Lags, Open Ends and a Crew Cap on a Plant Shutdown Overhaul

Why blanket deletion is wrong

• A lag is a duration with no work attached. Deleting all lags produces an earlier, fictitious finish and removes constraints that are real - the schedule becomes optimistic rather than accurate.
• The construction manager is effectively buying a date with a modeling change, which is the opposite of what a baseline is for.

Triage the lags by what they represent

• Regulatory or contractual waits (permit approval): this is real work with a queue, a reviewer and a risk of rejection. Replace the 10-day lag with a permit-approval activity of realistic duration, ideally on the same calendar as the authority, so it can be tracked and can slip visibly.
• Resource or physical waits (cure, delivery, mobilization): model as an activity if effort is involved, or keep as a lag only if genuinely passive - and document the basis.
• Arbitrary padding: delete, but through the estimation discussion, not unilaterally.

Control point

• Any of this changes the baseline before sign-off, so it is estimation, not change control. Once baselined, the same edits require a formal change.
• Watch for lags that mask a missing relationship - a lag is often a symptom of logic the planner did not finish.

Open ends (no successors)

• CPM calculates float backwards from the project finish. An activity with no successor has no path to the finish, so its float is undefined or artificially large, and it cannot contribute to the longest path.
• Consequences: the critical path is incomplete, negative float may be created or hidden, and activities that can slip without consequence are indistinguishable from activities that drive the end date.
• Every activity except the project finish milestone needs at least one successor, even if that successor is a summary or a project-completion milestone.

Start-to-start with no lag

• SS with zero lag means the successor may begin the same day as the predecessor. If the intent was 'start after some progress,' the lag is missing and the successor is scheduled too early.
• Verify intent with the construction manager for each of the 9, then either add the intended lag or convert to FS where the sequence is genuinely finish-to-start.

Why before baseline

• A baseline with broken logic cannot be used for float analysis, time-impact analysis or claims later. Fixing it after the baseline means either re-baselining or working forever with a corrupt reference.
• Run a structured schedule health check (open ends, lags, leads, hard constraints, high-duration activities, float distribution) and record the results with the baseline.

Distinguish the two techniques

• Smoothing uses the float that already exists: tasks are shifted within their late dates, peaks are flattened, and the completion date does not move. Try this first - it costs nothing in schedule.
• Leveling respects the cap by delaying tasks beyond their float; it extends the project. The 3-week extension is the price of the 22-crew constraint, and it is a true result, not a software artifact.

Then test the constraint rather than only accepting it

• Is 22 a physical limit (site access, supervision ratio) or a budget limit? If budget, the comparison is explicit: 3 weeks of production loss versus the cost of a second crew or a temporary subcontract.
• Can the work be re-sequenced by area so that electrical work follows civil release in a smoother wave, reducing the peak without delaying completion?
• Confirm the leveling is not splitting activities, respecting the correct calendars, or producing negative float from a bad constraint - those produce false extensions.

Decide and record

• Present the extension, the smoothing-only alternative, and the added-crew alternative with cost and risk. The shutdown window's production-loss cost is the yardstick; whichever option is chosen, the basis goes into the scheduling approach document.