The DCMA Checks That Actually Matter

The DCMA 14-point assessment has become the default health check for CPM schedules, and mostly that is a good thing. It gives owners and contractors a shared vocabulary, it is fast to run, and it catches genuinely broken schedules. But treating all 14 points as equally important is a mistake I see constantly, in both directions: teams burning days chasing a metric that does not matter on their project, and teams waving through a failure that predicts real trouble. Here is how I actually weight them.

The checks that predict failure

Missing logic. This is the big one. An activity without a predecessor or successor is a hole in the network, and a network with holes cannot forecast. Every open end is a place where a delay silently fails to propagate. The 5% threshold is generous; on a well-built schedule this should be near zero, with the exceptions explained.

The critical path test. Not one of the numbered metrics so much as the point of the whole exercise: push a driving activity and watch whether completion moves. If it does not, the schedule's logic is decorative, and every float value and forecast date in the file is suspect. A schedule can pass the other 13 points and fail this one, and it is the only one that is disqualifying by itself.

Negative float. Any negative float means the plan, as logically constructed, does not meet its own dates. Fine as a temporary condition in an update with a recovery plan attached. As a standing condition, it means the schedule has stopped being the plan and become a record of wishful thinking.

Hard constraints. Every mandatory date overrides logic, and every overridden relationship is a place where the network lies to you. A handful of contractual constraints are legitimate. Dozens of them mean someone is steering the dates by hand.

The checks that are context

High float matters, but as a symptom, not a disease. Large float populations usually trace back to thin logic, and the fix is in the relationships, not the float number. High duration is similar: a 60-day activity is not wrong because it is long; it is wrong if nobody can say what being 50% done means. Leads and lags deserve scrutiny because they hide work and distort float, but a small number with documented reasons is normal construction scheduling.

The checks that are mostly noise

The relationship-type ratio (90% finish-to-start) is a style preference dressed up as a rule. Renovation and phased work legitimately runs on start-to-start and finish-to-finish relationships, and forcing them into FS chains to satisfy a percentage produces worse logic, not better. Resource loading is binary: either the contract requires it or it does not. And the BEI and CPLI indices compress rich information into single numbers that get gamed the moment anyone manages to them. I report them because reviewers expect them. I have never made a decision because of them.

What this means for your next review

When a schedule crosses my desk, the first hour goes to the network, not the checklist: trace the critical path end to end, test whether it responds to delay, find the open ends, and read the float pattern for what it says about the logic. The 14-point run comes after, as documentation and as a common language for the review meeting. The checklist is a screen, not a verdict, and the reviewers you should worry about, federal owners' reps especially, know the difference.

If your owner runs the 14 points as a pass/fail gate, know your flags before they do, and walk in with the explanation. A flagged metric with a good reason, stated plainly, builds more credibility than a gamed metric that technically passes.