The metric that hangs on every jobsite trailer wall
Walk into any construction trailer and you'll find a percent-complete number taped to the wall. "We're 62% done." Somebody in the Monday owner meeting will nod at it like it means something. It doesn't, not by itself. I've watched a job sit at 62% for three weeks while the schedule quietly slid a month to the right, because percent-complete measures how much concrete you've poured, not whether you're going to finish on time.
Good scheduling metrics tell you where you're losing the job before the job is lost. Bad ones give everybody a comfortable number to point at while the float bleeds out. The difference isn't the software you use to track them — it's knowing which handful of numbers actually predict a miss, and which ones just make the report look busy. This is about the second kind: the small set of measurements a superintendent can read in ten seconds and act on before the next pour.
Start with reliability, not percent-complete
The single most useful number in short-interval scheduling isn't on most owner reports at all. It's Percent Plan Complete — PPC — and it comes straight out of the Last Planner discipline. Every week you commit to a set of tasks in your weekly work plan. At the end of the week you count how many you actually finished, all the way done, no "90% there." Divide completed by committed. That's your PPC.
A crew running 50% PPC is finishing half of what it promised. That's not a productivity problem — it's a planning-reliability problem, and it's the leading indicator that your three-week look-ahead is fiction. Healthy jobs live in the 70–85% range. Below 65% and your subs have stopped believing the schedule, so they stop staffing to it, and now you're in the death spiral where the plan and the field have divorced.
The gold is in the misses, not the number. Every incomplete task gets a reason code: prerequisite work wasn't done, materials didn't show, RFI still open, manpower short, weather, inspection failed, changed priorities. Tally those reasons over a month and the jobsite tells you exactly where it's broken. On one podium job we ran PPC for six weeks and 40% of our misses traced back to one thing — the deck wasn't being cleaned and swept ahead of layout, so the layout crew kept getting bumped. No metric on the master schedule would have surfaced that. The reason codes did.
Schedule health: float, variance, and the critical path
Reliability tells you if the plan is trusted. Health metrics tell you if the plan itself is in trouble. Three worth watching:
- Total float / float erosion. Float is your cushion — the days an activity can slip before it drags the finish date. Watch the trend, not the snapshot. If your critical-path activities had 12 days of float a month ago and 3 days now, you're consuming float faster than you're spending time, and that's a red flag even if nothing looks "late" yet. Float that goes negative means you're already committed to missing the date unless you compress.
- Critical path length and stability. A critical path that jumps to a completely different chain of work every update isn't a healthy dynamic schedule — it usually means your logic is thin or your durations are guesses. When the critical path keeps moving, trust it less and go walk those areas yourself.
- Schedule variance in days, not dollars. Earned-value schedule variance gets reported in money, which is nearly useless to a foreman. Convert it to time: are we ahead or behind, and by how many working days on the activities that matter? Days are what you can act on.
Here's the rule of thumb I hand younger supers: a metric you can't tie to a decision you'd make this week is a metric you're collecting for the report, not for the job. Float erosion passes that test. "Overall percent complete" usually doesn't.
Progress: measure completion, not effort
Progress metrics are where jobs lie to themselves the most. The trap is measuring effort — hours burned, activities "started," material installed — and calling it progress. What matters is activities actually completed against activities planned to complete in that window, and milestones hit against milestones due.
Count a task done only when it's inspection-ready and the trade behind it can start. A wall that's framed but not fire-caulked isn't done for the drywaller. In-wall rough-in that hasn't passed rough inspection isn't done for the insulator. If you let partial credit creep into your progress numbers, you build a backlog of near-done work that all comes due at the same moment — usually the week before the drywall hangers show up, and now three trades are stacked in one room.
Milestone hit-rate is the honest cousin of percent-complete. Did you top out the deck by the date you committed? Did you get the building dried-in before the finishes trades were scheduled to load in? Milestones don't give partial credit, and that's exactly why they're worth tracking.
