Earned Value Management (EVM) explained

Everything in EVM is derived from three quantities measured at a point in time, plus the total budget. Get these right and the rest is arithmetic.

Planned Value (PV) is the budgeted cost of the work that should have been completed by now according to the baseline. Earned Value (EV) is the budgeted cost of the work that has actually been completed — value is "earned" at the budgeted rate, not at what it really cost. Actual Cost (AC) is what you have genuinely spent to complete that work. Budget at Completion (BAC) is the total planned budget for the whole project.

From PV, EV and AC you get the two variances that tell you whether you are ahead or behind. Schedule Variance (SV = EV − PV) compares work done to work planned: a negative SV means less value has been earned than the baseline promised, so the project is behind. Cost Variance (CV = EV − AC) compares the value of work done to what it cost: a negative CV means you have spent more than the work was worth, so the project is over budget.

Expressing schedule in dollars of value rather than calendar days is what makes EVM powerful — and also where its first blind spot hides. A project can show a healthy schedule variance while a single late task on the true constraint quietly threatens the finish date, because SV treats all earned value as equal regardless of which work earned it.

Variances are absolute; indices make performance comparable across projects of any size. The Schedule Performance Index (SPI = EV ÷ PV) is efficiency against the schedule: 1.0 is on plan, below 1.0 is behind, above 1.0 is ahead. The Cost Performance Index (CPI = EV ÷ AC) is efficiency against the budget: 1.0 is on budget, below 1.0 means each dollar bought less work than planned.

These two ratios are the heartbeat of EVM reporting. An SPI of 0.9 and a CPI of 0.85 says you are getting 90 cents of scheduled progress and 85 cents of budgeted value for every dollar and day you spend — a concise, defensible signal for a board deck or a sponsor review.

EVM's most valuable output is the forecast. Estimate at Completion (EAC) projects what the project will ultimately cost given performance so far. The common formula, EAC = BAC ÷ CPI, assumes today's cost efficiency continues to the end — so a CPI of 0.8 on a $1M budget forecasts a $1.25M finish. Estimate to Complete (ETC = EAC − AC) is what remains to be spent from here.

A Variance at Completion (VAC = BAC − EAC) then tells you the projected overrun or underrun before it happens. This is the discipline's real payoff: a credible, math-backed answer to "where will we land?" computed from progress to date rather than optimism.

EVM is rigorous, but it is fundamentally backward-looking and deterministic. Every metric reports a single number with no probability attached: an EAC of $1.25M sounds precise, but it carries no confidence interval, no P50 or P80, and no sense of how wide the real range of outcomes is. It extrapolates the past in a straight line and says nothing about the risk events that have not occurred yet.

For R&D programs the gaps are sharper still. EVM has no model of uncertainty — a biotech CMC vendor slip or a deep-tech TRL gate failure that has not happened yet is invisible to SPI and CPI until it lands. It treats all earned value as equal, so it can mask a constraint on the critical chain. And SPI famously drifts back toward 1.0 near the end of a project even when the finish date is in jeopardy, because remaining planned value shrinks. EVM tells you where you have been with authority; it does not tell you how likely you are to hit a future date.

EVM and probabilistic methods are complements, not rivals. EVM answers "how have we performed against the plan to date?" with cost-and-schedule discipline. Monte Carlo schedule risk analysis answers the question EVM cannot: "given the uncertainty and the risks still ahead, what is the probability we finish by a given date?" — producing P50/P80/P90 finish dates, a criticality index, and a tornado of the biggest drivers.

Critical Chain Project Management adds the constraint-aware view: it protects the finish with project and feeding buffers and tracks buffer consumption on a fever chart, so you are watching the one number that actually guards the deadline. Run together, EVM tells you where you stand, Monte Carlo tells you where you are likely to end up, and CCPM tells you whether your protective buffer can absorb what is left. That stack is far more honest than any single technique — and it is exactly the combination a deterministic Gantt or a standalone EVM report cannot give you.

CritPath AI computes EVM — SPI, CPI, EAC and the full BAC/PV/EV/AC set — directly from the same schedule and cost model that drives its risk engine, so your earned-value report and your Monte Carlo outcomes never disagree about the underlying plan. EVM sits alongside, not instead of, the probabilistic view: P50/P80/P90 finish dates, criticality index, tornado sensitivity, CCPM buffers, and a live fever chart.

An AI copilot built on Claude and Gemini, grounded in your actual dependency graph, reads all of it together — explaining why your CPI is drifting, which task is eroding your buffer, and what a gate decision does to your forecast finish. It is $10 per user per month with every standard feature, and AI usage billed separately by metered usage, in a modern web app rather than a legacy desktop tool.