Agile & Project Management
Cost Management and Earned Value: A Worked EVM Example
Advanced7 min readUpdated
Earned value management answers the question a spend report cannot: are you getting what you paid for? This guide builds a cost baseline, then works one project through every earned value formula with real numbers, interprets the indices honestly, and covers the technique's limitations and how it behaves on iterative deliveries.
What you will learn
- Build a cost baseline from estimates, contingency and management reserve
- Compute planned value, earned value and actual cost at a reporting date
- Derive schedule and cost variances and their indices and explain what each means
- Forecast completion with three EAC formulas and choose the right one
- State the limitations of earned value and how it applies to iterative delivery
A spend report tells you how much money has left the account. It cannot tell you whether the work bought with it is worth what it cost, and that gap is where cost overruns hide for months. A project five months in, having spent forty per cent of its budget, might be exactly on plan, comfortably ahead, or heading for a fifty per cent overrun — and the finance system reports the same number in all three cases.
Earned value management closes the gap by measuring progress in the same unit as spend. It is the most quantitative technique in project management, it has genuine limitations, and it is worth learning properly because the arithmetic is simple and the interpretation is not. This guide builds a cost baseline first, works one project through every formula, and then discusses where the method misleads.
Estimating and budgeting
Cost estimating uses the same three families as duration estimating, applied to money. Analogous estimating scales from a comparable past project; parametric estimating multiplies a measured rate by a quantity, such as a daily rate by estimated person-days; bottom-up estimating prices every work package in the breakdown structure and sums upward, which is the most accurate and the most expensive to produce. Most real budgets combine them: bottom-up for the well-understood parts, parametric for the rest.
Include everything that will actually be charged. Direct costs — people, licences, hardware, contractors — are obvious. Indirect costs such as shared infrastructure, office overhead and management time are frequently omitted and then appear as an overrun. Recurring platform costs after go-live belong to operations rather than to the project, but they still need an owner; the cloud cost management guide covers that side.
Aggregating estimates gives the cost baseline, in layers:
| Layer | Contents | Who authorises spend |
|---|---|---|
| Work package estimates | Priced deliverables from the breakdown structure | Project manager |
| + Contingency reserve | Identified risks, sized from the risk analysis | Project manager |
| = Cost baseline | The time-phased budget performance is measured against | — |
| + Management reserve | Unknown unknowns, typically 5–10 % | Sponsor |
| = Total project budget | The full authorised funding | — |
Time-phasing the baseline across the schedule produces the familiar S-curve: cumulative planned spend, flat at the start while the team ramps up, steep through the build, flattening as work completes. The shape matters because the flat start is why a project can look under budget for months while already being behind.
The three measurements
Everything in earned value derives from three numbers taken at a reporting date.
Planned value (PV) — the budgeted cost of the work that should have been completed by now, read off the baseline S-curve.
Earned value (EV) — the budgeted cost of the work that has actually been completed. Note the word budgeted: earned value is measured at baseline rates, never at what the work really cost. This is the number that requires discipline, because it depends on an honest rule for crediting partial completion. The safest rule is 0/100 — credit nothing until a work package is finished — which is conservative and difficult to game. The 50/50 rule credits half on start and half on completion. Percentage-complete claimed by the person doing the work is the least reliable and the most common.
Actual cost (AC) — what has genuinely been spent on that work, from the finance system, including committed costs on invoices not yet paid.
Also fixed: budget at completion (BAC), the total cost baseline.
A worked example
A portal replacement has a budget at completion of 480,000 over twelve months. At the end of month five, the reporting pack gives:
BAC = 480,000 Budget at completion
PV = 200,000 Planned value at end of month 5
EV = 170,000 Earned value — budgeted cost of completed work
AC = 190,000 Actual cost incurred to date
Variances and indices
Schedule variance SV = EV - PV = 170,000 - 200,000 = -30,000
Cost variance CV = EV - AC = 170,000 - 190,000 = -20,000
Schedule performance index SPI = EV / PV = 170,000 / 200,000 = 0.85
Cost performance index CPI = EV / AC = 170,000 / 190,000 = 0.89
Both variances are negative and both indices are below 1.0, so the project is behind schedule and over cost. Concretely: only 85 % of the planned work has been done, and every currency unit spent has produced 89 cents of budgeted work.
The percentage view makes the same point in a form that stakeholders read faster — 35.4 % of the value has been earned while 39.6 % of the budget has been spent against a plan of 41.7 %. If you are converting figures like these for a report, the percentage calculator handles the arithmetic quickly.
Forecasting
Forecasts extend today's evidence to completion. Three formulas answer three different assumptions.
Remaining work at budget: BAC - EV = 480,000 - 170,000 = 310,000
1. Cost performance continues (systemic cause)
EAC = BAC / CPI = 480,000 / 0.8947 = 536,500
2. The overrun was a one-off; rest goes to plan
EAC = AC + (BAC - EV) = 190,000 + 310,000 = 500,000
3. Both cost and schedule pressure continue
EAC = AC + (BAC - EV) / (CPI x SPI)
= 190,000 + 310,000 / (0.8947 x 0.85)
= 190,000 + 407,600 = 597,600
From the chosen forecast, two further figures follow:
Estimate to complete ETC = EAC - AC = 536,500 - 190,000 = 346,500
Variance at completion VAC = BAC - EAC = 480,000 - 536,500 = -56,500
So on the middle assumption the project will overrun by about 56,500, roughly 12 % of budget, and needs a further 346,500 to finish.
Can it still be recovered?
The to-complete performance index answers that. It is the cost efficiency the remaining work must achieve to hit a given target.
