Steel-Procurement Lag and a Constraint That Hides the Critical Path
A hospital expansion has a 60-week baseline containing 2,100 activities. The contractor is preparing a delay analysis after a 45-day steel-delivery slip. The schedule highlights steel erection as critical, but the logic shows a 90-day lag on a finish-to-start relationship from 'Design complete' to 'Fabricate steel'; the lag is not a separate activity. Steel erection also carries a start-no-earlier-than date constraint tied to a client milestone, and the scheduler calculates 22 days of total float on it. The project manager argues that the critical path is the longest bar chain and that the 22 days prove the delay is non-critical. The client's reviewer notes that the steel procurement activity has no predecessor other than the lag and that the constraint date is before the current data date. The scheduler must determine the true critical path and decide what to report.
1. Identify the scheduling errors that make the reported critical path unreliable. Explain how each affects total float and the forward/backward pass.
Read the worked answer
The reported critical path is not defensible for three reasons. First, a 90-day FS lag is being used as a substitute for work. A lag has no resources and is not a true activity; it hides the fabrication and delivery process and can sit on the driving path without showing its own risk. The correct model is an activity such as 'Steel fabrication and delivery' with a 90-day duration and its own predecessor and successor logic. Second, the start-no-earlier-than date constraint overrides logic. Constraints are date overrides, not logic; they can force the forward pass to start an activity later than its predecessors require and can make total float appear larger or smaller than the logic would produce. A constraint dated before the data date is especially suspect because it may already be violated or may mask negative float. Third, 'longest bar' is not the critical path. The critical path is the longest path through the logic network, determined by forward and backward pass and confirmed by zero or minimum total float. To test, remove the constraint, replace the lag with an activity, re-run CPM, and check that the path is contiguous from project start to completion. The 22 days of float should be recalculated; if the steel path is truly driving, the float will collapse or go negative. Report the model errors and the corrected path, not the PM's conclusion.
2. Recommend how to repair the model and calculate a defensible critical path. Include what to do with the lag and constraint, and how to communicate the effect on the 45-day delay.
Read the worked answer
Repair the model before any delay analysis. Replace the 90-day lag with a real procurement activity and assign the same calendar and, where relevant, resources. Remove the start-no-earlier constraint or convert the client milestone into a logical predecessor or successor with a milestone activity; if the client date is contractual, keep it as a separate milestone, not as a constraint on the work. Add any missing predecessor to steel procurement, such as design release or purchase order. Run the CPM, check the longest path, and reconcile the data date. Then perform a time impact analysis: take the updated schedule at the data date, insert a fragnet for the delivery slip, and compare forecast completion with and without the slip. If the steel path becomes critical, the 45-day slip may drive completion; if it does not, quantify the remaining float. The key is to separate model repair from delay entitlement: first make the network valid, then measure impact. Do not accept the original 22-day float as proof until the constraint and lag are removed.
3. If the client demands a single completion date forecast, what schedule metric should the scheduler use, and what caveats must accompany it?
Read the worked answer
For a single completion forecast, use the critical path remaining duration from the data date, supported by earned schedule where EVM data are available. The forecast is data date plus remaining longest-path duration, adjusted for known risks and resource constraints; do not use total float as a schedule buffer or subtract it from the completion date. If using earned schedule, report SPI(t) and the time-based forecast, and explain that SPI is a cost-weighted index that can distort time forecasting. Caveats: the network must be constraint-free and lag-free, open ends closed, out-of-sequence progress corrected, and the forecast should be a range with confidence, not a single deterministic date. State that the 22-day float is model-dependent and not a contractual buffer.