Chapter 30 — Earned Value Management for Construction: Measuring Progress in Dollars and Days

The Hook: Two Reports and One Question

Nadia Haddad's office is on the fourth floor of Kestrel's building in Rivermont, and it has exactly one decoration: a whiteboard with four job names on it and a number next to each. She is the Vice President of Operations. The numbers change on the first Tuesday of every month.

It was the first week of February, Year 2. Month eleven of Northgate. She had two documents on the desk in front of her and she had read both of them before I sat down.

The first was Lorena Vasquez's month-eleven cost report. Cost codes, budget, committed, cost to date, cost to complete, forecast at completion. At the bottom of the forecast column, a variance: the job was $118,000 under budget.

The second was Wei Chen's month-eleven schedule update. Data date January 31. As-built logic through the curtain-wall start. Forecast substantial completion: September 27, Year 2 — nine calendar days past the contract date of September 18.

Nadia turned the first one ninety degrees so it faced me, put the second one next to it, and asked the question.

"So are we ahead or behind?"

I said the thing everybody says. "We're under budget and behind schedule."

"Ray. That is not an answer. That is two answers to two different questions, and neither one of them is the question I asked." She tapped the cost report. "This says we have money. That one" — the schedule — "says we don't have time. On this job time is money. Ten thousand six hundred and fifty dollars a day. So which is it?"

I could not answer her, and I want you to understand exactly why, because it is not because I was unprepared. I had read both reports twice. The problem was structural. The cost report was denominated in dollars. The schedule update was denominated in days. Neither document knew the other one existed. The cost report did not know how much building we had actually built; it only knew what we had spent. The schedule did not know what anything cost; it only knew what had finished. You can stare at those two pages for an hour and you will not find the answer, because the answer is not in either of them.

Wei had been sitting quietly at the end of the table with a single sheet of paper. They slid it across.

There were four numbers on it.

Wei let me look at it for about ten seconds and then narrated it in one breath.

"We planned to have $25,760,000 of the work in place by the data date. We actually have $24,970,000 of it in place, measured at the prices in our own budget — not at what it cost, at what we said it would cost. And putting that much work in place cost us $25,642,000. So we are behind by $790,000 of work and over by $672,000 of money."

Nadia looked at the cost report. "Then why does this say we're under budget?"

"Because it compares what we spent to what we planned to spend," Wei said. "Twenty-five million seven hundred sixty planned, twenty-five million six hundred forty-two spent. A hundred eighteen thousand under. That comparison is arithmetically correct and it is completely meaningless, because we haven't done the work yet. We're not under budget. We're behind, and being behind looks exactly like being under budget until the month you catch up."

That is the chapter. Four numbers, one unit, and a question that two separate reports could not answer between them.

🏃 Fast Track: If you already know PV, EV, and AC cold, skim 30.2 through 30.4 and go straight to 30.5 (the four estimates at completion and when each one is the honest assumption), 30.6 (the full Northgate work-package report and the money check on Kestrel's fee), and 30.9 — the section on where earned value misleads on a construction project, which is the half of this subject that certification courses skip.

🔬 Deep Dive: The arithmetic conventions and rounding rules are in Appendix A. Cost loading a schedule is the mechanical prerequisite and it starts in Chapter 14 §14.9; the CPM work that earned value cannot replace is in Appendix B and Chapter 29. Where the man-hour version of all of this lives is Chapter 20.

30.1 The Idea, Before Any Acronyms

Here is the whole concept in three sentences, and you should read them before you read a single Greek letter or capitalized abbreviation.

You cannot tell whether a project is doing well by comparing money spent to money budgeted. That comparison has a hole in the middle of it the size of the building: it does not know how much work you got for the money. A project that has spent 60 percent of its budget could have built 70 percent of the building or 40 percent of it, and the cost report looks identical either way.

Earned value fixes the hole by measuring, in dollars, the value of the work you have actually completed — and then comparing that one number against two others: what you planned to have completed by now, and what it actually cost you.

One unit. Two questions answered. That is the entire invention.

