SPI and CPI Interpretation: Three Ways They Mislead You

An SPI above 1.0 looks like good news. It is not always good news, and the gap between "looks fine" and "is fine" is exactly where most PMOs get surprised by a critical path that was late the whole time. SPI and CPI interpretation goes wrong in three specific, recurring ways: the ratio blends critical and non-critical progress into one number, it weights that blend by dollar value rather than by what actually determines the finish date, and it is mathematically forced toward 1.0 as the project nears completion. CPI has its own trap: a healthy-looking CPI can be propped up by cost that was earned but not yet invoiced.

The direct answer: SPI = EV / PV measures cost-weighted schedule efficiency; CPI = EV / AC measures cost efficiency. Both are useful, both are gameable, and neither should be read as a single number in isolation. A project can show SPI comfortably above 1.0 while its actual critical path runs weeks behind, because SPI averages every task's dollar-weighted progress rather than reporting the one task chain that sets the finish date.

SPI and CPI Interpretation Starts With What Each Ratio Measures

	SPI (Schedule Performance Index)	CPI (Cost Performance Index)
Formula	EV / PV	EV / AC
Measures	Schedule efficiency, in dollars	Cost efficiency
Above 1.0 means	Ahead of the planned earn rate	Under budget per unit of work
Below 1.0 means	Behind the planned earn rate	Over budget per unit of work
Blind spot	Blends all tasks by dollar weight	Assumes recorded cost is complete
Converges to at completion	Always 1.0	Reflects final cost, not forced

The full derivation of both ratios, plus EAC and ETC forecasting, is covered in earned value management explained; this post assumes you already have the numbers and focuses on where reading them goes wrong.

Three Ways an SPI Above 1.0 Still Means Late

1. SPI blends critical and non-critical progress into one number. Take a project with three open tasks: two large, easy, non-critical tasks worth $600,000 and $500,000 of planned value, both running slightly ahead (110 percent and 105 percent of planned value earned), and one small, $100,000 critical-path task that has only earned 60 percent of its planned value. Total planned value is $1,200,000; total earned value is $660,000 + $525,000 + $60,000 = $1,245,000. Blended SPI = 1,245,000 / 1,200,000 ≈ 1.04. The aggregate reads "ahead of schedule," and the one task chain that actually sets the finish date is running 40 percent behind.

2. SPI weights by dollar value, not by schedule importance. The example above is not a fluke of small numbers: SPI is deliberately dollar-weighted, so two large, easy, front-loaded tasks will always outvote one small, hard, critical one. A schedule with a lot of cheap, dependency-heavy work and a few expensive, easy tasks produces this pattern routinely, and the ratio has no way to know that the $100,000 task is the one that determines the finish date and the $600,000 task is not.

3. SPI mathematically converges to 1.0 at completion, whatever the real finish date is. This is the same structural flaw covered in earned schedule explained: once EV reaches the full budget, dollar-based SPI reaches 1.0 regardless of whether the project finished on time or eighteen months late. Read SPI(t) from earned schedule as the project nears completion instead of trusting dollar-based SPI at exactly the moment it becomes least reliable.

The CPI Gaming Check: Unbilled Work

CPI has its own version of the same trap, and it runs through the denominator instead of the numerator. CPI = EV / AC, and AC is only honest if it reflects everything actually spent, whether or not it has been invoiced yet.

Say a vendor performs $80,000 of real work in a period but has only submitted a $30,000 invoice by the reporting cutoff, with the rest to follow next month. If Actual Cost is recorded on an invoice basis, AC reads $30,000 this period even though $80,000 of value has been earned against it: CPI = EV / AC balloons to a number well above what the vendor's real spend justifies, and it will fall back down hard the moment the rest of the invoice lands.

The check that catches it: reconcile recorded Actual Cost against the vendor's accrued or committed cost for the period, not only against invoices received by the cutoff date. A CPI that jumps upward right after an unusually low-invoicing period, with no matching change in scope or productivity, is the specific pattern to flag; it usually means AC understated real spend rather than the project genuinely getting cheaper to run.

Read Both Ratios Against Their Trend, Not a Snapshot

A single period's SPI or CPI is a data point, not a diagnosis. Three consecutive periods of a declining CPI is a real pattern; one low period after a vendor invoicing delay is noise that resolves itself next month. The habit that actually protects a PMO from both traps in this post is the same habit either way: pair the blended ratio with the critical path status directly, so an SPI reading above 1.0 never stands in for having actually checked whether the one task chain that sets the finish date is on time. The rest of the blog's PMO and schedule controls library covers the surrounding practice, from earned schedule to baseline drift reading.