Short answer: a schedule delay analysis is easier for others to test and rely on when its method fits the records that exist, its inputs can be traced to those records, and its choices are stated and explained. No method is right for every project. AACE International Recommended Practice No. 29R-03, Forensic Schedule Analysis is a widely used technical reference that sorts methods into families. It describes itself as a reference rather than a prescriptive standard, so the analyst still has to explain why a method was chosen.
How RP 29R-03 organizes the methods
RP 29R-03 groups retrospective methods along a few dividing lines:
- Observational or modeled. Observational methods look at schedules as they are, without inserting or removing delays. Modeled methods change a schedule model, by adding or subtracting delay events, and compare the results.
- Static or dynamic logic. Static methods rely on a single set of schedule logic, such as the baseline. Dynamic methods follow how the logic changed from update to update.
- Additive or subtractive. Additive models insert delays into a schedule that does not contain them. Subtractive models start from the as-built condition and remove delays.
The labels below are common names. RP 29R-03 identifies nine method implementation protocols within these families.
As-planned versus as-built
What it does: compares the planned schedule with what actually happened, either for the whole project or period by period.
When it fits: when there is a reasonable baseline and a reliable as-built record, but schedule updates are missing, sparse or unreliable. It is observational and generally uses static logic, so it relies heavily on the quality of the baseline and the as-built dates.
Records it depends on: the baseline schedule, actual start and finish dates, and the field records used to confirm them, such as daily reports, inspection records and photographs.
Windows analysis
What it does: divides the project into periods ("windows") and measures how much the forecast completion moved in each one, and what drove that movement.
When it fits: when there is a sequence of schedules covering the project. "Windows" is a broad label: it can describe an observational review of updates or a modeled analysis performed period by period, so the label alone does not say how the work was done.
Records it depends on: a chain of schedule versions, plus the records explaining events in each period. How periods are chosen, whether monthly, by event or grouped, should be stated, because period boundaries can affect results.
Contemporaneous period analysis
What it does: examines the project's own schedule updates one after another, measuring the loss or gain of time along the controlling path between data dates, without inserting or removing delays. A variant uses an intermediate, or "half-step," schedule to separate the effect of progress from the effect of logic or duration revisions made in the update.
When it fits: when regular updates were prepared during the project and can be shown to be reasonably reliable. Because it uses the forecasts made at the time, it reflects what the project team saw and reported as the work progressed, when the updates are shown to be reliable.
Records it depends on: native schedule update files, submission correspondence, schedule narratives and comments, and records confirming reported progress.
Time impact analysis and fragnets
What it does: models a specific event as a fragment of schedule network, a "fragnet," inserts it into a schedule and recalculates to see how the completion date changes. It is a modeled, additive approach.
When it fits: prospectively, for evaluating a time request during the project, and in some retrospective assignments. Results depend on which schedule the fragnet is inserted into, its status date, and how the fragnet's activities, durations and logic ties were derived.
Records it depends on: the schedule current at the time of the event, the documents describing the event (RFIs, directives, change records), and the basis for each fragnet duration and logic tie.
Collapsed as-built
What it does: starts with a model of the as-built schedule and removes selected delays to estimate when the work would have finished without them, sometimes described as a "but-for" model. It is modeled and subtractive, and can be done on a single as-built model or on multiple models covering different periods.
When it fits: when a detailed as-built record exists, even if contemporaneous updates are limited. Because the as-built model's logic is often reconstructed, how that logic was built, and how each delay was defined and removed, are central to understanding the result.
Records it depends on: detailed as-built dates, the records used to reconstruct logic, and documentation defining each delay that is removed.
Why method fit matters more than method name
Two analyses using the same label can produce different results because of different inputs, period choices, logic adjustments or assumptions. The more an analysis changes the project's own schedules, the more important it is that each change is documented and traced to a record. Where a disputed fact affects the result, running the analysis under alternative assumptions can show how much the result depends on that fact.
Questions counsel can ask about any delay analysis
- Which method family was used, and why does it fit the records available?
- Which schedule versions were relied on, and how was their reliability tested?
- Were native files used, and were any schedules adjusted or reconstructed? If so, how, and on what records?
- How were analysis periods chosen?
- For modeled methods, where did each inserted or removed duration and logic tie come from?
- How was the critical path identified in each period, and how did it change?
- Which assumptions does the result depend on, and how sensitive is it to them?
- What contrary records were considered, and how were they addressed?
- What questions does the analysis not answer?
Related reading
For background on concurrent delay, including that how concurrent delay is treated depends on the contract and applicable law, which are for counsel and the tribunal, see What is concurrent delay? How AACE RP 29R-03 defines it. For selecting an expert, see What construction attorneys should ask a delay expert before retaining one. Oxley's construction delay expert witness work is led by Daniel R. Zorrilla, J.D., BSBC, CGC, CFCC.
This article is general information, not legal advice. An opinion in any matter depends on that matter's records and assignment. How a contract or the law applies depends on the facts; consult counsel.