where everyone is always waiting on somebody.
What has to be true before a foreman can promise anything
The reason promises break usually isn't that the foreman lied. It's that he committed to work that was never ready to be committed to. Before you accept a promise into the weekly work plan, a handful of things have to be genuinely true — not "probably," not "should be by then," but true.
- The predecessor is actually complete, not "basically done." "Basically done" is where schedules go to die. Framing that's 95% done still has the corner nobody can hang rock on. Walk it or don't count it.
- The material is on site and staged, not on a truck. A confirmed delivery date is not material in hand. If the promise depends on a Thursday delivery, the promise is really a bet on a freight company you don't control.
- The crew is real and available. Not the crew the foreman hopes to have — the crew he'll actually have, minus the guy he already knows is pulling to another job midweek.
- The information is clear. An open RFI on that wall means the work isn't ready to promise, no matter how much everyone wants it moving.
- The area is accessible and released. If the space is still owned by another trade, or a permit or prior inspection is still open, the work is blocked whether anyone admits it or not.
These are your constraints, and clearing them is the make-ready work that has to happen in the weeks before an activity hits the weekly plan. This is exactly what a rolling look-ahead is for. Six weeks out you're just seeing the shape of things. As work pulls into the three- and two-week windows, you're hunting constraints and assigning someone by name to knock each one down — "Mike, get the light fixture submittal approved by the 14th" — so that by the time you ask for a promise, there's nothing left standing in the way of a yes. A tool like LookAheadWall earns its keep here by making those constraints and their owners visible instead of living in your head and three separate email threads, but the discipline is yours to run.
Make the promise specific enough to fail cleanly
Vague promises can't be reliable because nobody can agree on whether they were kept. "Make progress on electrical" is unkillable — it's technically true if one guy pulled one wire. That's not a promise, it's a hedge.
A promise you can hold someone to has four things nailed down: what, where, how you'll know it's done, and by when. "Panel and branch circuits complete and ready for rough-in inspection in electrical room 102 by end of day Thursday" is a real promise. Either room 102 is ready Thursday or it isn't. There's no wiggle room to negotiate after the fact, which sounds harsh but is actually a gift to the foreman — he knows exactly what he signed up for, and so does the trade stacking behind him.
Location-based promises are where short-interval scheduling gets its teeth. "Electrical rough-in, 40 hours" tells you nothing about flow. "Rough-in, east wing, then north wing, then the corridor" tells drywall precisely where they can start and when. Half the trade-stacking fights on a job come from two crews both believing they had the same room this week because nobody ever pinned the promise to a place.
Let people say no — it's the point, not a problem
Here's the part that trips up old-school supers, and I was one of them: a foreman who declines a commitment is doing his job. A promise you squeezed out of a guy under pressure in a coordination meeting isn't reliable, it's just a date he'll break more politely later.
When you lean on someone until he says "yeah, fine, Friday" just to end the conversation, you haven't gained a commitment. You've laundered your own optimism through his mouth and made it his problem. The foreman who says "I can promise the east wing but not the corridor — I'm short a man Wednesday" is the most valuable person in that room. He just handed you accurate information you can plan around instead of a comfortable lie you'll find out about Friday at 3 p.m.
So build the room where no is allowed. Ask "what can you commit to?" and then actually let the answer be less than you wanted. A weekly planning meeting where every foreman promises exactly what the super hoped for is not a well-run meeting. It's a room full of people telling you what you want to hear.
Track whether promises get kept, then use the number
You measure this with Percent Plan Complete — commitments kept divided by commitments made, times a hundred. If the crew made 22 promises this week and kept 18, that's 82%. Simple to calculate, brutal to argue with.
A healthy PPC lands around 80 to 85%. Not 100 — a job running at 100% every week is a job where people are sandbagging, promising less than they know they can do so the number looks pretty. That's gaming the metric and it quietly destroys the whole point, because now the plan is deliberately soft and the schedule loses time nobody can see. Below 70% and you've got a real reliability problem worth stopping to fix.
But the PPC number itself is the least useful part. The gold is in the reasons. Every time a promise breaks, you write down why — one honest word or two. Do that for a month and the pattern jumps off the page. If eight of your last dozen misses trace back to material showing up late, your problem isn't the trades, it's procurement and staging. If they cluster on one sub, you've got a capacity or a management conversation to have with that sub, and now you've got evidence instead of a hunch. PPC tells you that you're bleeding. The variance reasons tell you where the wound is.
The failure modes that quietly wreck reliability
After enough jobs you start seeing the same reliability-killers over and over:
- Optimism bias. Every foreman alive underestimates how long the messy 20% takes. The trim, the punch, the "we'll knock that out Friday morning" stuff. Push back gently on the too-clean commitment.
- Hidden constraints that surface late. The classic is the promise made against an inspection everyone assumed had passed. Verify, don't assume, before it enters the plan.
- The disappearing crew. A foreman promises Monday morning, then loses two guys to another job Wednesday and never tells you. Mid-week check-ins catch this while there's still time to react.
- Coordination collisions. Two trades promise the same area the same week. This only surfaces if promises are location-specific and someone is looking across all of them at once — which is precisely the coordination view a shared look-ahead board gives you.
Notice most of these are catchable days before they hurt you, but only if you're running make-ready seriously and checking in mid-week instead of finding out at the Friday walk.
Why the software matters, and where it doesn't
None of this requires software. Guys ran the Last Planner System with a wall of sticky notes and a grease pencil for years, and a disciplined super with index cards will beat a sloppy one with the best app on the market every time.
What good software does is take the friction out of the parts you'll otherwise skip when you're slammed. It keeps every constraint visible with an owner and a due date so nothing falls through. It calculates PPC and trends it so you can see reliability climbing or sliding without doing arithmetic on a Friday. It puts promises on a shared board where the plumbing foreman can see what the electrician committed to in the same room, so collisions surface in the meeting instead of on the deck. And it makes the whole thing transparent — who's keeping promises, who isn't, and whether it's getting better — which is what turns reliability from a thing you nag people about into a thing the team can actually see and own. That's the honest role of a tool like LookAheadWall: it doesn't create the discipline, it makes the discipline cheap enough that you'll actually keep it up when the job gets hard.
Because in the end the promise is the product. Get your foremen making commitments they believe in, clear the constraints before you ask, let people say no, and measure whether the promises hold — do that and the schedule mostly runs itself. Skip it, and no software, no Gantt chart, and no amount of Friday-afternoon yelling will save the job.