Every job I've run generated more numbers than anyone looked at. Daily reports piled up, timesheets got approved and filed, the schedule got updated on Friday and printed. Nobody was doing anything with it. The data was there; the analysis wasn't. Then software started collapsing all that scattered record-keeping into dashboards, and the honest question became: which of these numbers actually change a decision you make on Monday morning, and which are just decoration on a report nobody reads past page two?
That's the frame worth holding onto. Analytics in construction management software are only worth the click if they tell you something you'd otherwise learn too late. Below are the metrics I've found earn their keep, how to read them without fooling yourself, and where they lie to you if you're not careful.
Start With Plan Reliability, Not Percent Complete
The single most useful number a scheduling tool can hand a superintendent is not how far along the job is. It's how often the work you committed to last week actually got done. In Last Planner terms that's Percent Plan Complete, or PPC: of the tasks you promised on the weekly work plan, what fraction were finished by the date you said?
Overall project percent-complete is a comfort number. It goes up no matter what, and it hides a job that's quietly falling apart. PPC does the opposite. If you're running around 80 percent PPC, your planning is honest and your crews are hitting their commitments. If you're stuck at 50, half of what you plan every week isn't happening, and the reason is almost never laziness. It's that the work wasn't really ready when you scheduled it, or a constraint you didn't see took it out.
Read the trend, not the single week. One bad week during a storm means nothing. Six weeks trending down means your look-ahead process has drifted into wish-listing, and you need to tighten how you screen work before it lands on the plan. Good short-interval scheduling software calculates PPC for you and shows the curve, so you're not doing it by hand in a spreadsheet at 6 a.m. LookAheadWall builds those weekly commitments visually by location and trade, which makes the follow-up conversation concrete: not "the schedule slipped," but "these three activities in Building B didn't close, here's why."
Track the Reasons Work Didn't Happen
PPC tells you something broke. The reason codes tell you what. This is the analytics feature most teams skip and later wish they hadn't. Every time a planned task doesn't get done, you tag why: prerequisite work incomplete, material not on site, RFI unanswered, inspection not passed, manpower short, weather, changed conditions, design conflict.
Do that for a month and a pattern jumps out of the data that no single daily report would ever reveal. Maybe 40 percent of your misses trace back to one thing — submittals approving late, or one sub who never has the crew he promised. That's no longer a gut feeling you can't defend in a meeting. It's a count. When you can walk into an owner meeting and say "we've lost eleven planned activities in the last five weeks waiting on the mechanical submittal package," the conversation changes. Constraint analysis is worth more than almost any other analytic precisely because it converts complaints into evidence.
Labor Productivity: Useful, and Easy to Misread
Productivity analytics compare planned hours or units against actual — square feet of drywall hung per labor hour, linear feet of pipe set per day, whatever your unit of work is. Done right, it's early warning that a trade is underwater before the budget shows it.
Here's where people fool themselves. Productivity numbers are only as good as the quantity tracking underneath them, and field quantities are notoriously sloppy. If the foreman guesses "we're about 60 percent done" every Friday until the last week when it suddenly becomes 100, your productivity curve is fiction. A couple of rules of thumb keep you honest:
- Trust trends over absolute numbers. A crew running consistently at a given rate that suddenly drops 30 percent is telling you something real, even if you don't fully trust the baseline.
- Watch for the end-of-job cliff, where reported progress catches up all at once. That's a data-collection problem, not a productivity story.
- Normalize for the obvious variables. Productivity on a wide-open floor plate is not comparable to the same crew fighting congestion in a mechanical room. Compare like to like.
- Feed the real numbers back into estimating. The best reason to measure productivity isn't to beat up a foreman this month — it's so next year's bid is built on what actually happened, not on a number someone copied from the last job.
Subcontractor Reliability, Measured Instead of Remembered
Ask any veteran super which subs are reliable and you'll get a confident answer built on memory and the last thing that went wrong. Memory is biased toward the recent and the dramatic. Commitment data isn't. If your look-ahead process captures who committed to what each week, you can compute a completion rate per trade partner across the whole job — and across multiple jobs if you're running a portfolio.
