08/26/2026
The job your team reviews every single week faded six-tenths of a point.
The two nobody reviews lost $1.67 million.
That's not a discipline problem. It's an attention-allocation problem, and it's structural: manual review is finite, so it goes to the biggest number on the page. The $24M interchange gets a weekly forecast. The $6.4M widening gets one at close.
Here's the part that makes it expensive in heavy civil specifically.
CFMA's 2025 Benchmarker puts heavy construction net income before tax at 8.3% of revenue. Standard retainage is 5–10% of contract value.
Read those two numbers together. The retainage balance on a typical civil job is worth 60% to 120% of the profit that job will ever produce.
So profit fade doesn't reduce your margin. It consumes a balance you haven't collected, can't easily influence, and won't see until final acceptance.
Which is why fade and retainage aren't two topics. They're one number:
Margin at Risk = retainage held ÷ current estimated gross profit at completion
Every point of fade pushes it up automatically — fade shrinks the denominator while the retainage numerator sits exactly where it was. Past 100%, the owner is holding more money than the job has profit left in it.
And none of this is an ERP failure.
Sage Intacct Construction's WIP feature does what it says: automates the monthly WIP statement, posts the over/under billing entries, reports margin movement. But percent complete is cost-to-date divided by estimated cost at completion — a fact divided by an opinion. The fact posts continuously. The opinion gets revised once a month, by a PM, during close, across a dozen jobs. Often in a system that isn't the ERP at all.
If one input arrives monthly, the output is monthly. However fast the ERP runs.
Full breakdown — including the four-job table above and how to read Margin at Risk per job: https://bit.ly/4gB5IF1
Profit fade is a reporting-latency problem, not an estimating one. How real-time WIP and retainage visibility catches margin erosion while you can still fix it.
