Contingency Reserve Formula¶
Reserve sizing rule — instantiates Reference-Class Planning Calibration
Converts a chosen percentile of the calibrated overrun distribution into a protected, evidence-linked reserve that cannot be shaved without moving the number.
A calibrated forecast can tell you a project's most likely cost sits at one figure while its comparable-case history has a long, ugly right tail — and still leave the budget defenceless, because a point estimate carries no reserve. Contingency Reserve Formula is the small, explicit rule that closes that gap: it names a target percentile on the calibrated distribution, reads the value there, and sets the reserve as the distance from the plan's base figure to that percentile. Its defining move is that the reserve is derived, not negotiated — the buffer is a function of an evidence-based distribution and a stated risk appetite, so anyone who wants a smaller reserve has to argue with the distribution or the percentile, not haggle the number down in a meeting. It sizes one dial (the buffer) from an uncertainty range it is handed; it does not build that range, and it does not itemize the individual risks that produced it.
Example¶
A film production company is greenlighting a mid-budget feature and, as a condition of financing, must post a completion guarantee — a reserve large enough that the completion guarantor is confident the film can be finished even if things run long. The line producer's schedule says principal photography is a most-likely 44 days. But the calibrated distribution, drawn from comparable shoots of similar cast size and location complexity, shows overruns bunched from a few days up to a two-week tail when weather and a lead actor's availability go wrong. The Contingency Reserve Formula does one thing: the guarantor sets the target at the 80th percentile of that distribution, reads off "52 days," and fixes the reserve at the eight-day gap — converted to dollars at the daily burn rate, plus a matching cost-side percentile for post-production. The reserve is now a stated rule: reserve = P80 − plan. When the producer later asks to trim the buffer to make the top-line look leaner, there is nowhere to push; the only honest moves are to lower the confidence target (and say so on the record) or to bring evidence that this shoot's distribution is genuinely tighter than its class.
How it works¶
- Take the distribution as input. The formula does not generate the overrun distribution — it consumes one produced upstream (a base-rate distribution, a Monte-Carlo roll-up, or a three-point spread).
- Fix the risk appetite as a percentile. Choose the confidence level the reserve must cover — P50 for a routine internal task, P80 or P90 for an exposed capital or safety-critical commitment. This is a governance choice, made once and recorded.
- Read and subtract. Reserve = value at the chosen percentile minus the plan's base figure. Do it separately for schedule and for cost, since their distributions differ.
- Bind the reserve to its evidence. Record which distribution and which percentile produced it, so any later reduction must change one of those inputs rather than quietly erasing the buffer.
The whole apparatus is deliberately mechanical. Its value is not sophistication but the way it removes discretion from the moment a reserve is most tempting to cut.
Tuning parameters¶
- Target percentile — the confidence the reserve buys. Higher percentiles cover deeper tails but tie up capital and invite the charge of padding; lower ones free money but leave the project exposed to ordinary bad luck.
- Schedule/cost split — whether one percentile governs both dimensions or each gets its own. Separate targets are more honest (a schedule can slip without a cost blowout) but double the bookkeeping.
- Reserve base point — whether the reserve is measured from the P50, the mean, or the inside-view plan. Measuring from an optimistic plan silently inflates the reserve; measuring from the P50 keeps it clean.
- Release schedule — how the reserve is drawn down as milestones pass without incident. A front-loaded release frees cash early but can strand a project that hits its worst risks late.
When it helps, and when it misleads¶
Its strength is depoliticizing the buffer. Because the reserve follows from a percentile on real comparable outcomes, it resists the classic failure the archetype warns about — contingency negotiated downward before anyone has looked at the overrun distribution. This is the logic behind the optimism-bias uplift that some public-investment guidance mandates: a percentage add-on calibrated from a reference class of past projects rather than argued case by case.[n1]
It misleads when the distribution feeding it is wrong or stale — a precise percentile computed from a flattering, cherry-picked class produces a confident but under-sized reserve, and the formula's tidiness lends that number false authority. The classic misuse is running it backward: picking the reserve the sponsor will tolerate and then choosing the percentile that yields it, which converts an evidence rule into decoration. The guarding discipline is to fix the percentile before reading the value, keep it visible, and re-derive the reserve whenever the underlying distribution is updated rather than defending the first figure.
How it implements the components¶
contingency_buffer_policy— this is its core output: a reserve set as a stated function of a percentile and an evidence-based distribution, deliberately hard to renegotiate.uncertainty_interval_and_tail_frame— it consumes the calibrated range and picks a specific tail point on it, making the reserve a chosen depth into the downside rather than a flat percentage.
It does not build the overrun distribution from comparable cases — that base_rate_distribution is the Historical Project Outcome Database's job — and it does not itemize the discrete risks or re-score them over time (rolling_recalibration_trigger, scope_change_ledger); that running catalog is the Schedule and Cost Risk Register. This formula reads one number off a distribution; the register produces it.
Related¶
- Instantiates: Reference-Class Planning Calibration — supplies the buffer that converts calibrated uncertainty into protected reserve.
- Consumes: Schedule and Cost Risk Register or Three-Point Estimate with Base Rates — either can supply the distribution the percentile is read from.
- Sibling mechanisms: Schedule and Cost Risk Register · Historical Project Outcome Database · Reference-Class Forecasting Workbook · Three-Point Estimate with Base Rates · Launch or Commitment Readiness Gate · Independent Estimate Round · Forecast Backtesting Review · Premortem as Auxiliary Probe
Editorial Notes¶
Form Classification¶
Form family: Rule, Policy & Commitment
Rationale: Converts a chosen percentile of the calibrated overrun distribution into a protected, evidence-linked reserve that cannot be shaved without moving the number, making its operative form a standing rule, threshold, contractual commitment, or policy constraint governing future conduct.
Independent corroboration: The frozen evidence defines Contingency Reserve Formula as 'Converts a chosen percentile of the calibrated overrun distribution into a protected, evidence-linked reserve that cannot be shaved without moving the number', so its operative form is Rule, Policy & Commitment.
Review outcome: Independent reviewer agreement; high confidence.
Origin Attribution¶
Primary origin: Operations Research
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Specialized
Rationale: Quantitative risk analysis cohered percentile-based reserve sizing from calibrated cost or schedule overrun distributions.
Related originating lineages:
- Economics & Finance — Financial risk management contributes explicit risk appetite and protected reserve amounts.
- Organizational & Management Science — Project controls supplied protected reserve governance.
- Public Administration & Policy — Public-investment appraisal institutionalized optimism-bias uplifts derived from reference classes.
- Statistics & Experimental Design — Probability modeling supplied the calibrated uncertainty distribution.
Review resolution: Quantitative risk analysis in operations research is primary, while finance, project controls, public cost governance, and statistical uncertainty modeling are genuine contributing lineages; the percentile-to-protected-reserve rule is specialized and synthesized.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Reconciled after independent review; high confidence.
Notes¶
[n1] An optimism-bias uplift is a percentage add-on to a project's estimate, calibrated from the observed overrun distribution of a reference class of comparable past projects — the mechanism by which some public-sector appraisal guidance turns reference-class evidence into a mandated, non-negotiable reserve rather than a case-by-case argument. ↩