Earned value management: CPI, SPI and estimate at completion

Earned value management puts three amounts beside each other: the budgeted work you planned to complete, the budgeted work you can prove is complete, and what that work has cost. CPI = EV / AC tests cost efficiency. SPI = EV / PV tests progress against the approved baseline. For a program finance lead, the first decision is whether the progress evidence is good enough to use either index.

You can make that assessment from the work-package budget, a dated progress export and ledger actuals. A warehouse project is not a prerequisite. The delivery owner still owns the evidence that a milestone has been accepted; a financial tool cannot infer completion from an invoice.

Start with scope, progress and one reporting date

Use an approved baseline with a budget for each work package and a schedule of when its work should be complete. Preserve its version, approval date and change history. Budget at completion (BAC) is the total approved budget for the scope being measured. Keep any reserve outside that scope separate rather than earning value against it.

For each package, agree a progress rule before reporting: accepted units, weighted milestones or a stated completion rule. The weights must add to the package budget. A manager's unsupported percentage is not progress evidence. Record actual costs through the same cutoff, including the accruals finance requires for work already performed.

PV and EV are measured in dollars of budgeted work, not sales, invoices or a valuation of the project's benefits. AC is dollars of cost. Keep all three on a consistent cost basis.

CPI and SPI formulas, with the variances beside them

Cost variance (CV) = EV - AC

Schedule variance (SV) = EV - PV

Cost performance index (CPI) = EV / AC

Schedule performance index (SPI) = EV / PV

A negative CV means the work completed has cost more than its budgeted value. A negative SV means less budgeted work has been completed than was planned by that date. SV is a dollar amount, not a number of days.

CPI and SPI above 1 indicate favorable performance on their respective comparisons; below 1 indicates unfavorable performance. Report the numerator and denominator beside each index. If AC or PV is zero, its index is undefined, not zero. Investigate missing costs or an unstarted baseline before interpreting it.

A worked $1 million program

This synthetic US systems-replacement program has a $1,000,000 approved baseline. At the September 30 reporting date, design is accepted. Build uses weighted, accepted milestones: 70% was planned and 60% is earned. Migration uses accepted units: half was planned and one-third is complete. These percentages describe the agreed work rules, not the share of cash spent.

The program has earned about $0.86 of budgeted work for each dollar incurred. It has completed about 86% of the budgeted work planned by the cutoff. It is both over cost for the work achieved and behind its baseline progress. The equal indices are a feature of these inputs, not a rule that cost and schedule performance move together.

Compute the program indices from summed dollars. Do not take a simple average of work-package CPI or SPI values: a small package would then have the same influence as a large one. Use unrounded values for the forecast calculations below.

Estimate at completion: state the assumption before the formula

The forecast identities are:

Estimate at completion (EAC) = AC + estimate to complete (ETC)

ETC = EAC - AC

Variance at completion (VAC) = BAC - EAC

ETC must include the full remaining cost of delivery. An open purchase order may support that estimate; do not add the same committed amount again if it is already in the remaining forecast. The project cost control guide covers budget, commitments, actuals and the remaining forecast together.

For the example, compare three explicit assumptions. They are alternative forecasts, not amounts to add together.

The CPI forecast uses $1,000,000 / (600,000 / 700,000), not the displayed four-decimal index. It is a check on the forecast, not proof that the eventual cost will be that amount.

Some organizations also use EAC = AC + (BAC - EV) / (CPI × SPI) when they expect both cost and schedule efficiency to affect remaining cost. That requires a defensible relationship between delay and cost. An SPI below 1 alone does not establish one. Choose the assumption with the delivery owner and finance, and retain the reason rather than selecting whichever result fits the budget.

Where it goes wrong

Progress follows spend. Paying a supplier does not earn the milestone. Require delivery evidence and reconcile financial cutoffs separately. If accepted progress is missing, flag EV as unverified and hold the indices out of the performance conclusion.

The baseline moves without a bridge. A revised schedule can make SPI improve without any additional work. Show authorized scope changes and compare versions. Do not silently overwrite the baseline used in last month's report.

Schedule performance becomes a finish date. SPI does not describe dependencies or the critical path. At completion EV and PV both reach BAC, so SPI returns to 1 even if delivery finished late. Keep the schedule forecast alongside it.

Missing costs create a favorable CPI. A late invoice or unrecorded accrual can make the program appear efficient. Reconcile AC to finance and list incomplete cost periods before reporting.

A forecast formula becomes a promise. CPI continuation assumes a repeatable cost pattern. New scope, vendor changes and one-time rework can break it. Keep the bottom-up remaining forecast visible and explain the difference.

Budgets and actuals use different scope. Unallocated costs, reserve movements or inconsistent overhead treatment can distort the comparison. Resolve the scope difference before interpreting an index.

Bring the budget and progress evidence to the same review

For a medium or large US program, the useful output is one table by work package: baseline version, progress rule, evidence date, PV, EV, AC and the approved remaining forecast. In Covirage, the agreed definitions and supplied files provide the inputs; Covirage's tools compute every figure and check the rollups. The external AI model explains the results. It does not establish milestone completion or invent the missing forecast, and your team does not need a data team to prepare a warehouse first.

See project and program cost control for the file-based review. For the adjacent question of why actual costs differ from plan, read variance analysis; earned value adds the evidence of work completed.