Everything else in this chapter — every acronym, every index, every forecast formula — is arithmetic on that idea. If you understand the idea, you can rebuild the rest from scratch on a napkin. If you memorize the formulas without the idea, you will compute a schedule performance index of 1.02 on a job that is thirty-four days late and you will not know why the superintendent is looking at you like that.

🧩 Productive struggle. Before you read on, take three minutes with this. You are running a $900,000 sitework package. The plan said you would have moved 60,000 cubic yards of the 100,000 CY by today. You have moved 48,000 CY. You have spent $460,000.

Write down, in your own words and without formulas, the answers to three questions: Are you ahead or behind on production? Are you spending efficiently? And what will this package cost when it is done? Then write down which single number you had to invent in order to answer the second question. Do not read on until you have written something.

30.2 The Three Base Measures (and Two Supporting Ones)

Three measurements. Every one of them is a dollar figure, and every one of them is stated as of a specific date.

PV — Planned Value

Planned value (PV) is the budgeted cost of the work scheduled to be complete by the data date. It is what the baseline said you would have earned by now. Older texts call it BCWS, the budgeted cost of work scheduled; you will still see that on federal reports and in some software. Same thing.

PV comes from the cost-loaded schedule — the baseline schedule with dollars attached to activities (Chapter 14 §14.9). If your budget lives only in cost codes and never touches a schedule activity, you cannot compute PV at all, and roughly half of the failed earned-value implementations I have seen died right there. More on that in 30.7.

EV — Earned Value

Earned value (EV) is the budgeted cost of the work performed — the value, at budget rates, of what you have actually built. Older name: BCWP, budgeted cost of work performed.

💡 Aha moment. EV is measured at budget rates, not at what the work cost you. This is the single most reliably misunderstood idea in the entire subject, so let me say it three ways.

• If your budget says drywall is $2.85 per square foot and you have hung 100,000 SF, your earned value is 100,000 × $2.85 = $285,000 — whether it cost you $250,000 or $340,000.
• Earned value never contains actual cost. If you find actual cost leaking into your EV, your CPI will be exactly 1.000 forever and your report is worthless.
• Earned value is a measure of production, priced in dollars so that it can be compared to something. The dollars are a unit of measure, not a payment.

The reason this matters: EV and AC are the two sides of the cost-efficiency question. If both sides come from the same actual-cost data, there is no question left to ask.

AC — Actual Cost

Actual cost (AC) is what you actually spent to perform the work that has been performed. Older name: ACWP, actual cost of work performed.

AC is not "cash paid." It is cost incurred — labor charged to the code, material received, sub work performed and approved, equipment consumed. Getting this right is Chapter 28's job, and the usual failure is timing: invoices arrive weeks after the work, so a naive AC understates cost and flatters your CPI. If you do not accrue uninvoiced work, you are not measuring performance; you are measuring your accounts-payable department's turnaround time.

BAC and the data date

BAC — budget at completion is the total budget for whatever you are measuring: a cost code, a work package, or the whole job. On Northgate the BAC for the cost of the work is $40,000,000 — the first line of the GMP build-up, sitting inside a $47,500,000 guaranteed maximum price.

The data date is the moment the report describes. Everything is "as of" it. Work performed after the data date does not count; cost incurred after the data date does not count. Northgate's month-11 data date is January 31, Year 2 — 334 calendar days into a 565-day contract.

A report without a data date is not a report. It is a rumor with a table in it.

The whole thing on one cost code

Abstractions are where good intentions go to die, so here it is on one activity, with small numbers.

Cost code 03-330 — slab on grade. Northgate's slab is 33,000 SF at 5 inches (510 CY of concrete). The budget is $214,500, which is $6.50 per square foot placed, reinforced, finished, and cured. The plan was six pours over six weeks; by the end of week four the schedule said pours one through four would be down — 24,000 SF.

At the week-four data date, Jamal Foster's crew has placed 21,450 SF. The cost report shows $148,900 charged to 03-330.

And here is why AC is not just labor — this is the whole cost, and if you leave any of it out you will build a beautiful, wrong report:

Now look at what a conventional cost report would say about this code. Budget to date $156,000, spent $148,900 — "under budget by $7,100." Everyone relaxes.

Here is the truth: you got $139,425 of slab for $148,900 of money. You are over cost by $9,475 and behind by $16,575 of production. The "under budget" reading was an artifact of not having done the work.

