Most jobs don't get better on their own. You run the same play week after week, hit the same walls with the same trades, and by the time the punch list rolls around everybody's blaming everybody. The Last Planner System exists to break that loop — not by adding another meeting, but by turning every week into a small, honest experiment. You commit to work, you find out what actually got done, and you figure out why the rest didn't. Do that consistently and the job teaches you how to run it better.
The catch is memory. A superintendent's head can hold maybe two weeks of "why did that slip." Past that it all blurs into "the drywall guys are always behind." Software's real job here isn't the Gantt chart or the pretty dashboard — it's remembering, so the pattern that took you three projects to feel in your gut shows up in the data by week six. That's what makes improvement systematic instead of accidental.
PPC is the score, not the goal
Percent Plan Complete is the number at the center of all this: of the tasks a crew committed to this week, how many finished exactly as promised. Committed to twenty, finished sixteen, that's 80% PPC. Simple to calculate, brutally honest, and widely misunderstood.
The first thing to know is that a task is either done or it isn't. No partial credit. "We got the north wall 90% roughed" counts as a miss. That feels harsh the first month and then it becomes the whole point — partial credit is how schedules lie to you. A wall that's 90% roughed still can't be inspected, still can't be closed, still holds up the next trade. Binary scoring forces you to plan work you can actually finish in the window.
The second thing is that 100% PPC is not the target, and if you're consistently hitting it, you're sandbagging. Crews that always make plan are committing to less than they can do, padding the week so nothing slips. A healthy job runs somewhere in the 70s and 80s and stays curious about the misses. When I see a foreman posting 100% three weeks running, I don't congratulate him — I ask what he left off the list. The number is a thermometer. You don't improve by staring at the thermometer; you improve by understanding the fever.
The real gold is in the misses
Here's the mechanism that actually drives improvement, and it's the part most teams skip: every time a committed task doesn't finish, somebody writes down why. One sentence, one reason code. Not to assign blame — to build a record. Over a few weeks those reasons sort themselves into a handful of buckets, and the buckets are where you find money.
The standard variance categories are worth memorizing because they cover almost everything that goes wrong on a plan:
- Prerequisite work — the trade ahead didn't finish, so the following trade couldn't start. Sequencing problem.
- Materials — the stuff wasn't on site, or was the wrong stuff. Procurement or logistics problem.
- Information — an RFI still open, a detail unclear, a submittal not back. Engineering or coordination problem.
- Labor — the crew didn't show, or showed short-handed. Staffing problem.
- Equipment — the lift, the pump, the crane time wasn't available.
- Prior contractor / rework — the work was there but wrong, and had to be chased down.
- Weather / conditions — the honest act-of-God bucket, which should be small.
- Changed priorities — somebody pulled the crew to a hot spot. This one's a management tell.
The insight isn't in any single miss. It's in the tally. If forty percent of your variances for the month land in "prerequisite work," you don't have a lazy-trade problem, you have a sequencing problem, and no amount of leaning on subs will fix it. If half your misses are "information," your RFI turnaround is strangling the field and that's a conversation with the design team, not the crews. The reason codes turn a pile of anecdotes into a diagnosis. That's the whole trick, and it's why capturing the "why" every single week matters more than the PPC number itself.
Plan, do, check, act — on a one-week clock
Underneath the Last Planner System is an old idea: the PDCA loop. Plan the work, do the work, check what happened, act on what you learned. What makes it powerful in construction is the clock. You're not running a quarterly improvement initiative. You're closing that loop every single week, fifty-plus times a job.
Plan happens in the weekly work plan meeting, where foremen commit to specific, finishable tasks for the coming week — work that's already been screened for constraints in the lookahead. Do is the week itself. Check is Friday or Monday: calculate PPC, log the misses and their reasons. Act is the ten minutes that most teams cut and shouldn't — you look at the misses, pick the one or two patterns worth chasing, and change something before you plan the next week.
Because the loop is so short, small corrections compound fast. A crew that's honest about its misses in week three is planning noticeably tighter by week ten. That's the "improve constantly" idea made concrete: not a poster on the trailer wall, just a fifty-repetition drill you happen to run every week.
Where the lookahead does the heavy lifting
PPC measures how well you executed. The lookahead is where you actually prevent misses before they happen. If you're seeing a wall of "prerequisite" and "information" variances, the fix lives upstream, in a proper three-to-six-week lookahead where you screen tasks for constraints before anyone commits to them in the weekly plan.
