We went looking for leaks in a $5.2M flooring contractor. We found $434,000.

It's worth exactly what it earns, and here's what it earns: at 16% gross margin, a $6,200 garage job contributes about $990 before a dollar of overhead. Mobilization, a two-visit install, and callback risk are nearly the same as a job three times the size.

Fill work has real value — but only if it's priced as fill work, not chased as a growth segment. The fix isn't dropping residential. It's (1) repricing it to a 25% floor, (2) taking it only inside a 30-minute radius, and (3) stopping paid advertising for it entirely. Let it come to you at a price that pays; spend the effort where margins are double.

Gross margin by segment, monthly — not blended. And its sharper cousin: gross profit per crew-day by segment. Industrial here produced ~$1,860 of gross profit per crew-day; residential produced ~$610. Same crew, same day, three times the money. When you see that number, scheduling decisions make themselves.

Because material markup and the industrial segment were quietly subsidizing it. The company was profitable — at 4.1% net. The point isn't that $34 bankrupts you; it's that every labor hour ships $5.90 of margin out the door, and the healthy segments absorb the loss so nobody sees it.

The fix took one afternoon: a burden worksheet per role, refreshed twice a year, feeding the estimating template automatically. New floor rate: $47/hour (loaded cost plus target margin). Win rates on industrial bids didn't move. Residential win rates dropped — which, per Finding 01, was the goal.

Because the system made flagging expensive. A change order meant: foreman calls the PM, PM emails the estimator, estimator prices it in the spreadsheet, someone types it into an email for the GC, and — if everyone remembers — it reaches the QuickBooks invoice weeks later. Five handoffs. On a job site, five handoffs means zero.

The fix was making capture cheaper than absorption: a foreman photographs the condition, taps a CO request with a price range on their phone, the client approves by text, and it lands on the next progress invoice automatically. Capture went from roughly 40% of eligible events to 91% in one quarter. The crews didn't change. The friction did.

Two numbers. CO capture rate: change orders invoiced ÷ change events logged (crews log events in seconds if it's one tap). And estimated vs. actual hours per job — when actuals consistently run over with no CO attached, that overage is your unbilled work, priced at your loaded labor rate. This company ran 6.4% over estimate on average. That's not an estimating problem; that's a billing problem wearing an estimating costume.

Correct — and that's precisely why the budget drifted residential: residential is the only segment you can buy with a credit card. Industrial spend looks like: memberships and actual attendance where facilities managers are (ABC, IFMA chapter), lunch-and-learns for GC estimating teams, case-study one-pagers per vertical, a referral structure for the two trades that see floors before you do (concrete polishers, mechanical contractors), and disciplined follow-up on every plant manager who ever said "not this year."

It's slower and less measurable per-click. It also produced jobs at $71 acquisition cost against $412 from lead marketplaces — for work at double the margin.

Three things, in order of leverage. First, the 9-day lag is entirely yours — it existed because progress billing required the PM to reconcile hours from the time app against the estimate spreadsheet, then hand totals to the office. Automating that turned lag into same-week invoicing: ~$106K of working capital released permanently, no conversation with any client.

Second, billing terms on new work: deposit on mobilization, materials billed on delivery (resin isn't cheap and it's yours the moment it ships), and progress billing weekly instead of monthly. GCs push back far less than owners fear — the ones who push hardest are usually the ones who were going to pay slow anyway, which is worth knowing at bid time.

Third, chase retainage like it's revenue, because it is. $186K held with no owner. Assigning one person a monthly retainage-release checklist recovered $61K of aged retainage in the first 90 days.

Five years ago, fair. The economics changed: a focused operating platform — estimating with your burden rates, CO capture from the field, progress billing pushed to QuickBooks — is now a mid-five-figure build, not a seven-figure IT project. The question isn't "can we afford custom." It's whether $90K over three years to run your process beats $160–210K to rent someone else's.

And it's not for everyone — under ~$2M revenue or with dead-simple workflows, an off-the-shelf tool is genuinely the right answer, and we say so in diagnostics. The math turns at about this company's size, when per-seat pricing and process mismatch start compounding.