“Earned Value Formulas: PV, EV, AC, CPI, SPI and More”

What are the four inputs every formula uses?

earned value formulas is a key topic in modern project management and teamwork. Four numbers, defined once, then reused everywhere. Getting these wrong is the root of most bad earned value reporting, because every downstream formula inherits the error.

Two disciplines make the whole system trustworthy. First, EV must come from completed work packages valued at their budgeted cost, not from a percentage someone estimated. “We are 60% done” is an opinion; “six of ten work packages signed off, budgeted at $102,000 each” is earned value. Second, PV freezes at the baseline. If you rebaseline and keep reporting against the new curve, the variances stop meaning anything — you are comparing reality to a plan that was edited to match it.

The method itself is codified in the ANSI/EIA-748 standard and forms the cost-management core of PMI’s guidance, which is why the acronym set is stable across industries even when the reporting format is not.

What are the core earned value formulas?

All of them, with what each one actually answers. Formulas are grouped so you can see the pattern rather than memorise a list.

The pattern: variances subtract, indices divide, forecasts divide by an index. If you remember only that structure, you can reconstruct any of them under pressure.

How do you read the variances?

Two subtractions, four possible signs, and one trap.

CV = EV − AC. Negative means you have spent more than the work is worth. Positive means you are under budget for the work completed.

SV = EV − PV. Negative means less value has been earned than planned. Positive means ahead.

The trap is reading SV as a schedule measure. It is expressed in currency, which makes it sensitive to your cost baseline rather than to time. A project with a $4M budget can show SV = −$400,000 and be three weeks late; the same SV on a $400,000 project represents a different delay entirely. For schedule, use SV as a directional signal and the critical path or earned schedule for the magnitude. And note that on a late project, SV converges to zero at completion for the same reason SPI converges to 1.0 — all planned work eventually gets earned.

How do you read the indices?

CPI = EV ÷ AC. At 0.89, every dollar spent has bought 89 cents of budgeted value. Below 1.0 is over budget.

SPI = EV ÷ PV. At 0.85, the project has earned 85% of the value the plan called for by this date. Below 1.0 is behind.

Both are ratios, so both are comparable across projects of different sizes — which is why portfolio reporting runs on indices and not on variances. What they are not is stable. A single month’s CPI on a project with lumpy invoicing can swing 15% for reasons that have nothing to do with efficiency, so the reading that matters is the three-period trend.

For what a CPI of 0.89 implies about recovery, what thresholds are worth escalating, and why cost indices tend to persist rather than self-correct, see the cost performance index guide. Here, the formula is enough.

Which EAC formula should you use?

Four variants, four different assumptions about the future. This is the part of earned value where analysts disagree, because the choice is a judgement about what will happen next, not a calculation.

Two supporting formulas complete the set. ETC = EAC − AC tells you the cash you still need. VAC = BAC − EAC tells you the finish-line variance; negative VAC is the expected overrun.

The rule that keeps forecasts honest: pick the variant before you see the answer, write down why, and do not switch variants mid-project because the new one looks better. Switching methods when the number gets uncomfortable is how forecasts lose credibility with sponsors.

Worked example: a $2.4M project at month 9

Eighteen-month programme, month 9 status date, half the schedule elapsed.

Working through every formula:

• CV = 1,020,000 − 1,140,000 = −$120,000 (over budget for work done)
• SV = 1,020,000 − 1,200,000 = −$180,000 (behind plan)
• CPI = 1,020,000 ÷ 1,140,000 = 0.89
• SPI = 1,020,000 ÷ 1,200,000 = 0.85
• EAC (typical) = 2,400,000 ÷ 0.89 = $2,682,000 (a $282,000 overrun)
• ETC = 2,682,000 − 1,140,000 = $1,542,000
• VAC = 2,400,000 − 2,682,000 = −$282,000
• TCPI = (2,400,000 − 1,020,000) ÷ (2,400,000 − 1,140,000) = 1,380,000 ÷ 1,260,000 = 1.10

Now the same project under the other two EAC assumptions:

• Atypical: EAC = 1,140,000 + (2,400,000 − 1,020,000) = $2,520,000. This says the 42.5% delivered was delivered badly, but the remaining 57.5% will be fine. Possible; rarely true once a trend has formed.
• Compounding: EAC = 1,140,000 + [1,380,000 ÷ (0.89 × 0.85)] = 1,140,000 + 1,824,000 = $2,964,000. This says delay is now driving cost, which is what happens when a fixed overhead team runs three months longer.

Three defensible forecasts between $2.52M and $2.96M, from identical data. That spread is not a flaw in the formulas — it is the uncertainty, made visible. Present the range and the reasoning, not a single number with false precision.

The TCPI of 1.10 is the most actionable output. It says the remaining work needs 110% cost efficiency against a project currently running at 89%. That is not a plan; it is a description of how unlikely the current budget is to hold.

Worked example: what a funding cap actually requires

The same project, now with a sponsor who refuses to fund the $2.68M forecast and caps total spend at $2.6M. The question changes from “what will it cost” to “what would it take.”

TCPI against the capped budget:

TCPI = (BAC − EV) ÷ (EAC_cap − AC) = (2,400,000 − 1,020,000) ÷ (2,600,000 − 1,140,000) = 1,380,000 ÷ 1,460,000 = 0.95

That is a very different message. To finish inside $2.6M, the remaining work must run at 95% efficiency — better than the current 89%, but not heroic. The gap is 6 index points. At 0.89, the forecast is $2.68M; at 0.95, it is $2.6M; the difference is about $80,000, which is roughly 3% of the total.

