Earned value management

Earned value management (EVM) integrates scope, schedule and cost to measure project performance objectively. Instead of asking only "how much have we spent?" or "are we on schedule?", it asks "how much of the planned work have we actually done, and what did it cost?" EVM gives early warning of trouble, supports forecasts of final cost and duration, and is a reliable source of calculation questions on the PMP exam. This lesson defines the three core measures, the variances and indices, the forecasting formulas and when to use each, the to-complete performance index, ways of measuring progress, earned schedule, and how to interpret results. CON-310 applies EVM to construction cost control in more depth; here the focus is on the PMP formulas and their interpretation.

The three core measures

Planned value (PV) is the authorized budget assigned to the work scheduled to be done by a given date. It is read from the time-phased cost baseline (the S-curve from lesson 10). Earned value (EV) is the budgeted value of the work actually completed by that date, measured using the budget for that work, not what it actually cost. Actual cost (AC) is the cost actually incurred for the work performed by that date. Budget at completion (BAC) is the total planned value for the project: the cost baseline total.

A simple way to remember them: PV is what we planned to have done, EV is what we actually did (in budget terms), and AC is what we paid for what we did. EV is the key: comparing EV with PV shows schedule performance, and comparing EV with AC shows cost performance. Comparing PV with AC directly is meaningless, because it mixes how much work was done with how much it cost.

Variances and indices

Variances express performance in money; indices express it as ratios, which are easier to compare across projects and over time.

A useful memory aid is that EV comes first in all four formulas, and that negative or below-1 is bad. CPI is often regarded as the most important single EVM measure, because research on large projects has found that cumulative CPI tends to stabilize early, once a project is about 20 % complete, and rarely improves dramatically afterwards. That finding comes mainly from large government projects, and it is a tendency, not a law, but it warns against assuming that a cost overrun will recover by itself.

# Earned value for Project Compass at the end of month 6 ($000, invented data)
BAC = 2400
PV_cum = [150, 360, 620, 900, 1180, 1440, 1700, 1940, 2180, 2400]   # cost baseline
EV_cum = [140, 330, 560, 820, 1060, 1290]
AC_cum = [150, 360, 610, 880, 1150, 1400]

print("Mon    PV    EV    AC    SV    CV   SPI   CPI")
for m, (ev, ac) in enumerate(zip(EV_cum, AC_cum), 1):
    pv = PV_cum[m - 1]
    print(f"{m:3d} {pv:5d} {ev:5d} {ac:5d} {ev - pv:5d} {ev - ac:5d} {ev / pv:5.2f} {ev / ac:5.2f}")

PV, EV, AC = PV_cum[5], EV_cum[-1], AC_cum[-1]
CPI, SPI = EV / AC, EV / PV
eacs = {
    "BAC / CPI (current cost performance continues)": BAC / CPI,
    "AC + (BAC - EV) (remaining work at planned rate)": AC + (BAC - EV),
    "AC + (BAC - EV) / (CPI x SPI) (cost and schedule)": AC + (BAC - EV) / (CPI * SPI),
    "AC + bottom-up ETC of 1,180": AC + 1180,
}
print()
for k, v in eacs.items():
    print(f"EAC {k}: {v:,.0f}  VAC {BAC - v:+,.0f}")
print(f"TCPI to meet BAC: {(BAC - EV) / (BAC - AC):.3f}")
print(f"TCPI to meet EAC of 2,560: {(BAC - EV) / (2560 - AC):.3f}")

# Earned schedule: when should the current EV have been earned?
k = max(i for i, pv in enumerate(PV_cum, 1) if pv <= EV)
ES = k + (EV - PV_cum[k - 1]) / (PV_cum[k] - PV_cum[k - 1])
print(f"Earned schedule {ES:.2f} months at actual time 6: SV(t) {ES - 6:+.2f}, SPI(t) {ES / 6:.3f}")
print(f"Forecast duration at SPI(t): {10 / (ES / 6):.1f} months (plan 10)")

Mon    PV    EV    AC    SV    CV   SPI   CPI
  1   150   140   150   -10   -10  0.93  0.93
  2   360   330   360   -30   -30  0.92  0.92
  3   620   560   610   -60   -50  0.90  0.92
  4   900   820   880   -80   -60  0.91  0.93
  5  1180  1060  1150  -120   -90  0.90  0.92
  6  1440  1290  1400  -150  -110  0.90  0.92

EAC BAC / CPI (current cost performance continues): 2,605  VAC -205
EAC AC + (BAC - EV) (remaining work at planned rate): 2,510  VAC -110
EAC AC + (BAC - EV) / (CPI x SPI) (cost and schedule): 2,745  VAC -345
EAC AC + bottom-up ETC of 1,180: 2,580  VAC -180
TCPI to meet BAC: 1.110
TCPI to meet EAC of 2,560: 0.957
Earned schedule 5.42 months at actual time 6: SV(t) -0.58, SPI(t) 0.904
Forecast duration at SPI(t): 11.1 months (plan 10)

At the end of month 6, Project Compass planned to have done $1.44 million of work, has actually done $1.29 million of work and has spent $1.40 million. The schedule variance is −$150,000 and the SPI is 0.90: the project has achieved about 90 % of the work planned by now. The cost variance is −$110,000 and the CPI is 0.92: for every dollar spent, the project has earned about 92 cents of planned value. The monthly table shows both indices have been steady at about 0.90 to 0.93 since month 1, suggesting a consistent pattern rather than a one-off problem.

