SPI and CPI: the two ratios that say if you are late or over budget

Reading project performance · Part 2 of 2

Two ratios off one earned value figure. SPI divides it by the plan and reads schedule. CPI divides it by the bill and reads cost. Below 1 is bad news in both, for different reasons.

Two ratios off one earned value figure. The schedule performance index divides it by the work the baseline planned by now, so below 1 means less is done than scheduled. The cost performance index divides it by what that work has cost, so below 1 means it cost more than budgeted. Two denominators, two independent verdicts.

Two ratios off one earned value figure

Both indices are built from the same three quantities, and the article on earned value, planned value and actual cost is where those come from. In one line: planned value is the budget for the work the baseline said would be finished by today, earned value is the budget for the work that actually is finished, and actual cost is what finishing it cost.

Subtract those and you get variances in pounds. Divide them and you get the two indices, and the division is what makes them travel. A schedule variance of minus £25,000 means nothing to a reader who does not know whether the project is worth half a million or half a billion. An SPI of 0.86 means the same thing on both, which is why the ratio form is what ends up on the report and the money form is what stays in the working file.

ReadingSPI, scheduleCPI, cost
FormulaEV / PVEV / AC
Below 1Behind. Less work done than the baseline planned.Over. The work done cost more than its budget.
Exactly 1Work done matches the baseline to date.Work done cost exactly its budget.
Above 1Ahead. More work done than planned.Under. The work cost less than its budget.
Blind spotBlind to sequence. A late critical task and a late task with float count the same.Blind to quality. Cheap work still reads as efficient.

The shared numerator is the thing to hold on to. Get earned value wrong and both indices move together, in the same direction, by the same proportion. A pair that always agrees is not confirmation. It is a single number reported twice.

The schedule index counts work, not weeks

SPI is earned value divided by planned value. Take a £500,000 project four months in, with the baseline scheduling 35 percent complete by now and 30 percent actually complete: £150,000 earned against £175,000 planned, so SPI is 0.86. For every pound of work the plan expected by today, the project has produced 86 pence of it.

Notice what that sentence does not say. It does not say the project is 14 percent late, or two weeks late, or any number of days late at all. SPI is a ratio of two money figures, so its verdict is denominated in work rather than in time, and the conversion between the two is not arithmetic. A project can be 0.86 on the index and one day late because the missing work is a single trivial package, or 0.97 and three months late because the missing work is the one activity everything else waits on.

Never quote SPI as though it converts to a date. Turning 0.86 into a forecast finish requires going back to the schedule, finding which activities are actually missing, and asking whether any of them sit on the critical path. The index tells you to go and look. It does not tell you what you will find.

That distinction is why the index and the schedule get read together or neither is worth reading. The GAO’s schedule assessment guide makes confirming a valid critical path one of its ten best practices, on the grounds that the critical path is what determines the earliest completion date and what tells a team which activities decide the outcome. An index built from summed budgets has no view of that path at all. It weighs a pound earned on a task with six weeks of float exactly the same as a pound earned on the task holding the end date.

The cost index counts efficiency, not spend

CPI is earned value divided by actual cost. Our project has earned £150,000 and spent £160,000, so CPI is 0.94. For every pound spent, 94 pence of budgeted work came back.

Read that carefully, because the common misreading is that CPI compares spend against budget. It does not. Comparing spend against budget tells you whether the money is going out faster than planned, which is a cash question and a useful one, but it says nothing about whether the money is buying anything. CPI compares spend against work delivered. A project underspending because nothing is happening looks calm on a cash report and shows up on CPI only when the work it did produce turns out to have cost too much.

CPI is also the more stable of the two, and the more useful for forecasting. Divide the budget at completion by the current cost performance index and you get a rough estimate at completion: £500,000 divided by 0.94 is roughly £532,000, which is the answer to the question of what the whole project costs if today’s efficiency holds to the end. That forecast is crude, it assumes a rate that no project actually holds flat, and it is still the first number worth putting in front of a sponsor, because it converts a ratio nobody feels into a figure everybody does.

Set the threshold before you need it. A band of 0.9 to 1.1 with a defined action at each edge turns the indices into a control rather than a commentary, and it stops the monthly conversation restarting from first principles about whether 0.94 is worrying. Agree the band at baseline, when nobody is defending a number.

Four combinations, four different conversations

Two indices, each above or below 1, gives four quadrants, and each carries a genuinely different management response.

