People do not need to read every formula to trust a model. They do need to know where its inputs came from, which assumptions they may change, and how the headline result is calculated.
Start with a clear separation between actual data, editable assumptions, calculations, and final outputs. Color can help, but labels and instructions matter more. An input should never look like a calculated result. A historical figure should never quietly change when someone selects a future scenario.
Add a few simple checks: do totals reconcile with the underlying detail, are dates aligned, and are essential inputs missing? Show a warning when the model cannot support a conclusion. A blank input should not turn into a confident-looking zero without explanation.
Document the important choices in plain language. If a cost is allocated among locations, say how and why. If an output is an estimate rather than a posted accounting figure, label it. If a formula changes, record the reason.
A trustworthy model makes it possible for another person to question it constructively. That is a strength. A useful challenge may reveal an assumption that should be updated or a process that should be clarified.