DCMA-14 Metric 3 — Lags: When They're Valid and When They're Gaming

A lag is valid in a construction schedule when it represents a real waiting period with no work, such as concrete curing or a fixed review period, or a genuine overlap, and when it is documented. It is not valid when it holds back float or stands in for work that should be its own activity. The DCMA 14-point assessment caps lags at 5% of logic links.

This article, metric 3 in the series, separates the cap from the judgment call: the 5% limit is a number, but whether a given lag is legitimate is a test you apply link by link.

What does the DCMA lag metric measure?

The metric is:

Lag % = logic links with a positive lag ÷ all logic links

The DCMA pamphlet says the result should not exceed 5%. It counts links, not tasks, so a schedule with 800 logic links can carry up to 40 lagged links before reaching the limit.

The pamphlet's rationale is short and pointed: "per the IMS DID, lag should not be used to manipulate float/slack or to restrain the schedule." A lag is acceptable as a modeling tool. The concern is that it can quietly decide the dates.

Leads, the negative version, have their own metric with a goal of zero. See why leads are flagged. The two metrics are separate counts, and a link is one or the other.

What is the test for a valid lag?

Use a three-part test. A lag passes if the answer to all three is yes.

1. Is it passive time? No crew, equipment or cost is attached to the wait. Concrete curing before stripping forms, paint drying between coats, a fixed agency review period and fabrication lead time with no further input from the contractor all qualify.
2. Is it documented? The relationship note (or the schedule narrative) states the basis: the spec section, the mix design, the review period in the contract, the vendor quote.
3. Does it stay out of the float game? The value is the real duration of the wait, not a number chosen to land an activity on a target date or to absorb float.

If any answer is no, convert the lag to an activity or remove it.

Common valid lags in construction

Where the wait is long, uncertain or needs tracking, an activity is usually better. Mosaic Projects' white paper notes that lags "hide detail" and "cannot be statused like normal activities," and that they should typically be replaced with activities.

What does gaming with lags look like?

Lags become a problem when they carry judgment that should be visible. Four patterns recur.

Placeholder for work. A 10-day lag between rough-in and drywall stands in for inspections, rework and punch. None of it can be progressed or resourced, so the schedule can't show whether the work is happening.

Float consumption. A scheduler adds a 6-day lag to push a successor off a milestone so that float looks adequate, or removes it to make a path look critical. Dates change, nothing physical does.

Compensating for a hard constraint. A lag is added so that the logic lines up with a constrained date instead of removing the constraint. See hard constraints for why that pairing hides problems.

A large lag on a link that should be an SS or FF. A 30-day FS lag between two phases of work that overlap in practice moves the overlap into an unobservable number.

GAO's Schedule Assessment Guide, GAO-16-89G points the same way in its sequencing practice, which says date constraints and lags should be minimized and justified, which is the same standard the three-part test applies.

How do I find and review lags in P6?

Oracle's Schedule Check Report covers lag-related checks, and the negative lags section reports lags in hours. For positive lags, review the relationships and sort by lag value, then calculate the percentage yourself. Menu names vary by version, so confirm the fields in yours.

A review pass is a table. Assume the following links came from a mid-rise building schedule (illustrative example):

With, say, 800 links in the schedule and 22 lagged, the metric is 2.75%, under the cap. The lag count is not the finding. The second row is.

Worked example: a lag converted to an activity

Illustrative example with assumed numbers. Two links on a floor-by-floor sequence, before and after converting a 10-day lag labeled "inspections and rework" into an activity:

The dates and float do not change, which is the point: the conversion is neutral for the schedule and gives the 10 days an owner, a status and a place in the resource plan. If field conditions later show the inspection takes 6 days, you update one activity instead of explaining a lag.

Now suppose the 10 days was padding. The conversion exposes it, because nobody can describe the work in activity I300. Removing it pulls drywall 10 days earlier, and the float it had absorbed shows up elsewhere.

The SS lag calculation option

P6's scheduling options matter for progressed work. Oracle's progress options include a setting for how start-to-start lag is calculated, from the predecessor's actual start or its early start. If you carry SS lags on in-progress activities, your choice changes the successor's dates, so state which one you use. The related relationship types and retained-logic behavior are covered in the relationship types article.

How should I fix a lag that fails the test?

Choose by what the lag stands for:

• Work or testing becomes an activity with a duration, a crew and a status. The successor links finish-to-start from the new activity.
• A long wait becomes a procurement or review activity so it can be tracked and delayed on its own.
• A genuine overlap stays an SS or FF link, with the lag matching the field offset and a note.
• A padding value is removed, and the float it was hiding is allowed to show.

Reschedule after each batch of changes, because removing a lag moves dates. Expect high-float counts to shift too. The high float article explains why padding can inflate float that DCMA later flags.

What will a reviewer ask about lags?

Three things, usually in this order. What is the percentage, and what denominator did you use? Which lags are the largest, and what is each one's basis? And did lag counts or values change since the last update?

A defensible answer is a one-page register: link, value, basis, owner. Keep it with the schedule narrative and update it whenever a lag is added or changed, so the next reviewer starts from your explanation instead of a blank list. Where the project has more than one scheduler, assign one person to approve every new lag. DCMA's own text sets the tone. The metric is a prompt for questions, so a documented answer ends the conversation.