02 Independent Assurance / Worked example
One fact, many values
Ask what a project’s budget is and the file will give you several answers. Each document looks authoritative. Most of the differences are not mistakes: a figure includes VAT, or was true last quarter, or is a forecast called a budget. But until someone establishes which is which, nobody can tell the real problem from the noise.
Much of what an assurance review finds is really an information failure. This example shows it, and what fixes it: a governing record, with one value per fact and a definition, source and owner for each.
Illustrative. The project, documents, values and dates are invented to show the method. The kinds of difference shown are common in real project files.
Fact / Document
Full business caseMar 2025
Funding agreementApr 2025
Gate review reportMay 2025
Contract dataJun 2025
Board paperJun 2025
Press releaseJul 2025
Contractor's programmeJul 2025
Risk registerAug 2025
Monthly cost reportAug 2025
Contractor's forecastAug 2025
Annual reportSep 2025
Approved budget
Contract sum
Client contingency
Completion date
Estimate confidence
AgreesDifferent definitionDifferent dateDifferent quantity, same nameRoundedWrong
| Fact | Governing value | Definition | Source | Owner | Variants reconciled |
|---|---|---|---|---|---|
| Approved budget | £48.6m | Approved total cost at outturn, including contingency, excluding VAT | Full business case, approved Mar 2025 | Senior responsible owner | 5 |
| Contract sum | £39.8m | Tendered total of the Prices at contract award | Contract data, signed Jun 2025 | Commercial lead | 3 |
| Client contingency | £4.4m | Client-held contingency at award, set at the P80 of quantified risk | Full business case, Mar 2025 | Senior responsible owner | 2 |
| Completion date | 30 Sep 2027 | Contractual Completion Date | Contract data, Jun 2025 | Commercial lead | 7 |
| Estimate confidence | P80 | Confidence level of the approved budget, from quantitative risk analysis | Full business case, Mar 2025 | Senior responsible owner | 4 |
What it shows
Three findings
- Most differences are not errors. In this file, two values out of thirty are wrong. The rest are different definitions, different dates or different quantities that share a name. Treating them all as errors wastes a review; ignoring them hides the two that matter.
- The dangerous ones sit in plain sight. A contractor’s forecast seven weeks beyond the contract date, and a board told the estimate is P90 when it is P80. Both are visible in the file. Neither is visible to anyone reading one document at a time.
- The fix is structural, not heroic. A governing record with a definition, source and owner for each key fact lets every document be checked against it. Differences then become labelled variants or flagged errors, instead of a reviewer’s judgement call.
Next step
We carry out independent reviews for boards, sponsors, funders and regulators, and help clients build the governing record that makes their own information assurable. Discuss an independent review →