Run Planning Uncertainty Analysis

This guide adds uncertainty to a reviewed planning model and interprets the result for one decision.

Before You Start

• Use a model with reviewed formulas and actuals.
• Choose one output and period.
• Identify the small number of assumptions whose ranges matter.
• Keep the deterministic base and scenario available for comparison.
• Save a named version before changing established assumptions.

1. State the Question

Write the question in range form.

Examples:

• What range of ending cash does the model produce for December?
• What range of revenue follows from uncertain conversion and pricing?
• What range of payroll follows from uncertain hiring dates?

Choose one primary output. Do not begin by simulating every variable.

2. Select Uncertain Inputs

Choose assumptions that are both uncertain and material.

Good candidates include:

• conversion rate
• customer arrivals
• collection timing
• price realization
• hiring volume
• usage or demand

Do not add a probability function to an imported actual.

3. Choose a Distribution

Use the simplest distribution that reflects the assumption.

• Use uniform(from, to) for an evenly plausible bounded range.
• Use triangle(from, to) when the middle is more plausible.
• Use normal(mean, variance) for symmetric variation around a mean.
• Use normal_from_interval(from, to, confidence) when you can defend an interval more easily than a variance.
• Use sample(value1, value2, ...) for discrete cases.
• Use poisson(lambda) or binomial(n, p) for appropriate event counts.

Confirm the function parameters in formula suggestions.

4. Check the Central Value

Save the probability formula and review the ordinary grid.

The grid shows a stable central value when no simulation is running. Confirm that value is reasonable and that the downstream output still calculates.

If the fixed model is wrong, correct it before running uncertainty.

5. Run the Simulation

1. Open Formula uncertainty.
2. Select Run simulation.
3. Wait for the 200 model runs to complete.
4. Select the output variable.
5. Select the decision period.
6. Record P10, P50, and P90.

6. Interpret the Range

Compare:

• the deterministic base value
• the selected named scenario
• P10
• P50
• P90

Ask:

• Is the range large enough to change the decision?
• Which assumption creates most of the spread?
• Does the range cross a cash, covenant, margin, or hiring threshold?
• Does the modeled downside remain operationally acceptable?

The percentile range is conditional on the model and distributions. It is not an external forecast guarantee.

7. Challenge the Assumptions

For each probability formula, record:

• the owner
• the source or rationale
• why the distribution shape is appropriate
• why the bounds or parameters are defensible
• the period where the assumption applies

Use row comments when the rationale needs collaborative review.

8. Compare With a Named Scenario

A named scenario is easier to discuss than a percentile alone.

Create or keep a downside scenario that represents a coherent operating case. Compare it with the simulated range.

• If the downside sits near P10, explain why.
• If it lies outside the simulated range, review the distributions.
• If P10 is worse than the named downside, check compounding uncertain drivers.

9. Save the Review Context

1. Save a view with the uncertain inputs and selected output.
2. Include the relevant periods.
3. Add comments that explain the distribution choices.
4. Save a named model version.
5. Export a point-in-time file when the review needs one.

Record the active scenario and simulation date with the result.

Troubleshoot

No uncertainty variables appear

Confirm a supported probability function feeds the model and run the simulation again.

The simulation returns an error

Check invalid parameters, upstream formula errors, and the model calculation status.

The result barely changes

Confirm the uncertain input feeds the selected output and increase the range only when business evidence supports it.

The result is implausibly wide

Check variance units, long-tailed functions, duplicated uncertainty, and compounding drivers.

P10 and P90 appear reversed for the business decision

The percentile order is numeric. Interpret whether higher or lower is favorable for the selected variable.

Next Steps

• Planning Uncertainty
• Compare and Share Planning Scenarios
• Collaborate on a Planning Model