Every jobsite generates a mountain of data whether anyone looks at it or not. Every missed commitment, every crew that showed up to a wall that wasn't ready, every inspection that failed on a Friday afternoon — all of it is telling you something. The trouble is that most of that information dies in a superintendent's head or a foreman's truck, and by the next job the same mistakes get made again because nobody wrote down why the drywallers were three days late last time.
Analytics in a scheduling app is just the discipline of catching that information before it evaporates and turning it into something you can act on. Done right, it's the difference between a super who thinks the electrician is always the bottleneck and one who can prove it, name the reason, and fix it. Done wrong, it's a dashboard full of pretty charts nobody opens. This article is about the handful of numbers that actually move a job, what they mean when they go sideways, and how to use them without turning into a spreadsheet jockey.
Start with PPC — the one number that keeps you honest
If you take one metric away from this, make it Percent Plan Complete. PPC is dead simple: of the tasks you committed to this week, what percentage did you actually finish? Twenty tasks planned, sixteen done, that's 80 percent. It comes straight out of the Last Planner methodology, and it's the closest thing our trade has to a reliability score.
What makes PPC powerful isn't the number itself — it's what it forces you to confront. A crew that hits 45 percent week after week isn't lazy; it's a crew whose commitments are being wrecked by something upstream. Maybe the material's not landing, maybe the prior trade never finishes clean, maybe the foreman is committing to work he already knows he can't reach. PPC drags all of that into the light.
A few rules of thumb from running this on real jobs. Anything above 85 percent consistently and your look-ahead is working. Sitting in the 60s and 70s is normal on a busy job with a lot of trade stacking, but it means you've got real constraints to chase. Below 50 and something is structurally broken — usually the plan is a wish list, not a plan. And beware the crew that reports 100 percent every single week. That's not excellence, that's a foreman padding the plan with easy wins and never committing to anything risky. You want honest 80s, not fake 100s.
Variance analysis: the "why" behind the miss
PPC tells you a task slipped. Variance analysis tells you why, and that's where the real money is. Every time a committed task doesn't get done, somebody should log a reason — and the reason needs to fall into a real category, not "ran out of time." The categories that earn their keep on most jobs:
- Prerequisite work — the prior trade wasn't finished or wasn't done right.
- Materials — didn't arrive, arrived wrong, or arrived damaged.
- Manpower — the crew didn't show up in the numbers promised.
- Information / RFI — waiting on a detail, a submittal, an answer from the architect.
- Equipment — the lift, the pump, the crane time wasn't there.
- Inspection / approval — the AHJ hadn't signed off so the next step couldn't start.
- Weather — the honest one, and the one people hide behind when the real cause was one of the above.
The magic happens when you stack a month of these up and look at the distribution. If 40 percent of your misses are tagged "prerequisite work," you don't have a scheduling problem, you have a hand-off problem — your trades are stepping on each other. If materials dominate, your procurement lead times are wrong and no amount of clever sequencing will save you. I've watched a job blame the framers for weeks until the variance log showed that two-thirds of the framing misses were actually late window deliveries. Nobody had connected those dots because nobody was writing the reason down in a place you could add it up.
Constraint trends and the make-ready process
Constraints are the things that have to be cleared before a task can even be committed — the missing submittal, the crane pick that isn't scheduled, the inspection that's not booked. A good look-ahead is really a machine for surfacing constraints six weeks out and grinding them down to zero before the work hits the weekly plan.
The two analytics worth watching here are constraint aging and constraint source. Aging is simply how long a constraint sits open before it's cleared. A constraint that's been red for four weeks is a task that's quietly about to blow your schedule, and it should be screaming at you from the board long before its start date. Source tells you who keeps generating the choke points. When the same design discipline shows up as the origin of a third of your open constraints, that's a conversation to have with the design team, backed by data instead of a hunch.
The practical target: nothing should enter the weekly work plan with an open constraint. If it does, you're planning to fail. Tracking the ratio of constraints identified versus constraints actually cleared before the commitment window is one of the truest measures of whether your make-ready process is real or theater.
Trade reliability — earned, not assumed
Once you've got PPC and variance flowing for a few weeks, you can slice it by trade, and this is where it starts to change how you run the job. Some subs commit conservatively and hit their number every time. Others over-promise in the coordination meeting and under-deliver on the wall. Both feel the same in a Tuesday meeting; they look completely different in the data.
