Every Last Planner meeting you run leaves a trail. Somebody promised drywall would be hung on the third floor by Thursday. Thursday came, half of it went up, and the reason it didn't finish gets a one-line note in the plan: "waiting on the electrician to finish rough-in." Multiply that by six trades, forty commitments a week, and forty weeks on the job, and you're sitting on a mountain of data whether you look at it or not.
The Last Planner System was never really about the paperwork. It's about making reliable promises and learning from the ones you break. The software just makes the learning possible at a scale you can't hold in your head. The trick is knowing which numbers actually move the job and which ones are just noise on a dashboard. I've watched plenty of teams generate beautiful reports nobody uses. Let's talk about the handful of measures that earn their keep, and what to do with them on a Monday morning.
Start with PPC, but don't stop there
Percent Plan Complete is the one number everyone knows: of the tasks you committed to this week, what fraction did you actually finish? Count each commitment as done or not done — no partial credit. If 34 of 41 promised tasks closed out complete, your PPC is 83%.
A single week's PPC tells you almost nothing. A crew got rained out, a truck was late, somebody's kid got sick — one bad week is weather, not a trend. What matters is the line over time. Pull eight to twelve weeks and look at the slope. A team that's genuinely learning drifts from the low 50s up toward the 70s and 80s as the plan gets more honest and the constraints get cleaned out earlier. A flat line in the 50s week after week means the planning isn't improving — you're committing to work that isn't ready, and everybody at the table has quietly accepted it.
Here's the part people miss: high PPC isn't automatically good. If a foreman is hitting 95% every single week, he's not a genius — he's sandbagging. He's only committing to work he already knows is bankable and leaving the risky, coordination-heavy tasks off the plan. That protects his number and starves the job of the hard conversations. When I see a PPC that's suspiciously high and flat, I go looking for the work that isn't on the plan at all. The goal is reliable promises, not pretty percentages.
The reasons-for-variance log is where the real gold is
PPC tells you how often you missed. The variance reasons tell you why, and that's the data that actually changes how you build. Every time a committed task doesn't finish, someone should record the reason, and those reasons should fall into consistent buckets — otherwise you can't count them. Typical categories:
- Prerequisite work not complete — the trade ahead of you didn't finish, so you couldn't start.
- Materials — not on site, wrong material, or damaged.
- Manpower — the sub didn't field the crew he committed to.
- Information / RFI — waiting on a detail, a submittal, or an answer from design.
- Equipment — lift down, no crane window, tool not available.
- Change / rework — direction changed, or something had to be torn out and redone.
- Weather / access — the honest ones you can't control.
Now run a simple Pareto on a month of these. Nine times out of ten you'll find that two or three categories cause 70 to 80 percent of your misses. That's your leverage. If "prerequisite work not complete" is eating your plan, the problem isn't the trade that got blamed — it's your sequencing and your hand-off timing. If "information" dominates, your RFIs aren't getting answered fast enough and you need to escalate the log with the design team, not nag the field. Chasing every category equally is how you spread yourself thin and fix nothing. Fix the tall bars first.
One caution: the reasons are only as honest as the room. If nobody wants to say "manpower" out loud because the sub is sitting right there, everything gets logged as "weather" or "coordination" and your data turns to mush. Make it clear you're mining the reasons to fix the system, not to build a case against anybody. The day people start gaming the variance log, your analytics are worthless.
Slice PPC by trade and by area
Project-level PPC is a blunt instrument. The same 68% can hide a framing crew running at 85% and an MEP sub dragging along at 45%. Break the number down and the picture sharpens fast.
Slice it by trade and you'll spot the sub who consistently over-promises and under-delivers. That's a conversation worth having early — usually his problem is upstream (he's short on labor across three jobs), and you'd rather know in week 6 than in the last month when he's blowing your finishes. Slice it by area or floor and you'll often find one zone that keeps failing — congested mechanical rooms, a stair core everybody's fighting over, a level where three trades are stacked on top of each other. That's a trade-flow problem, and it usually means your work is tripping over itself in space and time. This is exactly where a location-based look-ahead earns its keep: when you plan work by zone instead of by a flat task list, the collisions show up on the wall before they show up as a missed commitment.