The discipline is called making work ready. A task doesn't earn its spot on next week's plan until the constraints are cleared: material on site or confirmed delivering, RFI answered, prior trade's work verified complete, permit or inspection scheduled, crew assigned, equipment reserved. Anything with an open constraint stays in the lookahead as a "not ready" item with an owner and a due date. You're deliberately building a backlog of work that's fully cleared to go, so that when a crew commits, it's committing to something that can actually be done.
This is where a look-ahead scheduling tool earns its keep. Tracking a rolling six-week window of tasks, each with its own list of constraints and owners, gets unmanageable on a whiteboard past a couple dozen activities. In LookAheadWall the plan is visual and location-based — you can see which crews are stacked in the same area, follow a trade-flow sequence across zones, and flag the constraints that would otherwise ambush you Friday afternoon. The point isn't the software; it's that you can't clear constraints you can't see, and the lookahead is how you see them early enough to do something.
Reading the trend, and knowing when it's lying
One week of PPC tells you almost nothing — a rain day or a no-show can swing it fifteen points. The trend is where the signal lives. Plot PPC over eight or ten weeks and a few honest stories emerge.
A steady climb from the 60s into the 80s means your planning is maturing and constraint-screening is working. A plateau in the high 80s that won't budge usually means you've fixed the easy stuff and the remaining misses need a different lever — often something outside the field's control, like procurement lead times or design turnaround. A slow decline is the one to jump on early: it usually means the discipline is slipping, the "act" step is getting skipped, or the team's quietly gone back to committing to hopeful work instead of ready work.
Watch the reason mix alongside the number, because the mix can rot while the number holds. If total PPC is flat but "changed priorities" is creeping up, your job is fighting fires — someone keeps yanking crews to hot spots, and that churn will surface as a schedule problem a month from now whether the current PPC shows it or not.
Trade-level truth, without the blame
Break PPC down by subcontractor and you learn things that are awkward but useful. One trade running consistently below the job average isn't necessarily lazy — often they're the trade that inherits everyone else's delays, always starting late because the work in front of them keeps slipping. The data lets you tell the difference between a sub who can't deliver and a sub who's getting set up to fail by the sequence.
That distinction changes the conversation. Instead of "your guys are always behind," you sit down with a month of reason codes and say, "eighteen of your twenty-two misses were prerequisite — the trade ahead of you. Let's fix the handoff." That's a coordination fix, and it earns you credibility with the sub because you did the homework instead of pointing a finger. Reliability across the whole job improves when the trades believe the scoreboard is fair.
Turning learning into an action you can check
The step that separates jobs that actually improve from jobs that just measure themselves is closing the loop with a real action. A pattern in the data means nothing until someone owns a countermeasure with a due date. Keep it dead simple, four questions: what specifically will we change, who owns it, by when, and how will we know it worked?
Concretely: "Half our March misses were materials showing up late. Countermeasure — we move the material-readiness check to two weeks out in the lookahead instead of one, the PM owns confirming deliveries, starting next cycle, and we'll know it worked if the materials variance drops below ten percent by end of April." Now it's testable. In April you look and either it moved or it didn't. If it didn't, you learned the real bottleneck was upstream of your ordering, and you chase that instead. That's the difference between spinning your wheels and actually converging on the problem.
What makes it stick — and what kills it
The system runs on honest misses, which means it lives or dies on trust. The fastest way to destroy it is to turn PPC into a stick. The first time a foreman gets chewed out for an 80% instead of thanked for an honest number, everybody learns to game it — commitments get vague, misses get reasons that don't sting, and the data quietly becomes fiction. A number that's safe to be honest about is worth infinitely more than a number that's high because everyone's protecting themselves.
The other quiet killer is skipping the "act" step when the job gets busy. Measuring PPC and logging variances but never actually changing anything is just paperwork — you get the cost of the discipline with none of the payoff. If you're short on time, cut something else and keep the ten minutes where you look at the misses and change one thing.
Improvement also plateaus, and that's normal, not failure. When the easy gains are captured and PPC levels off, you set a new focus — maybe shrink constraint-resolution time, maybe attack one stubborn variance category, maybe tighten the handoffs between two specific trades. Fresh eyes help; a new foreman or PM will see friction the veterans have stopped noticing. The goal was never a perfect score. It's a job that runs a little smarter every week than it did the week before, and a crew that expects to keep getting better. Do that across a few projects and it stops being a technique and becomes how your organization builds.