That number is gold at buyout. A sub who quietly hits 90 percent of his weekly commitments is worth more than one who bids two points lower and blows a third of his dates, because the reliable one isn't dragging four other trades behind him every time he slips. Performance-based selection only works if you have the performance data, and the weekly work plan is where it lives. The point isn't to build a blacklist. It's to know before award which relationships need a tighter leash and more lead time.
Forecasting: Where the Schedule Is Actually Headed
Backward-looking metrics tell you what happened. The more valuable question is where you'll land if nothing changes. Predictive analytics use your recent completion rate to project a finish date that reflects how the job is really moving, not the baseline someone drew a year ago in the trailer.
The value here is lead time. If your current pace projects a completion three weeks past the contract date, finding that out in month four gives you room to add a crew, re-sequence, or renegotiate. Finding it out in month eleven gives you a claim and a fight. Treat the forecast as a smoke detector, not a crystal ball — it's directional, and it assumes the future looks like the recent past, which it won't exactly. But a projection quietly drifting later, week over week, is one of the earliest honest signals a job is in trouble.
Reading Dashboards Without Being Fooled By Them
A dashboard's job is to get the important thing in front of you in ten seconds — from the truck, on your phone, before the morning huddle. That's a real benefit; nobody digs through raw tables at 6 a.m. But a clean dashboard is also the easiest place to lie to yourself, so a few habits matter.
- Know what's behind every gauge. A green "92% complete" tile means nothing if it's counting planned progress instead of verified installed work. Ask what feeds each number before you trust it.
- Green isn't the goal. A board that's all green usually means the metrics are too soft to catch anything. You want indicators sensitive enough to turn yellow when something's actually off.
- Prefer leading indicators to lagging ones. Constraints unresolved for next week's work is a leading indicator — you can still act. Recordable incident rate is lagging; by the time it moves, the thing already happened.
- Watch the plumbing. Analytics are only as current as the field data behind them, and the field is busy. If crews aren't updating status, your dashboard is a confident picture of last Tuesday. Half of getting value from analytics is keeping the input honest and easy, which is exactly why a tool crews will actually update beats a powerful one they won't.
The Metrics That Round Out the Picture
A few others deserve a place on the board without needing their own sermon:
- Constraint lead time — how many days before scheduled work your constraints get resolved. Resolving them the morning work is supposed to start means you're firefighting, not planning. Pushing that lead time out to a week or two is the whole point of a rolling look-ahead.
- RFI and submittal turnaround — response times on the information flow. Slow turnaround upstream shows up as missed field commitments downstream two weeks later, and the analytics let you connect the two instead of arguing about them.
- Rework and deficiency rates — punch and inspection failures by trade. A rising rework rate on one trade is cheaper to catch in the data than in a warranty claim after closeout.
- Safety leading indicators — observations, near-misses, and pre-task planning completion, not just the incident count. Lagging safety numbers only move after someone gets hurt.
Turning Numbers Into a Decision
Analytics don't manage the job. You do. The discipline that separates teams who get value from a dashboard from teams who just have one is simple: every metric you watch should be tied to an action you'll take when it moves. PPC drops two weeks running, you audit how work gets screened onto the plan. One constraint type dominates your misses, you fix that upstream process. A sub's completion rate sags, you shorten his leash and lengthen his lead times.
The tooling — short-interval scheduling, weekly work plans, trade-flow sequencing in something like LookAheadWall — exists to make that loop cheap enough that you actually run it every week instead of once a quarter when the owner asks. Data-driven doesn't mean drowning in reports. It means a handful of numbers you trust, checked on a rhythm, each one wired to something you'll do differently. Get that loop turning and the job starts telling you where it's going while you can still steer. Ignore it, and you'll read the same story in the final schedule analysis, after it's too late to change the ending.