Forecasting: EAC, ETC and VAC

The estimate at completion (EAC) is the expected total cost of the project when complete. The estimate to complete (ETC) is the expected cost of the remaining work, so EAC = AC + ETC. The variance at completion (VAC) is BAC − EAC: a negative VAC means the project is expected to finish over budget. There are several ways to calculate EAC, each based on a different assumption about future performance. Choosing the right one is a common exam skill.

In the exam, the scenario tells you which assumption applies. If it says the variances are expected to continue, or gives no reason to think otherwise, use BAC / CPI. If it says a one-time problem caused the overrun and it will not recur, use AC + (BAC − EV). If it says the project must meet a fixed date and both indices matter, use the CPI × SPI version. If it says the original estimates were fundamentally flawed, re-estimate. For Project Compass, the overrun comes from consistently higher construction unit costs and slower tagging, so the team believes the pattern will continue. BAC / CPI gives about $2.6 million, and the team's bottom-up re-estimate of $2.58 million is close, which increases confidence.

The to-complete performance index

The to-complete performance index (TCPI) is the cost performance that must be achieved on the remaining work to meet a specified goal. It is the ratio of the work remaining to the funds remaining.

For Project Compass, finishing within the original $2.4 million would require a CPI of 1.110 on all remaining work, compared with the 0.92 achieved so far. That is unrealistic. A TCPI significantly above the current CPI, especially above about 1.1 when the cumulative CPI is below 1, is a signal that the original budget will not be met without major changes. If the board approves a revised target of $2.56 million, the required CPI falls to 0.957, which is achievable with modest improvement. TCPI turns a forecast into a performance target the team can understand.

Measuring earned value

EV is only as good as the method used to measure progress. Common methods include the following.

Subjective percent complete estimates ("about 80 % done") are notoriously unreliable: work often sits at 90 % for a long time. Objective methods, based on observable milestones or quantities, give better EV. Level of effort should be limited to genuine support work, because it always reports SV = 0 and can hide problems if overused.

Earned schedule

SV and SPI have a known weakness: they are expressed in money, not time, and at the end of a late project, EV eventually equals PV (all work is done, all planned value earned), so SV returns to zero and SPI to 1 even though the project finished late. Earned schedule (ES) addresses this. It finds the time at which the current EV should have been earned according to the baseline, then compares that with actual time (AT). Schedule variance in time is SV(t) = ES − AT, and SPI(t) = ES / AT.

For Project Compass, an EV of $1.29 million should have been earned at about 5.42 months according to the baseline, but the actual time is 6 months. SV(t) is −0.58 months, about two and a half weeks, and SPI(t) is 0.904. Dividing the planned duration by SPI(t) forecasts about 11.1 months instead of 10. Earned schedule is not universally used, but the PMBOK Guide and practice standards recognize it, and it gives a more intuitive answer to "how late are we?". Treat any EVM-based duration forecast as a cross-check on the critical path analysis, not a substitute for it, because EVM does not know which work is critical.

Interpreting and acting on EVM

EVM figures are indicators, not explanations. A poor CPI could come from higher prices, lower productivity, rework, poorly measured EV or costs recorded in the wrong period. A poor SPI could come from late starts on non-critical work that will not affect the finish, or from critical delays that will. Always look behind the numbers, at control account level, before deciding what to do.

Thresholds help. Many organizations set variance thresholds, for example CPI or SPI below 0.95 triggers an analysis and below 0.90 triggers escalation. When a threshold is crossed, the project manager investigates the cause, considers corrective actions such as resequencing work, improving productivity or renegotiating supplier prices, updates forecasts and reports honestly. If the forecast exceeds the budget and reserves, a change request may be needed to rebaseline, with the sponsor's approval.

For Project Compass, the project manager reports a CPI of 0.92 and SPI of 0.90 to the board, explains that construction unit costs are higher than estimated and tagging is slower than the pilot suggested, forecasts an EAC of about $2.6 million, and notes that the remaining contingency and management reserve may not cover the full overrun. She recommends a third encoding station to speed tagging, value engineering on remaining finishes, and a decision by the board on whether to fund a revised budget. PMP-121 showed how to present such figures in plain language to stakeholders.