Green Dashboard, Red Numbers
Northwind Health's claims platform migration has a nine-month baseline and a $1,480,000 budget. At the end of month 5 the PMO extracts the following: planned value $820,000, earned value $705,000, actual cost $845,000. The monthly dashboard still shows an amber-to-green trend because the milestone list contains no overdue items. Delivery leads have logged 1,900 hours of overtime since month 2, and two senior engineers were reassigned to another programme in week 18 without a formal resource change. Fourteen severity-2 defects remain open, and the regression suite has drifted from 480 to 391 automated tests as features were added. Eleven change requests were approved; four are pending, and the change board has not met for six weeks. The sponsor wants to skip the full regression cycle to protect a statutory reporting date, arguing that the numbers look fine and that the last release had no production incidents.
1. Using the figures given, calculate cost performance index, schedule performance index, estimate at completion (CPI-based) and to-complete performance index. What do these tell you, and where does the dashboard mislead the sponsor?
Read the worked answer
Calculations
- CPI = EV/AC = 705,000/845,000 = 0.83
- SPI = EV/PV = 705,000/820,000 = 0.86
- EAC (CPI-based) = BAC/CPI = 1,480,000/0.83 ≈ $1,783,000; VAC ≈ −$303,000
- TCPI = (BAC−EV)/(BAC−AC) = 775,000/635,000 ≈ 1.22
Interpretation. Each dollar spent is returning about 83 cents of earned value, and the project has earned less than planned. Recovering the baseline requires future work to perform at 1.22 efficiency - roughly 22% better than plan and 46% better than the current run rate. That is not a realistic recovery without a scope, schedule or cost decision, so the honest forecast is overrun plus lateness, not green.
Where the dashboard misleads.
- Milestone-based RAG only displays items that are overdue; work slipping inside a milestone window stays invisible.
- The milestone list was never updated for the resource reassignment, so capacity changes did not propagate into the plan.
- The 1,900 overtime hours are suppressing visible schedule variance - SPI looks survivable only because people are absorbing it.
- Defect counts and the shrinking regression suite are leading indicators of future cost that a cost-only dashboard cannot show.
Present the position to the sponsor in the units they care about: the statutory date, the ~$303k forecast overrun, and the residual risk being carried into production.
2. The sponsor directs you to skip the full regression cycle. What is your response, and what must happen before any decision is taken?
Read the worked answer
Refusing flatly is not the job; making the trade-off explicit and owned is. Sequence:
- Establish whether the test is discretionary. A statutory reporting date usually brings mandated controls (financial reporting, data integrity, safety). If the regression suite evidences a regulated control, skipping it is not the sponsor's call - it is a compliance decision requiring compliance or audit sign-off in writing.
- Quantify, do not assert. State the residual defect load (14 open severity-2 defects plus 89 tests removed from the suite), the historical escape rate, and the relative cost of a defect found in production versus in test. Offer three options: full regression with the date moving; a risk-based subset covering regulated and highest-traffic paths with a documented rationale; or deferral of named low-risk modules with a post-go-live plan.
- Route it through change control, not email. A quality-baseline change of this size needs a change request, an impact assessment on cost, schedule and risk, and a decision by the accountable owner - not the PM alone.
- If overridden, record the decision, the data presented, the residual risk, the accepting authority and the mitigation plan in the risk register and decision log.
The real question is not whether go-live can happen; it is who owns the consequence and whether that is documented.
3. What governance failures does the evidence suggest, and which one would you correct first?
Read the worked answer
The evidence points to at least five failures; fix the one that lets the others persist.
- Change control cadence. The board has not met for six weeks while 11 changes were approved, meaning approvals happened elsewhere and impact was never assessed. Fix: reinstate a weekly board with delegated limits and a mandatory impact assessment per change.
- Resource change control. Two engineers were removed without a formal change, invalidating the capacity baseline.
- Baseline and configuration control. The test suite shrank from 480 to 391 without a recorded decision.
- Reporting design. RAG is built on overdue milestones rather than forecast, which structurally hides slippage.
- EVM hygiene. PV was not replanned after approved changes, so variances are understated.
First correction: change control, because it is the mechanism that would have caught both the resource removal and the test-suite reduction. Concretely: re-baseline with sponsor approval, publish a decision log, and require every change to carry cost, schedule, quality and risk impact before approval. Everything else depends on a valid baseline - without one, the other fixes are guesswork and the next dashboard will be just as misleading.