A plausible number is not enough.

Specify missing-answer behavior explicitly; a legitimate zero must not be confused with an unanswered input.

Keep each cost component visible enough to explain what changes when the visitor changes scope.

A tier needs an explicit boundary rule: test just below it, exactly at it and just above it.

Decide where rounding occurs and keep displayed figures consistent with the saved result.

Reuse a stable expression within a funnel when it improves consistency; do not assume cross-funnel synchronization.

A lookup table can simplify maintenance, but its publication and synchronization rules are part of the requirement.

Write expected outcomes independently, then compare normal, missing, boundary and revised-answer cases.

Check where calculation occurs and when the visitor last changes an input; arithmetic alone does not establish freshness.

A revised answer should lead to a revised, identifiable result—not a mixture of old and new assumptions.