Schedule performance 0.86 £150,000 earned against £175,000 planned.
Cost performance 0.94 £150,000 earned for £160,000 spent.
Estimate at completion £532,000 Budget at completion divided by CPI, if the rate holds.
The four combinations of schedule performance index and cost performance index, with the verdict each one carries A square divided into four quadrants by a vertical line at schedule performance index 1.0 and a horizontal line at cost performance index 1.0. The top right quadrant, both indices above 1, is ahead and under budget. The top left, schedule below 1 and cost above 1, is late and cheap. The bottom right, schedule above 1 and cost below 1, is on time and being bought with overtime. The bottom left, both below 1, is late and over budget. A marked point sits in the bottom left quadrant at 0.86 by 0.94, labelled month four. CPI, cost performance1.31.00.70.71.01.3SPI, schedule performanceSPI below 1, CPI above 1Late and cheap. Work is nothappening. Look for blockers,not for savings.SPI above 1, CPI above 1Ahead and under budget. Rareenough to check the baselinewas not simply generous.month fourSPI 0.86, CPI 0.94SPI below 1, CPI below 1Late and over budget. Neitherindex recovers on its own.SPI above 1, CPI below 1On time, and being bought.Overtime, extra people, rework.
The vertical line is SPI 1.0 and the horizontal line is CPI 1.0. The worked example sits in the bottom left at 0.86 by 0.94, the only quadrant where neither index is doing you a favour. The two quadrants either side of it are the interesting ones, because a single healthy index there is usually paying for the unhealthy one.

The two mixed quadrants are where the reporting usually goes wrong, because each of them contains a healthy-looking number that is being paid for by the other one. A project ahead of schedule and over cost has bought its date with overtime, contractors or rework, and the cost index is the receipt. A project behind schedule and under cost has usually not saved anything at all; it has simply not done the work yet, and the underspend is a bill deferred rather than avoided. Neither situation is visible from one index alone, which is the practical reason the pair is always quoted as a pair.

Our worked example is in the bottom left, and the bottom left is the honest quadrant. Both numbers agree, nothing is being disguised, and the only question left is which activities are missing and why.

Where both indices quietly stop working

Three limits, and the first is the one that catches experienced people.

SPI drifts back to 1.0 as a project ends, whether or not the project is late. Earned value climbs to the budget at completion on every project that finishes, and planned value has already arrived there, so the ratio of the two converges on 1 by construction. The APM’s earned value group describes the classical schedule variance as a v shaped curve that will “slowly start to return to zero, as it charts the difference between EV and PV”, in a paper on earned schedule written precisely because of that behaviour. A project reading 0.78 at month ten will read 0.95 at month fifteen and 1.00 on the day it hands over, three months after the date it promised. SPI is a mid-flight instrument. On the last approach it stops reading, which is exactly when people quote it hardest.

Both indices are lagging. They describe work already done and money already spent, and they turn a corner only after the thing that caused the turn has finished happening. Read as a trend across six periods they are genuinely predictive, because efficiency is sticky and a project running at 0.94 rarely recovers to 1.0 without something changing. Read as a single point against last month’s single point, they are noise with a decimal place.

And neither index has any view of quality. Earned value credits work as complete against a crediting rule, not against whether the output survives testing. Rework arrives as fresh actual cost with no fresh earned value behind it, so a project quietly redoing its own work shows a falling CPI and an SPI that looks untroubled for months.

An index is a ratio of two numbers somebody chose. It cannot be more honest than the baseline underneath it.

Which is the real caveat under all three. SPI and CPI inherit everything from the plan they divide by: the decomposition, the crediting rule, the resource rates, the sequencing. Recalculate them against a baseline that was reset in April and they will report a healthy project with no history, because every rebaseline resets both indices to 1.0 and throws away the evidence that got them there. The ratios are not a verdict on the project. They are a verdict on the project measured against a document, and the document is the part worth arguing about first.

Common questions

What is the difference between SPI and CPI?
The difference is the denominator. Both indices divide earned value, the budget for the work genuinely complete, but the schedule performance index divides it by planned value and the cost performance index divides it by actual cost. SPI therefore answers whether enough work is done, and CPI answers whether the work that is done cost what it was budgeted to cost. One can sit above 1 while the other sits below it, which is the normal case rather than the odd one.
What is the schedule performance index?
The schedule performance index is earned value divided by planned value, expressed as a ratio rather than a sum of money. An SPI of 0.86 means the project has completed 86 pence of work for every pound the baseline scheduled by that date. Above 1 is ahead of the baseline, exactly 1 is on it, below 1 is behind it. The index measures how much work is done, never which activities are late.
How do you calculate SPI and CPI?
Divide earned value by planned value for the schedule performance index, and earned value by actual cost for the cost performance index. On a project that has earned £150,000 against a planned £175,000 for a spend of £160,000, SPI is 150 over 175, which is 0.86, and CPI is 150 over 160, which is 0.94. Both figures must be cumulative to the same status date, or the two ratios describe different projects.
What does an SPI or CPI below 1 mean?
Below 1 means the project is getting less than a pound of value for every pound of the thing in the denominator. A schedule performance index below 1 means less work is complete than the baseline scheduled, so the project is behind. A cost performance index below 1 means the completed work cost more than its budget, so the project is over. Neither figure tells you by how many days or which activities, only by how much on the whole.

Filed under Performance and metrics

Read next

Private beta, building in the open

Get the next one

Practitioner writing on project delivery, plus progress notes as the tool takes shape.

One email at launch. Unsubscribe in one click.