Track constraints as a leading indicator
PPC and variance are lagging indicators — they tell you what already went wrong. If you want to see trouble coming, watch your constraint log, which is a leading indicator by nature. Every task in the six-to-two-week window ahead should get screened: is the design released, materials ordered, prior work scheduled to finish, permits and inspections lined up, manpower committed? Anything that isn't ready is a constraint, and it gets an owner and a need-by date.
Two numbers here are worth watching over time. First, how early you're catching constraints — a healthy job identifies most of them three or four weeks out, with time to actually clear them. If constraints are showing up the same week the work is supposed to start, your look-ahead isn't looking far enough ahead. Second, average time to clear a constraint by type — long-lead material and RFI answers routinely take longest, which tells you those have to be screened earliest. A rule of thumb that's saved me plenty of grief: don't let a task onto the weekly work plan until every constraint on it is cleared. "We'll figure out the material by Tuesday" is not a plan, it's a wish, and wishes tank your PPC.
Balance the number you can see with the one you can't
A quiet trap in all of this is worshipping PPC because it's easy to graph. Pair every lagging measure with a leading one. PPC tells you how last week went; the size of your workable backlog — the pile of fully-constraint-free tasks a crew could pick up if their planned work fell through — tells you how resilient next week is. A team with a deep backlog absorbs a bad break and keeps producing. A team with no backlog sends people home the minute one hand-off slips. Two crews can post the same PPC and be in completely different shape underneath.
Turn the data into a Monday-morning habit
Analytics that live in a monthly report don't change anything. The cadence that works is short and boring:
- Look at the trend, not the week. Open the PPC line and the variance Pareto before your weekly planning meeting, not after.
- Pick one root cause. Take the tallest bar on the Pareto and ask the room what's really behind it. One cause, addressed properly, beats five noted and forgotten.
- Assign a countermeasure with a name and a date. "Order the long-lead switchgear this week — Mike owns it, due Friday." Vague action items are how good data dies.
- Check next month whether it moved. If "materials" was 30% of your misses and it's still 30% after the fix, the fix didn't work. Try something else. This is the whole loop — measure, act, re-measure.
That's it. You don't need a data scientist. You need fifteen honest minutes a week and the discipline to act on what the numbers are telling you.
What the software actually has to do
You can run all of this on a spreadsheet — plenty of great superintendents did for years. What good look-ahead software buys you is speed and honesty: it captures commitments and variances at the moment they happen, so nobody's reconstructing last week from memory on Friday afternoon, and it does the slicing and trending for you so the patterns are visible without a night of pivot tables. That's really the point of a tool like LookAheadWall — the plan lives on a visual, location-based wall the whole team can see, trade-flow sequences are connected instead of implied, and the reliability data falls out of the normal workflow instead of being a separate reporting chore. The measure of any of these tools is whether the data quality survives contact with a busy jobsite. If entering a variance takes so long that the foreman skips it, your analytics are built on sand no matter how slick the dashboard looks.
A few practical guardrails on data quality, because they make or break everything above. Keep your variance categories short and stable — if the list changes every quarter, you can't trend anything. Capture the reason the same day the task misses, while the memory's fresh and honest. And give each audience the view they'll actually use: an executive wants the portfolio trend, a PM wants the variance patterns, and the foreman on the deck wants this week's committed work and nothing else. Bury a superintendent in portfolio charts and he'll close the app and go back to a legal pad.
The bottom line
Your Last Planner data isn't there to grade people. It's there to show you where your building process keeps breaking so you can fix the process instead of just leaning harder on the crews. Watch the PPC trend for honesty, mine the variance reasons for your two or three real problems, catch constraints early enough to clear them, and keep a backlog so one bad break doesn't stall the job. Do that consistently and the numbers stop being a report you file and start being the reason your next project finishes cleaner than your last one.