Most schedules are a bet you make once and then spend the rest of the job defending. You build the CPM early, print it, tape it to the trailer wall, and by the time drywall shows up it's a work of historical fiction. Nobody learns anything from it because it never changes — it just gets more wrong.
A rolling look-ahead is different, and the difference is the whole point of this article. Because it rebuilds itself every week, it hands you something a bar chart never can: a fresh, honest comparison between what you said would happen and what actually did. Do that fifty-two times a year and you've run fifty-two experiments on your own operation. That's not a planning tool anymore. That's a learning loop. The crews and companies that get quietly, steadily better are the ones who treat it that way.
The loop, and why the weekly cadence matters
Strip away the jargon and continuous improvement is just a short feedback loop run over and over: you plan, you execute, you look at the gap, you figure out why the gap exists, and you change something before the next round. The magic isn't in any one step. It's in the frequency.
Annual lessons-learned meetings fail because the loop is too long. You're trying to remember why the underground got behind eleven months ago, and everyone who knows has moved on to the next job. A weekly look-ahead closes that loop while the evidence is still warm. The plumber who missed his commitment is standing right there Monday morning, and the reason — the fitting didn't show, the inspector no-showed, the GC changed the pour sequence Friday afternoon — is fresh enough to actually fix.
Here's the sequence that works on a real jobsite:
- Plan the week. Every task on the weekly work plan is a specific commitment by a specific person — not "start drywall" but "hang and fire-tape Level 2 north corridor, Wednesday, done Friday." Vague commitments can't be measured, so they can't teach you anything.
- Execute and record. Walk the work. At week's end, each committed task is either done or it isn't. No partial credit — 90% hung is not done, because the taper can't start on 90%.
- Measure the gap. Count what you committed to and what you finished. That ratio is your Percent Plan Complete.
- Find the reason. For every miss, capture why in one honest sentence.
- Change something. Feed that reason back into next week's plan and into how you screen work going forward.
Miss any of those and the loop breaks. Most teams do the first two and skip the last three, which is why they run look-aheads for years and never get better — they're planning, but they're not learning.
PPC: the one number that tells you if your plan is real
Percent Plan Complete is completed commitments divided by total commitments. Commit to twelve tasks, finish nine, you're at 75%. That's the whole formula, and it's the most useful number on the job that nobody tracks.
What PPC actually measures is the reliability of your promises, not the productivity of your crews. Those are different things and confusing them is the classic rookie mistake. A crew can be hustling all week and you can still post a lousy PPC because the work you promised was never really ready to go. Low PPC almost never means people are lazy. It means your planning is optimistic — you're committing to work that has constraints hiding in it.
Some real numbers to calibrate against. Teams new to this routinely start around 50 to 60%. That feels bad until you realize it means half of what everyone confidently promised each week wasn't happening — you just never measured it before, so it hid inside float and finger-pointing. A crew that's dialed in lives in the high 70s to mid 80s. Above 85% sustained is genuinely good. And if you're posting a clean 95% week after week, don't celebrate — you're sandbagging. You've learned to only commit to dead-certain work and you're leaving production on the table. The target isn't 100%. The target is honest.
Track it as a line on a chart, week over week, and post it where the trades can see it. The trend matters more than any single week. One bad week is weather. A three-week slide means something in your make-ready process is breaking down, and now you know to go look.
Variance reasons: where the actual learning lives
PPC tells you that you missed. The reason code tells you why, and the why is where every dollar of improvement hides. When a committed task doesn't finish, log it against a short, standard set of categories so you can count patterns later. Keep the list tight — five to eight buckets — or people won't use it:
- Prerequisite work not complete (the trade ahead of you ran over)
- Materials or equipment not on site
- Missing or incomplete information (RFI open, detail unresolved)
- Labor short or reassigned
- Change in priorities / directive from above
- Inspection or approval not obtained
- Weather or site conditions
- Underestimated the work (bad duration)
One week of reason codes is noise. Ten weeks is a map. When you tally them up, a pattern almost always jumps out, and it's rarely the thing everyone was yelling about. Maybe 40% of your misses trace back to prerequisite work — which means your problem isn't the trade that got blamed, it's the sequencing two steps upstream. Maybe materials keep showing up late, which points at procurement lead times nobody's managing, not at "unreliable subs." The reason codes turn a wall of individual excuses into a short list of systemic problems you can actually solve.
