09 · Advanced Metrics (EVM Deep Dive & Forecasting)

Level 2 introduced CPI and SPI as a health check. This module goes further: the full family of EVM formulas, three different EAC forecasting methods that give different answers depending on what you assume about the future, and the To-Complete Performance Index that tells you exactly how hard the remaining work has to go to hit a target.

The complete EVM formula reference

Metric	Formula	Meaning
PV (Planned Value)	Budgeted cost of work scheduled	What you planned to have spent by now
EV (Earned Value)	Budgeted cost of work actually performed	What the completed work is "worth"
AC (Actual Cost)	Actual cost of work performed	What you actually spent
CV (Cost Variance)	EV − AC	Positive = under budget
SV (Schedule Variance)	EV − PV	Positive = ahead of schedule
CPI (Cost Performance Index)	EV / AC	>1 = efficient spending
SPI (Schedule Performance Index)	EV / PV	>1 = ahead of schedule
BAC (Budget at Completion)	Total planned budget	The original target
EAC (Estimate at Completion)	See three methods below	Projected final cost
ETC (Estimate to Complete)	EAC − AC	Projected remaining cost
VAC (Variance at Completion)	BAC − EAC	Projected final over/under
TCPI (To-Complete Performance Index)	(BAC − EV) / (BAC − AC)	CPI required on remaining work to hit BAC

Three EAC forecasting methods

The choice of EAC formula is a statement about what you believe caused past variance and whether it will continue.

Method	Formula	Assumption
EAC (typical variance)	BAC / CPI	Past cost performance will continue exactly as-is
EAC (atypical variance)	AC + (BAC − EV)	The variance so far was a one-off; remaining work proceeds at the original budget rate
EAC (both CPI and SPI matter)	AC + [(BAC − EV) / (CPI × SPI)]	Both cost and schedule performance will continue to affect remaining cost

Worked example: three EACs from one dataset

A project: BAC = $2,000,000, PV = $900,000, EV = $750,000, AC = $825,000.

python3 -c "
BAC, PV, EV, AC = 2000000, 900000, 750000, 825000
CPI = EV/AC
SPI = EV/PV
eac_typical = BAC / CPI
eac_atypical = AC + (BAC - EV)
eac_both = AC + (BAC - EV) / (CPI * SPI)
print(round(CPI,4), round(SPI,4))
print(round(eac_typical,2), round(eac_atypical,2), round(eac_both,2))
"

Result: CPI = 0.9091, SPI = 0.8333.

Method	EAC
Typical variance (BAC/CPI)	$2,200,000
Atypical variance (AC + BAC − EV)	$2,075,000
Both CPI and SPI (AC + (BAC−EV)/(CPI×SPI))	$2,475,000

The spread between $2,075,000 and $2,475,000 — a $400,000 difference from the same underlying data — is exactly why a PM must state which method and which assumption they're using whenever they quote an EAC. Reporting "$2.2M" with no method named lets the number quietly become "the number" without anyone checking whether the underlying assumption (past performance persists unchanged) still holds. The "both" method is the most conservative here and the right default when a project is both over budget and behind schedule, since it assumes neither problem self-corrects.

TCPI — what the remaining work must achieve

python3 -c "
BAC, EV, AC = 2000000, 750000, 825000
tcpi_to_bac = (BAC - EV) / (BAC - AC)
print(round(tcpi_to_bac,4))
"

→ TCPI = 1.0638. To still hit the original $2,000,000 budget, every remaining dollar spent must return $1.0638 of planned value — a CPI 6.38% better than the original plan assumed, on the harder, later part of the project. Compare this to the current CPI of 0.9091: the team would need to improve performance by roughly 17% just to hit the original number. This is the single number that tells a steering committee whether "we'll still hit budget" is a plan or a hope — a TCPI more than about 5–10% above the project's demonstrated CPI is a hope.

Variance thresholds and RAG bands

Metric	Green	Amber	Red
CPI	≥ 0.95	0.90 – 0.94	< 0.90
SPI	≥ 0.95	0.90 – 0.94	< 0.90
TCPI vs. current CPI gap	≤ 3% harder	3–10% harder	> 10% harder

At CPI 0.9091 (Red) and a TCPI-vs-CPI gap of about 17 percentage points (Red), this project's dashboard entry is unambiguous: it needs the recovery process from module 08, not a "monitor and continue" note.

Forecasting with a trend, not a single snapshot

A single period's CPI can be noise. Track it across periods before treating it as a trend:

Period	CPI	SPI	Trend interpretation
Month 1	0.98	0.97	Normal
Month 2	0.95	0.93	Slight softening — watch
Month 3	0.91	0.88	Consistent decline — this is a trend, not noise
Month 4	0.91	0.83	Confirmed — trigger recovery review

Three consecutive periods of decline (months 2–4) is the general rule of thumb for distinguishing a trend from a single bad sprint — one bad month after five good ones is usually noise; three in a row is a signal.

How It Actually Works

The three EAC formulas from Level 2 Module 3 are, formally, three different statistical models of how a cost variance propagates, and choosing between them is a modeling decision with a right answer given the data: EAC = AC + (BAC − EV) implicitly assumes the variance-to-date was a non-recurring shock and future work reverts to the original planned rate (appropriate only if you can point to a specific, resolved, one-off cause); EAC = BAC / CPI assumes the current cumulative efficiency is the best predictor of all future efficiency (appropriate for a systemic cause, like consistently underpriced labor, that won't self-correct); and the "both" method, EAC = AC + [(BAC − EV) / (CPI × SPI)], additionally discounts remaining work by schedule performance, on the empirical basis that projects running behind schedule tend to also spend more per unit of remaining work (overtime premiums, expediting costs, resource contention) — so it's the appropriate model specifically when both indices are underperforming together, not when only one is. Running all three against one dataset and getting three materially different numbers isn't a methodology failure, it's the model correctly telling you that the choice of assumed root cause changes the forecast by that much — which is itself diagnostic information worth reporting alongside the number.

Exercise

A project: BAC = $1,200,000, PV = $500,000, EV = $410,000, AC = $470,000.

1. Calculate CPI, SPI, and all three EAC methods, verifying every step with python3 -c. State which method you'd report to a steering committee and why, given this project's specific CPI/SPI combination.
2. Calculate TCPI to BAC, and compare it to the current CPI to say whether hitting the original budget is realistic.
3. Assign this project's CPI and SPI a RAG color using the thresholds table, and state what additional data (beyond this one snapshot) you'd want before recommending a full recovery process.