Accurate rows, wrong total
The assumption to test is simple: if every budget row is correct, their sum must be correct. Imagine materials of $4,000 and labor of $6,000, plus an installation package of $10,000 that already includes both. Every figure can be accurate, yet adding all three gives $20,000 for $10,000 of work. Correct arithmetic has counted the same work twice. These examples are illustrative, not reported test results.
Ask what each amount covers
Treat each amount as a claim about a defined set of costs. Record its deliverable, period, estimate version, and whether it is a component or a rollup. Include either the package or its covered components. An addendum may revise an existing allowance: although its name suggests addition, it can replace an amount already recorded. Read the cost coverage before deciding to increase the total. Similar labels alone cannot establish overlap. The same description can cover different sites; different descriptions can refer to one invoice. Use the underlying scope to decide.
A hypothetical boundary case
Hypothetically, a revised package costs $12,000, while a separate survey line shows $2,000. The earlier $10,000 package is superseded. Decision rule: use $12,000 if the revision includes the survey, or $14,000 if it explicitly excludes it. If coverage is unresolved, show a provisional $12,000–$14,000 range and request a scope breakdown before approving a single figure. This range assumes the survey’s inclusion is the only uncertainty.
Run a coverage check
Choose one deliverable and mark every row that claims to pay for it. Trace those rows to the quote or estimate, then explain each overlap. Repeat for the remaining deliverables before comparing the budget with available funding. Keep excluded rows visible for audit purposes, with a reason, while leaving them outside the sum. Where components are available, rebuild the package subtotal independently and investigate any unexplained difference. A total earns confidence when each cost appears once within the chosen scope.