🔄 Check your understanding. A package has PV of $400,000, EV of $340,000, and AC of $325,000. Is it under budget?

30.3 The Variances: Two Subtractions, One Trap

Two subtractions. Both use EV as the anchor, which is the point — everything is measured against what you actually built.

Cost variance:

CV = EV − AC

Positive means the work was worth more than it cost you. Negative means you spent more than the work was worth. On Northgate at month 11: CV = $24,970,000 − $25,642,000 = −$672,000. You have spent $672,000 more than the work you have in place is worth at your own budget prices.

Schedule variance:

SV = EV − PV

Positive means you have completed more work than planned. Negative means less. On Northgate: `SV = $24,970,000 − $25,760,000 = −$790,000`. You are $790,000 of work short of the plan.

The trap, and it is a big one

Schedule variance is denominated in dollars, not days. It is a measure of work volume, not of time. And it has a property that makes it dangerous if you do not know about it:

SV converges to zero at completion — on every project, including one that finishes a year late.

Here is why, and the mechanism is trivial once you see it. At completion, all the work is done, so EV = BAC. And the baseline eventually runs out of planned work too, so PV = BAC. Therefore SV = BAC − BAC = 0.

A project that finishes fourteen months late will show a schedule variance of exactly zero on its final report, and a schedule performance index of exactly 1.000. The metric does not know what a calendar is. It knows how much work is left, and at the end there is none left, so it reports success.

🔍 Why this works — and why it fails. SV measures the gap between two cumulative curves that must both end at the same value. Any two curves that start at zero and end at BAC will converge, no matter how differently they got there. The gap between them is largest in the middle of the job and mathematically must be zero at the end. That is not a flaw in anyone's implementation. It is a property of the construction of the metric.

Practically, SV is a useful early- and mid-project indicator and a useless late-project one. On a 19-month job, trust it through about month 14 and stop trusting it after that. What replaces it is the CPM schedule — the forward pass, the backward pass, the total float, and the critical path from Chapter 14 and Chapter 29. There is also a formal extension called earned schedule that converts the variance into time units; we get to it in 30.9.

The one conversion you are allowed to make, carefully

People want SV in days. You can approximate it, and you should label the approximation loudly.

Divide SV by the planned burn rate at the data date. Northgate's baseline planned $3,800,000 of work in month 11 (cumulative $25,760,000 less month 10's $21,960,000), across 31 calendar days:

Planned daily rate = $3,800,000 ÷ 31 CD = $122,581 per calendar day SV in days ≈ −$790,000 ÷ $122,581 = −6.4 calendar days

So earned value says roughly six and a half days behind. The CPM says nine. Those two numbers are not the same number, and they are not measuring the same thing. The six-and-a-half is "how much work volume are we short, expressed in the currency of average days." The nine is "how far has the longest chain of dependent work moved." They agree here by rough coincidence and they will not always agree. When they disagree, the CPM wins, every time, because the CPM knows about logic and the earned value does not.

30.4 The Indices: Efficiency in a Single Ratio

Variances tell you the size of the problem. Indices tell you the rate, which is what you forecast from.

Cost performance index:

CPI = EV ÷ AC

CPI is how many dollars of work you get per dollar spent. At month 11 on Northgate:

CPI = $24,970,000 ÷ $25,642,000 = 0.974

For every dollar Kestrel spends on Northgate, it is getting 97.4 cents of budgeted work. That is a 2.6 percent inefficiency, and on a $40,000,000 base, 2.6 percent is over a million dollars.

Schedule performance index:

SPI = EV ÷ PV

SPI = $24,970,000 ÷ $25,760,000 = 0.969

The job has produced 96.9 percent of the work volume it planned to produce. Carry the same caveat forward: this is work volume, not time. SPI drifts to exactly 1.000 at completion for the same reason SV drifts to zero.

Reading the numbers

Two refinements that keep you from misreading the table.

Small denominators lie. A package that is 2 percent complete will produce indices that swing wildly on a single invoice. On Northgate's specialties package at month 11, EV is $23,000 against AC of $29,000 — CPI 0.793, which looks catastrophic and means almost nothing. Set a threshold (I use 20 percent complete or $250,000 of EV, whichever comes first) below which you look at the package but do not forecast from it.

