Every subcontractor management platform on the market will happily show you a dashboard full of gauges and trend lines. That's the easy part. The hard part — the part nobody sells you — is knowing which three numbers actually predict whether your framer shows up Monday with the right headcount, and which twenty are just decoration that make the software demo look impressive. After twenty years of running jobs, I've learned that most "analytics" get printed once, admired in a meeting, and never change a single decision. This is about the handful that should.
Start With the Question, Not the Chart
The mistake I see superintendents make is opening the analytics tab and asking "what does this tell me?" Backwards. You already know the questions that keep you up at night: Is drywall going to hit the north wing next week or slip again? Which of my subs is quietly running two other jobs and starving mine of manpower? Where do my inspections keep failing and costing me a re-inspection day? Good analytics answer questions you already have. If a metric doesn't map to a decision you'll actually make, it's noise.
So before you fall in love with a heat map, write down the five decisions you make every week that hurt when you get them wrong. Manpower commitments, sequence calls, what to escalate to the GC, who to put on notice, where to add a buffer. Then find the number that informs each one. Everything else, ignore for now.
PPC: The One Number That Earns Its Keep
If you take one thing from this article, take this: track Percent Plan Complete and track the reasons for the misses. PPC is dead simple — of all the tasks you committed to in this week's work plan, what percentage actually finished as promised. Twenty tasks planned, fourteen done and done right, that's 70% PPC. It's a batting average for reliability, and it's the single best leading indicator of whether your schedule is real or fiction.
Here's what the raw number is good for and where crews get it wrong. A PPC of 100% doesn't mean you're a genius — it usually means you sandbagged the plan and committed to less than the crew could do. Consistent 90%+ often hides padding. The healthy range on a well-run job tends to sit in the 65–85% band, climbing as the team matures. What matters more than the number is the trend and the reason codes behind the failures.
That reason data is the gold. When a committed task doesn't complete, the app should make you pick why: prerequisite work wasn't done, materials weren't on site, manpower short, RFI/design hold, weather, changed priorities, inspection failed. Log that every week and after a month you'll see the pattern plain as day — maybe 40% of your misses trace to "prerequisite not complete," which tells you your sequencing and trade-flow handoffs are the problem, not your subs' effort. That's a scheduling fix you own, not a beating you give the trades. I've watched a super blame the electrician for six weeks running when the variance data was screaming that the framer's late close-in was the real constraint every single time.
Reliability by Trade: Who Actually Keeps Their Word
Roll PPC up by subcontractor and you get something more useful than any reference check: a made-versus-kept commitment rate per trade, on your job, in the last 90 days. This is what tells you which subs to trust with a tight window and which ones need a buffer baked in.
Use it like this. A sub sitting at a 55% commitment-reliability rate isn't necessarily lazy — but you plan around them differently. You give them earlier notice, you sequence their work with slack in front and behind it, and you make their prerequisites bulletproof before you ever let them commit. A sub running 85%+ earns tighter sequencing and more trust. When you're building the look-ahead, this is the difference between a plan that holds and one that unravels by Wednesday. Software that ties each trade's history to the weekly work plan — which is exactly the point of a tool like LookAheadWall — lets you see that reliability score right next to the task you're about to commit, when the decision is actually being made, instead of in a post-mortem after it's blown up.
Variance Trends Beat Snapshots Every Time
A single week's data lies. Any trade can have a bad week — a guy quits, a truck breaks down, a slab pour runs long. The value is in the slope. Pull six to eight weeks of variance and the story stops being anecdote and becomes signal.
- Rising misses on one trade usually mean they're pulling manpower to another job. That's your cue to have the conversation now, not after they've abandoned you.
- Misses clustered on one location — same wing, same floor — point to an access, staging, or prerequisite bottleneck, not a labor problem.
- Misses spiking every time two trades share a zone flag a coordination gap you can fix by re-sequencing or splitting the area.
- A whole-job PPC that's been sliding for a month is the earliest warning you'll get that the schedule is drifting from reality. Catch it here and you've got weeks to recover; catch it on the master schedule and you're already late.
