Case 4 - Under budget or under delivering?
Marcus controls a 12-month core-banking data migration with a budget at completion of £800,000. At the end of month 5 the figures are: planned value £350,000, earned value £280,000, actual cost £340,000. Three of the eight data domains have failed reconciliation twice, and the team that should be cleansing reference data has been pulled onto a parallel regulatory report for six weeks. The finance director told the board last week that the project is tracking under budget because it has spent £10,000 less than planned. The sponsor proposes two remedies: add two data analysts at £9,000 per month each, and compress the testing window from 10 weeks to 6 weeks to recover schedule. The vendor's fixed-price contract covers extraction and loading only; cleansing and reconciliation are internal. A sister bank's migration failed after go-live for exactly this reason and its remediation cost £2.4m. Change control has been bypassed for two small scope additions.
1. Interpret the numbers and correct the finance director's statement. What does earned value not tell Marcus here?
Read the worked answer
The figures
- SPI = EV / PV = 280,000 / 350,000 = 0.80. The project has delivered 80% of what was planned by month 5.
- CPI = EV / AC = 280,000 / 340,000 = 0.82. For every £1 spent, about £0.82 of work has been earned.
- CV = EV - AC = -£60,000 (over budget on work performed); SV = EV - PV = -£70,000 (behind schedule).
- Forecast at completion using typical variance: EAC = BAC / CPI = 800,000 / 0.82 ≈ £971,000, a variance at completion of about -£171,000.
- Only 35% of the work is complete against 41.7% of the schedule elapsed.
Correcting the statement: spending less than plan is not a saving when output is lower still. Comparing actual cost with planned value conflates money spent with progress earned. The accurate statement is: behind schedule, over budget on work performed, with a credible forecast overspend near £170,000.
What EVM cannot show: technical quality. The two failed reconciliations in three domains are invisible in SPI and CPI. The project can be perfectly on cost and still be undeliverable because the data is wrong.
2. Evaluate the sponsor's two remedies. Which do you accept, reshape or reject?
Read the worked answer
Adding two analysts: reshape, do not simply approve.
£18,000 per month buys capacity, but only helps if the binding constraint is analyst hours. The stated constraint is that the cleansing team was removed to another project, and reconciliation is failing twice - a quality and ownership problem, not purely a throughput problem. Approve only with two conditions: the team is actually dedicated and backfilled on the regulatory report, and the analysts are assigned to reconciliation defect fixes with a measurable exit criterion.
Compressing testing from 10 weeks to 6: reject.
Testing is the only control that detects the failure mode already occurring. Cutting 40% of the test window to recover visible schedule converts a measurable risk into an unmeasured one, and this organisation has direct evidence of the cost: £2.4m of remediation at the sister bank. It also conflicts with the three domains that have failed reconciliation twice.
Better alternative: phase the go-live by domain, migrate the five reconciled domains first, and hold the three failing ones behind a data-quality gate with a named owner and date. That recovers perceived schedule without removing the control.
Test: does the remedy address the actual constraint or merely the visible symptom?
3. What control and reporting changes should Marcus make now?
Read the worked answer
Reinstate control
- Bring the two bypassed scope additions into change control retrospectively, with impact assessed and approved or reversed; do not leave them as precedent.
- Re-baseline formally with sponsor approval, since the original plan is no longer achievable. A re-baseline is legitimate when authorised and documented - quietly resetting figures is not.
Report differently
- Report CPI and SPI with EAC as a range, not a single number, and state the basis of the estimate.
- Add leading quality indicators alongside EV: reconciliation pass rate per domain, open severity 1 and 2 defects, and data-cleansing coverage. Cost and schedule alone would have shown this project as merely late.
Escalate the real constraint
The six-week removal of the cleansing team is a portfolio prioritisation decision, not something Marcus can resolve with his own resources. Take it to the portfolio owner with the £2.4m comparator.
Set gates
No domain goes live without two consecutive clean reconciliations and named business sign-off. Define the go/no-go criteria now, while there is still time to influence them.