Core idea: Validation asks whether a model is built correctly and whether it represents the decision well enough for its intended use.
When this model is useful
Use it before presenting a material model, especially when calculations are complex or the model was built by the person advocating one option.
The method is most useful when the question, alternatives and time horizon are stated before calculations begin. It should clarify tradeoffs and identify which assumptions deserve attention, not merely produce a score.
Inputs and evidence
Model logic, formulas, units, assumptions, historical outcomes, test cases, edge values and independent reviewers.
Use consistent units, definitions and periods across options. Mark estimates clearly, record their source and use ranges when precision is not supported. Evidence should be proportionate to the cost, risk and reversibility of the choice.
Step-by-step method
Perform face-validity review, trace formulas and units, test zero and extreme inputs, back-test where possible and ask whether users can explain what drives the result.
- Write the decision question and accountable owner.
- List realistic alternatives and eliminate any that fail hard constraints.
- Collect evidence and separate verified facts from assumptions.
- Run a base case and at least one downside test.
- Record the chosen option, accepted tradeoffs and review triggers.
Practical example
A model mixing monthly subscription cost with annual support cost may produce plausible totals while being wrong by a factor of twelve.
The point of the example is not the exact numbers. It is the discipline of using the same boundaries for every option and making the decision drivers visible.
Common mistake and limitation
A technically correct model can still answer the wrong question. Validate both the arithmetic and the decision frame.
Review whether the result changes under reasonable alternative assumptions. A close or fragile ranking should be presented as such rather than converted into false certainty.
Questions to ask before deciding
- What evidence would change the preferred option?
- Which consequence is missing because it is difficult to measure?
- Could a threshold or constraint override the numerical result?
- Who receives the benefits and who bears the costs or risk?
- When should the decision be reviewed?
Professional context: Legal, tax, investment, safety, medical, engineering and regulated decisions require qualified, jurisdiction-specific advice.