8.2 RP 29R-03 Taxonomy, Data Validation, and Method Selection

8.2 RP 29R-03 Taxonomy, Data Validation, and Method Selection

AACE RP 29R-03 organizes forensic schedule analysis methods so practitioners can describe and compare them consistently. Its purpose is to reduce avoidable procedural subjectivity, not to declare one method acceptable for every dispute. The document expressly recognizes that the contract, facts, law, documentation, and analysis objective affect method choice.

Core dimensions

Observational versus modeled

An observational analysis studies what the available schedules and records show without creating a counterfactual by inserting or removing impacts. A modeled analysis changes a schedule model—such as adding a fragnet or subtracting identified delay—to measure a hypothetical effect.

Static versus dynamic logic

A static-logic method relies primarily on one network representation. A dynamic-logic method evaluates logic and critical-path changes through multiple contemporaneous updates or periods. Dynamic analysis can better reflect an evolving project, but only if the updates are sufficiently reliable.

Contemporaneous versus retrospective perspective

A contemporaneous analysis asks what the parties could reasonably see at the time. A retrospective analysis uses the completed record. Neither perspective is inherently superior; each answers a different question and carries different hindsight risk.

Source data validation

Electronic schedule files are evidence, not self-proving facts. Validation commonly compares:

• data dates and update sequences;
• actual starts and finishes against daily reports, inspections, and correspondence;
• remaining durations against contemporaneous forecasts;
• logic, constraints, calendars, and coding changes;
• claimed progress against quantities, payment records, or turnover evidence; and
• schedule narratives against the native file.

Not every difference makes a schedule unusable. The analyst should identify material discrepancies, assess their effect, document repairs or assumptions, and preserve the original data. Quietly changing dates or logic to make a preferred method work undermines reproducibility.

Selecting a method

Start with the question. A prospective change evaluation may suit a time impact analysis using the current accepted update. A completed project with reliable monthly updates may support a windowed observational analysis. A project with poor or missing schedules may require an as-built reconstruction, with correspondingly stronger disclosure of inference and limitation.

Use a selection matrix:

Consistency and bias control

Window boundaries, criticality criteria, calendar treatment, and responsibility coding should be selected for a reason that can be applied consistently. A window may follow data dates, major phase changes, or critical-path shifts, but moving a boundary only to isolate a favorable event is result-driven.

A useful report states:

1. the question and scope;
2. sources received and validation performed;
3. method and why it was selected;
4. assumptions and repairs;
5. criticality and concurrency criteria;
6. calculations and results; and
7. limitations and alternative interpretations.

The goal is a reproducible technical analysis. RP 29R-03 supports that discipline while leaving room for justified departures based on professional judgment.

Applied review: choose a method after defining the question

Method selection begins with the assignment: what period, milestone, issue, and causal question must be evaluated? Then assess the available records, schedule quality, update frequency, baseline acceptance, model changes, and resources for analysis. A method that relies on reliable periodic updates may be unsuitable when only a late as-built record exists; a modeled approach may introduce excessive assumptions when contemporaneous updates already show changing critical paths.

RP 29R-03 provides a taxonomy and implementation considerations, not a mandatory ranking of methods. Document why the selected approach fits the data and question, what modifications were necessary, and how sensitivity checks affect the result. Preserve native files and an audit trail, validate dates against source records, and distinguish observed history from analyst-created modeling.

Communicate limitations plainly. If records cannot support day-level precision, do not manufacture it. Different reasonable methods may answer different questions or produce different numerical outcomes. The professional task is to make the reasoning reproducible and proportionate to the evidence, not to select the technique that produces a preferred claim value. Independent peer review can improve reliability and clarity.