The DCMA 14-point assessment, counted properly

Run the same programme through two schedule-checking tools and you'll often get two different DCMA scores. Neither tool is necessarily wrong. They're usually counting different things.

The population problem#

The DCMA 14-point assessment is a set of ratios: open ends as a share of activities, leads as a share of relationships, and so on. Change the denominator and the answer changes.

The question is which activities and relationships to count. On a programme that's half built, there are two very different answers:

• Every activity. Finished work is included. Its logic, lags and constraints get counted.
• Incomplete activities only. Only the plan that's left is measured.

DCMA's intent is the second. Finished work can't be re-planned, so its open ends and constraints say nothing about whether the remaining plan is sound. Counting it dilutes real problems in the work ahead, or punishes a plan for history nobody can change.

The same goes for relationships. A link into finished work no longer drives anything, so checks 2 to 4 (leads, lags and relationship types) should count only relationships whose successor is still open.

Three more counting rules that matter#

Lags: any positive lag counts. DCMA's check 3 counts every positive lag, not only long ones. Some tools set a threshold, such as five days. Both are defensible, but they aren't the same check.

Invalid dates use forecasts, not plans. A forecast in the past means the file wasn't rescheduled after statusing. The forecast is P6's early date from the last schedule run. Planned dates aren't forecasts, and using them produces false failures.

CPLI needs the right float. The float in CPLI should be the float on the path to completion, not the lowest float anywhere. An intermediate key date can set its own negative float. It's real, but it isn't the completion path.

So which score is right?#

The one whose rules you can state. A useful DCMA report says, for every check, what it counted and against what threshold, and lets you see the activities behind the number. If a reviewer's score differs from yours, switch the counting basis and compare like with like before arguing about the programme.

The checks that aren't defects#

Four of the fourteen measure performance, not construction: negative float, missed activities, CPLI and BEI. A late programme fails them honestly. That's not something to fix in the file; it needs recovery, or a notice under the contract. Editing constraints until negative float disappears doesn't make the job earlier. It only makes the programme less truthful.