Track each trade's plan reliability over the job and you get an honest scorecard — not for beating people up, but for planning around reality. If your MEP rough-in crew historically delivers 70 percent of what it commits, then when that foreman promises to finish a floor by Friday, you build your plan around Monday. You're not calling him a liar; you're respecting the pattern the numbers show. And when it's time to award the next project, a documented reliability history is worth more than a good handshake.
Duration actuals: stop guessing and start knowing
Ask ten supers how long it takes to hang and finish drywall on a typical floor and you'll get ten confident answers, most of them wrong in the same optimistic direction. The value of capturing planned-versus-actual durations is that it quietly builds a library of what things really take on your jobs, with your crews, instead of what the estimate hoped for.
Over a couple of projects this becomes a genuine planning asset. You learn that your assumed three-day activity is really four with cleanup and touch-up. You learn which trades' durations are reliable and which ones need a buffer baked in. A concrete example most jobs get wrong: frame-to-rough-in almost always wants a one-to-two-day buffer for punch, cleanup, and inspection that nobody drew on the bar chart — the frame gets called "done" and the electrician shows up to a floor full of debris and a wall that's not ready for layout. Capture that gap once and you'll never schedule those two back-to-back with zero float again.
Schedule health at a glance
All of the above rolls up into a simple question a project manager or owner actually asks: is this job on track or not? A schedule health view earns its place when it answers that in five seconds — trending PPC, count of aging constraints, number of activities behind their planned start, and whether the critical path still has any float left. Not forty KPIs. Four or five that a busy super will actually glance at every morning with his coffee.
The trap here is vanity metrics. A dashboard that's green because it's measuring easy things is worse than no dashboard, because it manufactures false confidence. Health indicators should make the uncomfortable stuff visible, not bury it. If your board is green and your gut says the job's slipping, trust your gut and go fix the metric.
Comparing across jobs — where the real learning lives
The single project teaches you about that project. The portfolio teaches you about how your company builds. When you can line up PPC, variance categories, and constraint patterns across several jobs, patterns jump out that no single super could see. Maybe every one of your multi-family jobs bogs down at the same phase. Maybe one team consistently runs cleaner hand-offs and it's worth finding out what their foreman does differently and spreading it.
This is also the honest antidote to the "every job is unique" excuse. Sure, every job is unique — and yet the same three variance reasons show up on all of them. Comparative analysis is how you separate the genuinely one-off problems from the systemic ones you keep re-buying.
A word on predictive claims
You'll hear a lot of noise about software that predicts your completion date or forecasts delays. Be skeptical, but not dismissive. Real forecasting isn't crystal-ball magic — it's just extending the trends you already have. If your PPC has drifted from 80 down to 55 over six weeks and your constraint backlog is growing, you don't need an algorithm to know you're heading for trouble; you need to act. The useful version of prediction is early warning: surfacing the slide while there's still time to add a crew, resequence, or lean on a sub. Anything that promises certainty about a construction outcome three months out is selling you something.
How to actually make this stick
The reason most jobsite analytics efforts die is that they get treated as a reporting chore instead of a habit built into the weekly rhythm. Keep it lightweight or it won't survive contact with a real job. A few things that work:
- Capture PPC and variance reasons in the same weekly work plan meeting you're already holding — not as a separate exercise. If it takes more than ten minutes it won't last.
- Log the reason for every miss in the moment, while the foreman still remembers why. A reason reconstructed a week later is fiction.
- Review the variance distribution monthly, not weekly. Weekly noise fools you; monthly patterns don't.
- Act on exactly one systemic issue at a time. Find your biggest variance category, kill it, then move to the next. Chasing all of them at once fixes none of them.
This is precisely the kind of work a purpose-built look-ahead tool is meant to carry for you. When PPC, variance reasons, and constraint aging are captured as a natural by-product of building the weekly plan — which is how something like LookAheadWall is designed to work — the analysis assembles itself instead of living in a side spreadsheet nobody updates. The point was never the dashboard. The point is that next quarter's job runs smoother than this one because you finally know, with evidence instead of instinct, what actually goes wrong on your sites and why.
Start with one number. Get honest PPC flowing for a month, add a reason to every miss, and read the pattern. That single habit will teach you more about how your jobs really run than any report a piece of software could hand you — and it turns your schedule from a plan you hope holds into a plan you can steadily make more reliable.