Sensitivity boundary

A sensitivity boundary separates a mechanical what-if result from evidence about what could occur. A sensitivity changes one or more model inputs and reports the resulting accounting or analytical output. It does not show that the change is reasonably possible, independent, probable, or caused by the input.

State the base case, changed input, range source, units, held-constant assumptions, tax treatment, interactions, nonlinear effects, and output. For example, changing a discount rate by 0.5 percentage points shows the model response under that rate. It does not establish that the rate change is supported for the reporting date.

Use inputs with clear provenance and reconcile the sensitivity to the approved base model. When inputs interact, test a combined case or explain why separate results cannot be added. Report direction and amount only at the precision the model supports.

Keep unsupported scenarios outside the accounting conclusion. A wide result can show decision exposure and prompt more evidence; it cannot select an estimate or override recognition guidance. The executive explanation may summarize a supported sensitivity, but it must carry the limitation. The release gate evaluates the underlying estimate and evidence separately from the mechanical what-if calculation.