Trade-flow and coordination metrics
The metrics that separate a smooth job from a brawl are the ones about hand-offs between trades. In a location-based schedule — the kind you build when you're thinking in terms of areas and floors flowing through a sequence of crews — you want to watch:
- Constraint-free look-ahead percentage. Of the tasks coming up in your two-to-three week window, what fraction are actually ready to go — materials on site, prior work complete, RFIs answered, permits and inspections cleared? This is the "make-ready" number. If only 40% of next week's plan is constraint-free, you don't have a plan, you have a wish list. Drive this number up in the coordination meeting and your PPC follows it up a week or two later.
- Buffer consumption between trades. Build deliberate buffers into your trade flow — frame-to-rough-in usually wants a day or two for cleanup, punch, and inspection before the next trade loads in; concrete wants its cure time before you stack load on it. Track whether crews are eating those buffers. When trades start overlapping in the same location because everyone's behind, quality craters and you start paying for rework you can't see yet.
- Stacking / crew density by area. How many trades are working the same zone at once? Two is coordination. Four is a fistfight over the same 400 square feet, and productivity drops for all four. This is a metric you feel before you measure it, but measuring it forces the conversation.
This is exactly the coordination a visual look-ahead tool is built to expose — when your weekly work plan is laid out by location and the trade-flow connections are drawn in, an area where four crews collide shows up as a traffic jam on the wall instead of a surprise on Tuesday morning. LookAheadWall does this because that's the whole point of a location-based plan, but the discipline matters more than the tool: if you can see the collision three weeks out, you can move somebody.
Quality metrics for the schedule itself
A schedule can be "on time" on paper and still be garbage. A few checks tell you whether the schedule is trustworthy:
- Update timeliness. A look-ahead updated once a month is a history book. Short-interval scheduling wants a weekly cadence, minimum. If your last actual-progress update is nine days old, every metric you're pulling from it is nine days stale.
- Open constraints aging. How long have RFIs, submittals, and missing materials been sitting on your constraint log? A constraint that's been "open" for 30 days isn't a constraint anymore, it's a wall you've decided to hit.
- Logic density. Activities with no predecessor or successor — "dangling" tasks — are activities that float free and lie about the finish date. A schedule where a third of the tasks have open ends isn't a network, it's a to-do list with dates.
Set targets, then let the alerts do the watching
A metric with no target is trivia. Decide up front what "good" looks like: PPC above 75%, constraint-free look-ahead above 80%, zero critical activities with negative float, updates no older than seven days. Now the numbers have teeth — you're not asking "is 68% good?" every week, you're asking "why are we under 75, and what's the top reason code?"
Then set thresholds that shout at you. You don't have time to stare at a dashboard, and you shouldn't have to. The value of automated alerts is that they surface the two or three numbers that just went the wrong way and leave the twenty that are fine alone. When float on a critical chain crosses into single digits, or a milestone's forecast date slips past its committed date, that's what should reach you — not the whole balanced scorecard every Monday. Good scheduling software earns its keep here by watching thresholds so you can spend your attention in the field, where it belongs.
Read the trend, distrust the snapshot
One last habit that matters more than any single metric: every number you track, track its direction. A 70% PPC that climbed from 55% over a month is a job getting healthy. A 70% that fell from 85% is a job coming apart, and it's the more dangerous of the two because the snapshot looks fine. Float, hit-rate, constraint-free percentage — plot the trend over the last four to six weeks and the trajectory tells you more than today's value ever will.
Keep the set small. Five or six metrics you actually act on beat thirty you glance past. PPC and its reason codes to see if the plan is trusted, float erosion and milestone hit-rate to see if the date's in danger, constraint-free look-ahead and buffer consumption to see if the trades are going to collide. Read them weekly, read the trend, and act on the worst one before the next plan goes out. That's the whole game — not the reporting, the acting. The metrics are only worth what you do about them by Friday.