Data migration: the finance director wants the variance hidden
You are project manager for a data and records migration in a government department: seven legacy systems consolidated onto one cloud platform. The approved budget (BAC) is £2.40m over 12 months. At the end of month 10 the finance officer reports planned value of £2.00m, earned value of £1.68m and actual cost of £1.85m. The variance is concentrated in two of the seven data domains, where source records were far dirtier than the sample used at estimating, and the team has been reworking migrated records twice. A fixed-price supplier is delivering data cleansing for £400k, paid on acceptance of each domain; three domains remain unaccepted. The sponsor holds a £150k management reserve outside the project budget for unidentified risks. The senior responsible owner has asked for a one-page forecast, and the finance director has suggested that the project draw on the management reserve to keep the reported variance neutral until after the annual report.
1. Calculate CPI, SPI, EAC, VAC and TCPI, and explain what these measures do and do not tell the senior responsible owner.
Read the worked answer
The numbers
| Measure | Calculation | Result |
|---|---|---|
| CPI | 1.68 / 1.85 | 0.91 |
| SPI | 1.68 / 2.00 | 0.84 |
| EAC | 1.85 + (2.40 - 1.68) / 0.91 | £2.64m |
| VAC | 2.40 - 2.64 | -£0.24m |
| TCPI | 0.72 / (2.40 - 1.85) | 1.31 |
What they tell the SRO. The project is delivering about 91 pence of value per pound spent and has earned 84% of the value planned by month 10. On current performance the outturn is £2.64m, an overrun of £0.24m, and the remaining work would have to be delivered 31% more efficiently than the original plan to hold the budget.
What they do not tell the SRO
- SPI here is a cost-based proxy for schedule, not a date. Report actual milestone slippage, and use earned schedule if a time forecast is needed.
- EAC assumes the remaining work will perform like the past. That is unsafe here, because the variance is concentrated in two domains with a known cause - dirty source data and double rework. Do a bottom-up re-estimate of the three unaccepted domains and the remaining migration work.
- Actual cost excludes commitments. The £400k supplier is paid on acceptance and three of seven domains are outstanding, so cash and accruals must be forecast separately.
- Nothing here shows whether the benefits - records availability and decommissioning of legacy systems - are still achievable.
Report the forecast outturn, the cause, the confidence range, and the three options: descope, fund, or re-baseline.
2. How should you respond to the finance director's suggestion about the management reserve?
Read the worked answer
Recommend refusing, politely and in writing.
Why it is wrong on the mechanics. Management reserve is held for unidentified risks - work not in the baseline because nobody knew about it. The £0.24m forecast overrun is not an unknown risk; it is a known performance variance caused by data quality and rework. Using reserve to absorb it converts a risk reserve into a slush fund and leaves the project unprotected when a genuine unknown arrives.
Why it is wrong on principle. The purpose of the suggestion is to keep the reported variance neutral until after the annual report. That presents a position you know to be inaccurate to the board, to the accounting officer and potentially to external auditors. It is a misstatement, and the project manager would be a party to it. It also displaces the governance principle that the SRO and board decide on the basis of the true forecast.
What to do
- Give the finance director the one-page forecast with the £2.64m EAC, the cause and the options.
- Explain what the reserve is for, and offer to document the variance transparently with the recovery plan, so the report shows control rather than concealment.
- Route the decision through the SRO and change board: descope (defer two domains), additional funding, or re-baseline with the forecast declared.
- If the instruction is repeated, escalate to the SRO and the finance business partner or audit, and record the request and your advice.
Tone matters. Assume the director is under reporting pressure rather than acting dishonestly, and give them a credible alternative that still lands well in the annual report.
3. Recommend what you would do about the remaining work to protect the outturn and the benefits.
Read the worked answer
Stop the rework before re-planning the work. Two migrations reworked twice means the process is producing defects; more effort applied to the same process will simply cost more.
Sequence
- Root cause. Profile the source data in the remaining domains before migration. Agree cleansing rules with the business data owners and freeze them. Add a quality gate before each domain migrates rather than inspecting afterwards.
- Re-estimate bottom-up. Apply the actual defect and rework rates from the two problem domains to the remaining three, and reforecast cost and duration. Replace the CPI-extrapolated EAC with this figure and explain the difference.
- Consider descoping. Deferring two low-value domains to a phase two may protect the go-live date and the remaining budget. Assess which benefits fall away and get the SRO to take that decision explicitly through change control.
- Test the supplier position. The £400k fixed price was based on a sample the client provided. If the data is materially dirtier than the sample, check whether this is a change or a compensation event under the contract, and whether acceptance criteria should include a pre-cleansing quality gate so rework is not billed twice.
- Tighten control. Weekly cost-to-complete, a change freeze on requirements, and agreed tolerances for the next reporting period so the board hears early rather than late.
- Protect the benefits. Reconfirm the business case still holds on the reduced scope before anything is deferred.