Compliance Data You Can Actually Act On
Insurance, licensing, and safety-doc tracking is where subcontractor management software genuinely earns its cost, because the failure mode is expensive and boring. A lapsed COI on the sub who's about to start overhead work is the kind of thing that gets a job shut down or, worse, leaves you exposed on a claim. The analytics that matter here aren't fancy — they're a clean expiration timeline and an exception list.
What you want is a single view of every active sub with their insurance expiration, license status, and required safety submittals, sorted by soonest-to-lapse. Set the alert at 30 days out, not the day of. The useful "analytic" is embarrassingly simple: how many days until this becomes my problem, and who's already out of compliance and shouldn't be on my site tomorrow. Don't overthink it. A super who's ever had a trade walked off by the GC's risk manager over a stale certificate will tell you this plain report saves more grief than any predictive model.
Where "Predictive" Is Real and Where It's Snake Oil
Vendors love the word "predictive." Be a skeptic. Real, useful prediction on a jobsite is mostly just honest extrapolation of trends you can already see: if a trade's PPC has dropped four weeks running and their variance reasons all say "manpower," you can predict with confidence they'll miss their next milestone. You don't need a machine-learning engine for that; you need the trend in front of you and the nerve to act on it a week early.
Be wary of black-box "completion probability" scores that can't show their work. If the tool can't tell you why it thinks you'll finish late — which tasks, which constraints, which trade — it's a magic 8-ball with a nicer font. The prediction you should trust is the one you can trace back to specific tasks in the look-ahead and specific reasons in the variance log. That traceability is the whole game.
Communication and Payment Data: The Quiet Tells
Two datasets get overlooked because they don't feel like "schedule" data, but they're early-warning gold. Response time to your RFIs and daily reports tells you who's actually engaged — a sub whose foreman was answering in an hour and now goes dark for two days is signaling something, usually that your job just dropped down their priority list. That change shows up in the data before it shows up on the wall.
Payment patterns tell the other half. A sub whose pay applications suddenly get sloppy, or who starts pushing hard for early release, may be having cash-flow trouble — which is often the last warning you get before they slow-walk your job or pull their best crew. None of this is about spying; it's about noticing the change in behavior early enough to have a real conversation while you still have options.
A Weekly Analytics Routine That Takes Fifteen Minutes
Analytics only matter if they change what you do. Here's the routine I'd run, and it fits inside your coffee:
- Open last week's PPC and the variance reason breakdown. One number, one chart. Ask: is the trend up, flat, or down, and what's the top failure reason?
- Glance at trade reliability. Any sub whose score dropped this week gets a phone call today, before you build the next plan.
- Check the compliance exception list. Anyone inside 30 days of a lapse gets chased now.
- Look at the top variance reason and ask if it's yours to fix. If it's "prerequisite not complete," fix your sequencing before you blame anyone.
- Carry those three or four findings straight into building next week's look-ahead — tighter commitments where reliability is high, buffers where it isn't.
That last step is the whole point. Data that doesn't flow back into the next weekly work plan is just accounting for a funeral. The reason a connected tool matters is that the reliability, variance, and compliance signals sit right where you're making commitments, so short-interval scheduling stops being a guess and starts being informed by what actually happened last week.
Garbage In, Dashboard Out
One honest warning: none of this works if the underlying data is junk. If your foremen aren't logging why tasks slipped, your reason codes are empty and your trends are lies. If commitments are entered loosely — "yeah, we'll get to that area" instead of a real, sized task — your PPC is meaningless. The analytics are only as good as the discipline in the weekly plan feeding them.
So invest in the input before you obsess over the output. Get the crew committing to specific, sized, location-based tasks. Get honest about why things miss — no blame, just the reason. Do that for a month and the analytics start telling you the truth about your job. Skip it, and the fanciest predictive dashboard on the market will just draw you a very confident picture of nothing.
Use your data, but use the three or four numbers that change a decision, and let the rest sit in the tab where you found it. A superintendent who tracks PPC, trade reliability, variance reasons, and compliance dates — and actually feeds them back into next week's plan — will run a tighter, more predictable job than one drowning in twenty charts nobody reads.