Project Forecast-to-Complete Override Rate

A project forecasting model predicts completion at a fixed cost based on past spending, but a manager changes the estimate because an unpriced supplier claim and future design work are known. Project forecast-to-complete override rate measures how often a forecast's calculated remaining-cost result is replaced by a documented human estimate.

Define forecast, since estimate to complete is remaining expected cost while estimate at completion combines incurred cost with expected remaining cost under a stated method. PMI discusses revised estimates at completion using performance, commitments and future conditions, and APQC publishes project budget-variance measures.

These sources support examining forecast assumptions, not a universal acceptable override rate. Choose the denominator by counting forecast submissions, project-periods or distinct line estimates with an eligible model output, and define override so that a normal update to actual cost is not treated as a discretionary replacement of the forecast result.

Record the original model-generated estimate and its input version before alteration, and record the revised figure with override amount, direction, reason, reviewer and date. Check evidence, since a supplier quote, approved scope change, productivity trend and risk event can justify changes.

Avoid unsupported optimism, because a lower manual estimate without a credible improvement plan can conceal a likely overrun, and avoid pessimism too, as adding all worst-case risks can overstate expected completion cost if they are not likely to materialize together. Check actuals, since incomplete cost posting can make a model appear low or high and source data should be corrected before overriding, and check commitments, because purchase orders and subcontract obligations can affect remaining cost before invoices arrive.

Check change control, as a proposed customer change is not necessarily funded or authorized and should be shown separately. Segment frequency: a routine monthly override may be part of the approved method, while unexplained repeated changes signal weak modelling.

Track size, since one large override can matter more than many small ones and value and direction should be reported, and show open review, because an override proposed but not approved should not become the official forecast. Distinguish reserve, as approved contingency and management reserve can have different treatment in the forecast, and avoid double counting a cost already included in model commitments by also adding it manually as a risk allowance.

Compare accuracy later, because a low override rate does not prove the model was right and realised project cost should be assessed separately. Review rationale, since a generic management judgment label is too thin to audit, keep period snapshots so last month's forecast is not overwritten when new information appears, and protect comparability, as projects at different stages may need different forecast methods and adjustment ranges.

Track reversals, so a supplier claim later withdrawn removes its forecast effect with a trace, check timing because an override after reporting cutoff may be a next-period decision, not a rewrite of an issued report, and audit approvals since finance and delivery may need different sign-off on material changes. Report uncertainty as a range where it communicates real unknowns better than an unjustified precise number, review reason distribution (overrides for known supplier commitments point to data-feed gaps while repeated unexplained reductions may point to governance pressure), align dates so a commitment signed after the forecast cutoff is not used to claim the prior model was wrong, review both directions on the evidence, and use override rate to locate where judgment adds information and where the model needs correction.