To finish within the original budget:
TCPI = (BAC - EV) / (BAC - AC) = 310,000 / 290,000 = 1.07
To finish within the revised forecast:
TCPI = (BAC - EV) / (EAC - AC) = 310,000 / 346,500 = 0.89
The first number is the important one. The team has been running at 0.89 efficiency and would now need to run at 1.07 — better than plan, for the entire remainder — to land on the original budget. That is a twenty per cent improvement with no stated reason to expect it, so the original budget should be treated as lost and the conversation moved to which scope is being cut or which reserve is being released. Research on large projects consistently finds that cost performance rarely improves after a project is roughly twenty per cent complete, which is precisely why a TCPI far above the current CPI is a red flag rather than a target.
Reading the indices
| SPI | CPI | Situation | Typical cause |
|---|---|---|---|
| < 1 | < 1 | Behind and over budget | Underestimated scope or complexity |
| < 1 | > 1 | Behind but cheap | Under-resourced; work has not started, not that it is efficient |
| > 1 | < 1 | Ahead but expensive | Overtime, contractors, or scope credited too generously |
| > 1 | > 1 | Ahead and under budget | Genuinely good, or the baseline was padded |
The second row is the one most often misread as good news. A project consuming less money than planned because nobody has been assigned to it is not saving money; it is accumulating a delay that will be paid for later.
Limitations
Earned value is only as good as the data underneath it, and five weaknesses are worth stating plainly.
It measures conformance to plan, not value. A project delivering exactly the wrong product on budget scores 1.0 on both indices. Nothing in the method asks whether the scope is still worth building.
The schedule index degrades near the end. Because all remaining work is eventually earned, SPI converges to 1.0 regardless of lateness. Past roughly eighty per cent complete it carries little information, and the critical path is the reliable signal.
It is sensitive to the earning rule. Optimistic percentage-complete claims inflate earned value directly, and the resulting indices look reassuring while the project deteriorates. This is the single most common way earned value reporting becomes worthless.
It has real overhead. A time-phased baseline, work-package-level cost collection and monthly recalculation cost effort. On a small project the reporting can cost more than it saves, and simple milestone tracking is a better trade.
Scope changes invalidate the baseline. Every approved change must be reflected in the budget at completion, or every subsequent index compares against a plan that no longer exists.
Earned value on iterative deliveries
The technique adapts if you decide what counts as earned. The common approach treats completed backlog items — or completed story points — as the earned quantity, with a budgeted cost per unit derived from total budget divided by total estimated scope.
Total scope estimated at 1,200 story points, BAC = 480,000
Budgeted cost per point = 400
End of iteration 10 (of 24):
Points completed 425
EV = 425 x 400 = 170,000
PV (planned by now) = 200,000
AC (10 iterations of a stable team) = 190,000
The arithmetic is identical from there. The precondition is a reasonably stable total scope: where the backlog grows by a third during delivery, the budget at completion keeps moving and the indices measure the estimate rather than the team. For genuinely evolving products, throughput-based forecasting as described in estimation and planning is more informative, and burn-up charts that show scope growth explicitly are more honest than any index.
Reserves deserve one final note. Contingency should be sized from the risk analysis, not set as a flat ten per cent, and drawing on it should be a recorded event tied to the risk that occurred. A contingency budget consumed by ordinary optimism was never contingency; it was a padded estimate.
Key takeaways
- The cost baseline is estimates plus contingency; management reserve sits outside it, and releasing it changes every earned value figure.
- Earned value is the budgeted cost of completed work, so it is only as honest as the rule used to credit partial completion — 0/100 is the hardest to game.
- SPI and CPI express schedule and cost efficiency as ratios; below 1.0 means less work delivered per unit of plan or per unit of spend.
- Three EAC formulas encode three different assumptions about whether the overrun is systemic, one-off, or compounding with schedule pressure.
- TCPI tests recoverability: if it sits well above the CPI achieved so far, the original budget is gone and the honest move is to cut scope or release reserve.
- The schedule index becomes unreliable near completion and the method says nothing about whether the scope is still worth building — read it alongside the critical path and the product's value case.
Sources and further reading
Frequently asked questions
- Why is the schedule performance index expressed in money?
- Because earned value measures schedule progress by the budgeted cost of the work completed, not by dates. That is its best-known weakness: as a project approaches its end, all remaining work eventually gets earned, so the index drifts toward 1.0 even on a late project. Always read it alongside the critical path, and never use it as the only schedule signal in the final quarter.
- Which estimate at completion formula should I use?
- It depends on why the overrun happened. If current cost performance reflects something systemic — a wrong rate, an underestimated codebase — divide the budget by the cost performance index. If the overrun was a one-off such as a hardware failure, add the remaining budgeted work to actual costs. If both cost and schedule pressure are expected to continue, use the formula that divides remaining work by the product of both indices, which is the most pessimistic.
- What is the difference between contingency and management reserve?
- Contingency reserve covers identified risks, sits inside the cost baseline and is spent by the project manager when a known risk occurs. Management reserve covers unknown unknowns, sits outside the baseline in the total project budget, and is released by the sponsor. Because management reserve is outside the baseline, drawing on it changes the budget at completion and therefore every earned value figure computed against it.
- Does earned value work on agile projects?
- It can, if you define what counts as earned. The usual approach is to treat completed story points or finished backlog items as the earned quantity and the budget per point as the rate, so earned value becomes points completed times budgeted cost per point. It works best where scope is reasonably fixed and the team is stable; where the backlog changes substantially, the budget at completion keeps moving and the indices lose meaning.