This is why TCPI deserves a place in every monthly report. It turns “we need to do better” into a number a team can accept or reject, and it lets you test a cap in thirty seconds rather than a week of re-estimating.

Worked example: why SPI lies at the end

A three-year infrastructure project, BAC $12M, baseline duration 36 months. It finishes in month 39.

The project is three months late, and its final SPI is a perfect 1.0. This is not a calculation error. PV stops accumulating at the baseline finish, while EV continues until all work is earned, so the ratio is guaranteed to converge — by definition, SPI at contract completion is 1.0.

Two consequences. Never use SPI as the primary schedule signal in the last third of a project, and never report a rising SPI as recovery without checking. A project can show improving SPI while slipping further, because the denominator has frozen.

The standard fix is earned schedule, developed by Walt Lipke in 2003. It converts EV into the point in time when that value was planned to be earned, then compares that month to the actual month: SPI(t) = ES ÷ AT. On the project above, at month 38 the earned value of $11.85M corresponds to roughly month 35.5 on the baseline curve, giving SPI(t) ≈ 0.93 — behind, and staying behind. Earned schedule does not converge to 1.0, because time keeps moving.

Earned schedule is an extension rather than a replacement: keep the cost formulas as they are, and add SPI(t) once you are past the halfway point.

What are the most common earned value mistakes?

Comparing AC to PV. A classic. AC is what you spent; PV is what you planned to have earned. Comparing them mixes cost and schedule and tells you nothing. AC compares to EV.

Using subjective percent complete as EV. If the percentage comes from a conversation rather than from completed, valued work packages, every formula downstream is an opinion with a decimal point.

Rebaselining to fix the variance. Rebaselining has legitimate uses after an approved scope change. Doing it because the variance is embarrassing destroys the baseline’s only function.

Reporting one EAC with no method. A forecast without a stated assumption cannot be audited or challenged, and it will be challenged the moment it is wrong.

Ignoring the trend. A single CPI reading is noise. Three consecutive readings moving in one direction is information.

Forgetting that EV is cost-weighted. Because EV is measured in budgeted currency, a project that front-loads expensive work will look ahead early and behind late even if the work is proceeding exactly on plan. Check the shape of your baseline curve before drawing conclusions in month two.

What are the limits of earned value?

Earned value measures cost and schedule against a baseline. It does not measure quality, benefit, or scope creep — a project can earn 100% of its value while delivering something nobody wants. It also assumes a stable baseline and reasonably granular work packages, which is why it fits construction, engineering, and government contracting better than discovery-phase product work.

The honest position for most teams: use the four inputs and four core formulas as a monthly discipline, add TCPI when a forecast needs to become a decision, and use earned schedule when the finish date matters more than the cost. Skip the full 32-criteria ANSI/EIA-748 apparatus unless a contract requires it — the marginal insight per hour of administration drops sharply after the basics.

How does Doitify fit into this?

Doitify is our product, so read this as disclosure. Earned value needs four numbers, and the reason teams skip it is that three of them are painful to assemble: planned value from the baseline, earned value from completed work packages, actual cost from finance.

In Doitify, tasks, sub-tasks, checklists, owners, and due dates carry the progress data, work reports roll up planned against actual effort, and Gantt charts and roadmaps hold the baseline that PV depends on. Project documents keep the budget and the change log beside the work. That does not compute EV for you — the valuation of a completed work package is still a judgement — but it removes the assembly step that usually kills the monthly cycle.

The value is cadence. Earned value once a quarter is an audit; earned value once a month is a control system, and control is the only reason to do the arithmetic.

assumes both cost and schedule pressure continue. A bottom-up ETC replaces the assumption with a fresh estimate.|Should I use SV or SPI for schedule status?::Use both as a directional check and neither as your only measure. SV is in currency and therefore sensitive to budget size, and SPI converges to 1.0 at completion. For late-stage projects, earned schedule (SPI(t)) is more reliable.|What is a good TCPI?::Below 1.0 means the remaining work can be less efficient than planned and still hit the target. Between 1.0 and about 1.10 is demanding but credible. Above 1.10 against a project already running below 0.95 is usually evidence the target budget is not achievable.|How often should earned value be calculated?::Monthly for most projects, matching the reporting cycle, and more often during a recovery period. Weekly EV on a project with monthly invoicing produces noise, because AC arrives in lumps and the index swings for reasons unrelated to performance.|Is earned value only for large projects?::No, but the overhead has to be proportionate. A simple version — four inputs, CV, SV, CPI, SPI, EAC — takes under an hour a month on a mid-size project and catches cost drift earlier than a burn-rate chart. The full ANSI/EIA-748 apparatus is where the cost becomes hard to justify.”]

Conclusion

Earned value formulas are a small, closed system. Four inputs — BAC, PV, EV, and AC — produce two variances, two indices, a forecast, and a required efficiency. Everything else is a variation on the assumption you make about the future.

Use the worked example above as a template, pick your EAC variant before you look at the result, and report the range rather than a single number. Then add the one formula most teams ignore: TCPI, which turns a forecast into a target the team can argue with. That is the difference between measuring a project and managing one.

If this post on earned value formulas was helpful, you might also enjoy How to Automate Task Assignment (Step-by-Step) and Project Documentation: The Complete Guide (Types, Examples, and Best Practices).