Windows analysis and collapsed as-built are two of the most commonly used forensic delay methods, and choosing between them isn't a matter of preference — it's a matter of what the project's records can actually support. Picking the wrong method for the available documentation is one of the fastest ways to undermine an otherwise legitimate delay claim.
This guide walks through how each method works, what it requires, and how HADVEN decides which one fits a given project record. For the related question of TIA versus after-the-fact analysis generally, see TIA vs. as-built: choosing a delay method.
How windows analysis works
Windows analysis divides the project's actual timeline into discrete periods, or "windows," usually bounded by the dates of contemporaneous schedule updates. Within each window, the analyst compares the schedule's critical path and projected completion date at the start of the window to the same at the end, and attributes any change in the forecast completion date to the specific delay events that occurred during that window.
The method's core strength is that it uses the schedule as the project team actually saw it in real time — each window's starting point is a schedule that was accepted or at least submitted contemporaneously, not reconstructed after the fact. That makes findings easier to defend, because the analysis is anchored to documents that existed before the dispute arose.
What windows analysis requires: a reasonably complete series of contemporaneous CPM schedule updates (monthly updates are typical), each one logically sound enough to support a valid critical path calculation, and clear documentation of what delay events occurred within each window.
How collapsed as-built works
Collapsed as-built starts from the opposite end: it begins with the as-built schedule — a reconstruction of what actually happened, activity by activity, built from daily reports, RFIs, submittal logs, and other field records — and then removes ("collapses out") specific delay events one at a time to measure how much each event affected the final completion date.
This method is often used when contemporaneous schedule updates are missing, inconsistent, or too unreliable to support a windows analysis — for example, on smaller projects where formal CPM updates weren't maintained, or where the update file history is incomplete. Because it works backward from the as-built record rather than forward through contemporaneous updates, it depends heavily on the quality and completeness of the underlying field documentation.
What collapsed as-built requires: a detailed, well-documented as-built schedule (which itself may need to be reconstructed from daily reports, RFIs, submittals, and correspondence if one wasn't maintained), and a clear, well-supported basis for identifying which activities were actually delayed and by how much.
Side-by-side comparison
- Data source: windows analysis uses contemporaneous schedule updates; collapsed as-built uses the reconstructed as-built record.
- Best fit when: windows analysis fits projects with a complete, logically sound series of accepted CPM updates; collapsed as-built fits projects where updates are missing, unreliable, or never existed.
- Defensibility basis: windows analysis is anchored to documents that existed before the dispute; collapsed as-built is anchored to the quality of the as-built reconstruction, which is built after the fact and can draw more scrutiny on how activities and durations were determined.
- Typical effort: collapsed as-built often requires more upfront work to reconstruct a reliable as-built schedule when one wasn't maintained during the project.
How HADVEN chooses a method
The choice is a records question, not a preference question. HADVEN's process on every forensic delay engagement starts with a review of what's actually available:
- Inventory the schedule update history. How many contemporaneous updates exist, how complete and logically sound are they, and do they cover the full disputed period?
- Assess as-built documentation quality. Are daily reports, RFIs, submittal logs, and correspondence complete enough to support a reliable as-built reconstruction if needed?
- Match the method to what the records support — not the other way around. A method chosen because it's theoretically preferred, but that the records can't actually support, produces a weaker analysis than a well-executed method that fits the available documentation.
- Stay open to a hybrid approach. Some engagements use windows analysis where contemporaneous updates exist and fill gaps with as-built reconstruction where they don't, rather than forcing a single method across a record with uneven quality.
This is the same records-first approach HADVEN applies to every forensic delay analysis engagement, and it's why the first deliverable on most engagements is a records assessment, not a finished analysis.