The discipline here is fighting the urge to write "labor" every time because it's the easy answer. If the real reason the framers didn't finish is that the slab wasn't cured and swept, "prerequisite not complete" is the truth and "labor short" is a lie that'll cost you the lesson.
Make-ready screening — and the lookback that sharpens it
The front half of a good rolling look-ahead is make-ready: you pull tasks off the master schedule six, four, two weeks out and screen them for constraints before they land in a weekly commitment. Materials ordered? Prior trade complete? Information in hand? Access clear? Inspection scheduled? Only work that's genuinely clear should ever become a promise for the coming week.
Here's the loop most people miss. Your variance data tells you how good your screening actually is. If a task blew up on a constraint you supposedly cleared during make-ready — the fitting was "ordered" but showed up wrong, the inspection was "scheduled" but the inspector no-showed — then your screening question was too soft. "Is it ordered?" and "is it on site and correct?" are different questions, and the second one is the one that matters. Every variance is feedback on your make-ready process, and over a couple of months your screening questions should get noticeably harder and more specific because the misses taught you where you were being lazy.
Durations get better the same way
Your duration estimates are guesses until you make them data. Track planned versus actual on the tasks you repeat, and the systematic errors show up fast. On multi-family work you'll often find rough carpentry running 15 to 20% longer than anybody plans, because the plan quietly ignores layout, cleanup, and the punch-and-fix churn between the framer and the inspector. MEP rough-ins tend to be more predictable per unit but murder you on coordination in tight ceilings. Finishes are where optimism goes to die.
Two rules of thumb worth stealing. First, the handoffs between trades usually want a buffer the schedule pretends they don't — frame-to-rough-in generally needs a day or two for cleanup, correction, and the framing inspection before the next trade can really move, and if you plan them nose-to-tail you'll miss both commitments. Second, when a duration estimate and a crew's field experience disagree, believe the crew. They know the work takes nine days. The schedule that says seven is the thing that's wrong, and pretending otherwise just manufactures a variance you could've predicted on day one.
The trap that kills the whole thing
All of this dies the instant the PPC board becomes a blame board. The first time a super uses the variance meeting to hang a foreman out in front of his peers, every reason code on that job turns to "weather" and "labor" forever, and you never get another honest number out of anyone. People will absolutely lie to a system that punishes them, and a look-ahead full of comfortable lies is worse than no look-ahead at all — it's the same fiction as the CPM on the wall, just with more meetings.
So run the variance conversation as forensics, not court. The only question is "what happened and how do we keep it from happening again," never "whose fault was it." When a task misses because a promise you made upstream fell through, say so out loud. Superintendents who own their own misses in front of the trades get honest reason codes back, and honest reason codes are the entire fuel supply for improvement. It's a cheap price for the truth.
Turning the whole thing into a habit
Improvement isn't a program you launch, it's a cadence you keep. A rhythm that holds up under real project pressure looks like this:
- Weekly: Score last week's PPC, log a reason for every miss, and carry the lessons into this week's commitments and next week's make-ready screen.
- Monthly: Step back and tally the reason codes. What's the number-one cause of variance this month, and what one process change attacks it?
- Per project: A real retrospective at closeout — where did PPC land, how did it trend, which fixes worked, what would you screen differently next time.
- Across projects: Push the durable lessons into how the next job gets planned, so a new team doesn't re-learn what the last one already paid for.
The reason this works better as a habit than as a heroic effort is that the rolling look-ahead already forces the loop every single week. You're rebuilding the plan Monday regardless — the only question is whether you spend twenty extra minutes asking why last week's plan was wrong. That twenty minutes is the difference between running the same schedule fifty-two times and running fifty-two better schedules.
Good software makes the bookkeeping disappear so you can spend the time on the thinking. When the tool scores PPC automatically, keeps the reason codes tallied, and lets you pull up what you actually committed to six weeks ago instead of arguing about it from memory, the loop runs on its own momentum. That's the practical case for building your weekly work plans and trade-flow sequences somewhere like LookAheadWall rather than in a spreadsheet that forgets everything the moment you overwrite it — the history is the whole asset, and a fresh blank grid every Monday throws it away.
None of this requires new talent or a bigger crew. It requires measuring the gap between your plan and reality, telling the truth about why it exists, and changing one thing before you do it again. Run that loop with any discipline at all and you won't just finish this week's work. You'll be measurably better at planning the next job than you were on this one — and that compounds, project after project, in a way that no static schedule on a trailer wall ever will.