Methods
Delay analysis methods, compared honestly
No method is inherently best. A method is fit when it answers the relevant question with the reliable data available, respects the contract and the governing law, and survives alternative explanations. In practice the records usually make the choice.
| Method | Perspective | Best use | Frequent failure |
|---|---|---|---|
| Time impact analysis | Prospective | Contemporaneous assessment of an extension of time, and any contract that asks for the effect of an event to be assessed at the time rather than argued at the end. | Hindsight smuggled in through the fragnet or the updated logic |
| Impacted as-planned | Prospective | Early, low-value or simple assessments, and as an illustrative cross-check rather than a primary retrospective analysis. | Tests the events against a plan rather than against what was happening on site |
| Windows and time-slice analysis | Retrospective | Retrospective analysis of a dynamic project where the controlling work changed over time and reasonable programme updates exist. | Windows chosen arbitrarily, or chosen to place events conveniently |
| As-planned versus as-built | Retrospective | Simpler works, or projects where programme updates are missing but the as-built record is good. | Planned logic quietly treated as actual logic |
| Collapsed as-built, or but-for analysis | Retrospective | Retrospective counterfactual testing where a validated as-built programme with defensible logic exists. | Logic invented for an as-built record that never had any |
| As-built critical path and longest path | Retrospective | Fact-led retrospective enquiry, particularly where the tribunal has signalled that it wants the facts rather than a model. | The longest chain mistaken for causation |
| Contemporaneous period analysis | Either | Reconstructing how assessments were, or should have been, made as the job ran, particularly under contracts that require assessment at the time. | Unvalidated updates carried forward without checking |
| Daily delay measure | Retrospective | High-resolution analysis of projects with genuinely reliable daily records, typically process, rail or infrastructure work with disciplined data capture. | Granularity creating false precision on a project whose records cannot support it |
- ProspectiveTime impact analysisUpdate the programme to the moment before the event, insert the event, and measure the forecast effect.
- ProspectiveImpacted as-plannedInsert the delay events into the baseline programme and read off the revised completion date.
- RetrospectiveWindows and time-slice analysisCut the project into consecutive periods and identify what controlled completion in each one.
- RetrospectiveAs-planned versus as-builtCompare what was planned with what happened, within periods, reasoning from the progress record.
- RetrospectiveCollapsed as-built, or but-for analysisTake the as-built programme, remove the delay events, and see when the works would have finished without them.
- RetrospectiveAs-built critical path and longest pathBuild the as-built programme, trace the longest continuous path to completion, and identify what drove it.
- EitherContemporaneous period analysisReconstruct the decisions period by period using the programme updates and narratives that existed at the time.
- RetrospectiveDaily delay measureCompare planned against actual at daily resolution across a maintained as-built model.
Three guidance families, side by side
Method taxonomies come from professional bodies rather than from law. The Society of Construction Law Delay and Disruption Protocol sets out practical principles for delay, disruption and records. AACE International’s recommended practice on forensic schedule analysis is the fullest taxonomy and expressly resists a one-size-fits-all prescription. ASCE published a consensus technical standard covering critical path, float, chronology, concurrency and responsibility.
All three are technical guidance. None is legislation, none is a contract term unless the parties adopted it, and none decides a case. Where this site describes a method as robust or weak, that is a statement about evidential quality, not about legal approval.
The validation protocol every model should pass
- Identify the forensic question before choosing the method.
- Inventory native files, calendars, codes, baselines and updates. Hash and preserve the originals.
- Check status dates, actual dates, data-date discipline, open ends, constraints, lags, out-of-sequence settings and calendar changes.
- Reconcile programme progress against diaries, quantities, photographs, procurement and payment evidence.
- Explain every added, deleted or revised activity and relationship.
- Identify critical and near-critical paths in each relevant period, and test criticality against what the site actually did.
- Build an event register tied to notice, cause, affected activity and evidence.
- Model claimant and non-claimant causes. Do not infer responsibility from the colour of a bar.
- Run sensitivity tests for contested logic, calendars, progress, pacing and concurrency.
- Make the calculation reproducible from the source files, and keep an audit log.
Built by AI Metric
The analysis on this site is only as fast as the evidence behind it. AI Metric builds bespoke systems for consultancies, contractors and claims teams: document and correspondence triage, event registers assembled from the project record, programme and cost reconciliation, and drafting support that always cites the document it came from. Built for review by your own experts, never to replace their judgement.
General explanation of how contract mechanisms, analysis methods and legal principles generally work. It is not legal or contractual advice, not an opinion on any project, and no standard-form contract wording is reproduced anywhere on this site. Standard forms are routinely amended, so every default described here, including every time period, can be different on your project. Your executed contract, as amended, and the governing law and forum always control. Deadlines may already be running: if an event has occurred, preserve your position and take qualified advice.