Source XER
Georgian_College_current.xer
For each documented delay event, the as-built network is cloned, the event's impact removed (by either shorten, reducing affected activity durations, or remove, deleting the affected activities and their TASKPRED edges), and the CPM forward pass re-run. The resulting project finish is the but-for [event] date. The differential between the as-built finish and the but-for finish is the event's attributed impact under the Modeled / Subtractive method.
Configuration. Default removal method is shorten (a per-event override applies when an event specifies its own removal_method). Contractor filter: No, every event in the input list contributes to the cumulative pass.
The subtractive method is sensitive to the input impact_days values. A ±20% input uncertainty on the per-event impact durations can swing the cumulative result by a similar amount. AACE 29R-03 §3.8 recommends the analyst run a schedule risk analysis (Monte Carlo with duration uncertainty distributions) against the input durations and report the P50 ± P90 range alongside the deterministic figures shown here. This dashboard reports a deterministic point estimate; opposing counsel should be expected to attack the input uncertainty on cross-examination. See the schedule-risk-analysis skill output for the corresponding probabilistic envelope.
When two delays affect the project simultaneously on overlapping paths, the subtractive method (this skill) and windows analysis (forensic-delay-analysis MIP 3.3) can produce different attribution. The cumulative pass treats all delays as additive; if the project has concurrent owner-vs-contractor delays or two parallel critical chains, run a concurrent-delay analysis (forensic-delay-analysis windows method, AACE 29R-03 §3.6) and reconcile against the output below. Dual-method gaps >10% are a strong signal that concurrent delays are present.
For each delay event below, the as-built network was cloned, the event's impact removed using the indicated method, and the CPM forward pass re-run. The resulting "but-for" finish and computed impact are shown.
| Event ID | Name | Affected Activities | Removal | As-Built Finish | But-For Finish | Impact (wd) | Responsible | Contractor? |
|---|---|---|---|---|---|---|---|---|
| E1 | AHJ / commissioning-agent hold on MEP and elevator commissioning | A4180 | shorten | 2026-08-31 | 2026-08-24 | 5 wd | Owner | N |
| E2 | Owner-directed re-sequence of integrated systems testing | A4190 | shorten | 2026-08-31 | 2026-08-24 | 5 wd | Owner | N |
| E3 | Contractor punch-list and deficiency re-work | A4240 | shorten | 2026-08-31 | 2026-08-17 | 10 wd | Contractor | Y |
For each event: the duration removed and the basis used, plus what drives the but-for finish. A completed activity whose finish is on or before the data date cannot pull a successor whose start is pinned at the data-date floor, so the impact can legitimately be 0 even when a duration was removed. That sensitivity is disclosed below.
Both methods converge within the 5% practitioner-consensus tolerance. The forensic conclusion is dual-method validated.
Thresholds: AGREE ≤ 5% · WARN 5-15% · INVESTIGATE > 15%. Bands are forensic-engineering practitioner consensus per Bramble & Callahan §9.05 and Wickwire et al. Ch. 10. They are NOT codified numeric cut-offs in AACE RP 29R-03 (which contemplates the cross-check itself in §3.8 but does not publish numeric tolerance percentages).