17.2 Data Accuracy, Assumption Control & Go/No-Go Decisions
Key Takeaways
- Study data are validated for origin, completeness, representativeness, accuracy, precision, timeliness, units, and fitness for the decision—not merely presence in a database.
- An assumption register identifies owner, basis, date, range, dependency, verification plan, and effect if wrong; high-value assumptions receive early testing.
- Mass, water, energy, area, schedule, and cash-flow reconciliations expose inconsistent workstream inputs before they enter the investment case.
- A decision gate uses pre-defined criteria for stop, recycle, hold, or advance and records dissent, conditions, residual risk, and accountable approvals.
- Optimism bias is controlled through reference-class evidence, independent review, scenario analysis, and transparent treatment of exclusions and contingency.
Feasibility failures often begin at interfaces: geology uses one density, mining another recovery, processing another feed tonnage, and finance an outdated schedule. Each model can balance internally while the project model does not. Decision quality therefore depends on controlled data, assumptions, and reconciliation.
Data Fitness Tests
For each critical dataset, ask:
- Origin: who measured or calculated it, by what method?
- Completeness: what intervals, locations, periods, or fields are missing?
- Representativeness: does it cover ore types, weathering, depth, season, and operating range?
- Accuracy and bias: what reference, calibration, QA/QC, or reconciliation tests it?
- Precision: how repeatable is it and what variability is natural?
- Timeliness: does it reflect current design, price, law, and site condition?
- Consistency: are units, coordinates, dry/wet basis, currency, and definitions aligned?
- Fitness: is the quality sufficient for the consequence of this decision?
A preliminary datum can be acceptable for screening but unacceptable for equipment purchase or reserve conversion.
Assumption Register
An assumption is a placeholder treated as true for analysis. Record:
| Field | Example |
|---|---|
| Assumption | Grid power available by commissioning |
| Basis/date | Utility letter dated 15 June |
| Owner | Infrastructure manager |
| Range | 0–9 month delay |
| Dependencies | Construction power, plant start, diesel backup |
| Validation | Connection agreement and design |
| Consequence | Capital, operating cost, emissions, schedule |
| Due gate | Before investment approval |
Do not hide an assumption in a spreadsheet cell. High-uncertainty, high-consequence assumptions should be tested first.
Reconciliation Checks
Mass
Ore mined plus opening stockpile must equal plant feed plus closing stockpile and accounted losses. Plant feed metal must reconcile to product metal, tailings metal, and inventory.
Water
Inflows minus outflows equal storage change for dry, average, wet, startup, upset, and closure cases.
Energy
Equipment load lists, operating hours, process models, and utility demand should agree, including starts and standby.
Schedule
Access, permits, engineering, procurement, construction, commissioning, development, and ore availability must follow logic. A date imposed without predecessor activities is not a schedule.
Economics
Revenue tonnes and grade must match the mine and plant schedule. Capital timing must match construction. Closure cost and working capital must enter cash flow at the correct time.
Risk, Opportunity and Bias
Uncertainty can be a symmetric range; risk is an uncertain event with consequence; bias is systematic optimism or pessimism. Contingency may address estimated scope uncertainty but cannot make an impossible schedule credible.
Use:
- sensitivity analysis for value drivers;
- scenarios for linked conditions;
- probabilistic analysis where inputs and correlations are defensible;
- quantitative schedule-risk analysis;
- reference-class comparison with similar completed projects; and
- independent technical and commercial review.
Beware double counting. If a cost distribution already includes quantified risk, adding the same risk again as flat contingency inflates cost. Conversely, excluding schedule delay from both risk and contingency understates exposure.
Decision Gate
Define criteria before reviewing the result:
- advance: evidence and value support the next commitment, with accepted residual risk;
- conditional advance: specified conditions must be completed before or during the next phase;
- recycle: redesign or collect defined information and return to the gate;
- hold: wait for an external event such as permit or market clarification; or
- stop: fatal flaw or insufficient value.
Record decision, evidence, dissent, conditions, owners, dates, and what would trigger re-review. Senior approval does not turn missing data into accurate data.
Example
A base case has positive NPV, but 70% of revenue comes from an inferred geological zone excluded from the Mineral Reserve and scheduled in the first two years. This is not a small sensitivity. The mine and cash-flow premise are inconsistent with the reserve basis. Rebuild the schedule using supportable material, obtain more drilling, or stop; do not approve the base case because its spreadsheet NPV is positive.
The best exam answer identifies the data defect, traces its effect through schedule and value, and chooses a gate response proportionate to consequence.
Assumption Expiry
Set an expiry or revalidation trigger for volatile assumptions such as price, exchange rate, tariff, fuel, law, land access, and vendor delivery. A figure that was accurate at study start may be stale at approval. Re-baseline material changes and show their effect rather than silently overwriting the record.
Challenge the Base Case Before Approval
A gate team should reproduce the material balances and value drivers independently, then compare downside combinations rather than only one-at-a-time sensitivities. Test delay with inflation, lower recovery with higher processing cost, or wet-season productivity with water constraints where dependencies are credible. Conditions of approval need an owner, evidence, deadline and consequence; otherwise a conditional advance becomes an undocumented unconditional commitment.
A feasibility model shows positive NPV, but most early revenue comes from material not included in the stated Mineral Reserve. What is the appropriate response?