Trend beats level. A CPI of 0.96 that has been 0.96 for six months is a known, priced condition. A CPI of 0.99 that was 1.03 two months ago is an emergency in progress.

The CPI-stability observation, presented honestly

There is a finding that gets repeated constantly in the project-controls world, originally from analysis of large government programs and echoed by practitioners across industries:

CPI tends to stabilize surprisingly early in a project — often somewhere around the 15 to 20 percent complete mark — and rarely improves much after that.

I want to be careful here. That observation comes largely from large, long-duration government acquisition programs, not from commercial construction, and I have not seen a body of published construction-specific research that establishes it with the same weight. Treat it as a widely reported practitioner finding, not a law. AACE International and PMI both teach forecasting approaches consistent with it; neither one promises it.

But treat it seriously anyway, because the mechanism is plausible and the failure mode is expensive.

🔍 Why this works. Consider what actually drives cost performance on a construction package: the accuracy of the estimate, the quality of the design documents, the price you bought the subcontract at, the site conditions, the crew, and the sequence. Every one of those is largely fixed before the work starts. Nothing about reaching the 40 percent mark makes your drawings better or your subcontract cheaper. And arithmetically, the denominator gets large: at 60 percent complete, moving a CPI of 0.95 back to 1.00 requires the remaining 40 percent to run at about 1.09, which is a performance nobody has demonstrated on this job.

So the honest reading of an early bad CPI is not "it is early, it will recover." It is: this is probably the number, and I have found out early enough to do something about it. Which is the entire point of measuring.

🏗️ From the field. I got this wrong once and it cost me a fee. A four-story senior-living project, month four of eighteen. CPI came in at 0.94. I wrote one sentence in the narrative — "early-job variance, expected to normalize" — and I believed it. Month eight: 0.94. Month twelve: 0.93. Month sixteen: 0.93. Final: 0.93, and $310,000 of a $340,000 fee gone. Nothing normalized, because nothing about the situation changed. The bad drawings were still bad, the mechanical sub was still underpriced, and the site still had the same access. The number was never noise. It was the project's actual efficiency, reported to me twelve months before anyone was willing to hear it.

🔄 Check your understanding. A package is 15 percent complete with a CPI of 0.92. Your superintendent says "we'll get it back on the repetitive floors." What do you need to hear before you believe that?

30.5 The Forecasts: The Part That Actually Drives Decisions

Variances and indices describe the past. Nobody spends money on the past. The forecast is where earned value earns its keep — and it is also where earned value is most often abused, because there are four standard formulas and they give four different answers.

ETC and EAC

Estimate to complete (ETC) is what the remaining work will cost from here. Estimate at completion (EAC) is the total: EAC = AC + ETC. And **variance at completion (VAC) = BAC − EAC` — the forecast overrun or underrun.

Here are the four ways to get EAC, and — far more important than the formulas — the assumption each one is quietly making.

Let me work all four on the same Northgate data, so you can see the spread. As of January 31, Year 2: BAC $40,000,000, EV $24,970,000, AC $25,642,000, CPI 0.974, SPI 0.969, and remaining budgeted work BAC − EV = $15,030,000.

The spread is $892,812 — from $40,672,000 to $41,564,812 — on the same four input numbers, on the same day, for the same job. Nobody is lying. Every one of those is a correctly applied standard formula.

Notice something about formula #2: its VAC is exactly −$672,000, which is exactly the cost variance to date. That is not a coincidence — formula #2 assumes the remaining work is perfectly efficient, so the only overrun that ever exists is the one you have already incurred. Formula #2 can never forecast a problem that has not already happened. That is its entire appeal to a project manager who does not want to deliver bad news, and it is why "we'll hold the rest to budget" is the most common sentence in a bad monthly report.

Which one do you report? My rule, and it is not complicated:

Report #1 as the baseline forecast, #4 as the checked forecast, and #2 only when you can name the event, show it is closed, and prove the remaining work is unaffected. If you use #2, write the justification in the narrative in one sentence. If you cannot write that sentence, you do not get to use the formula.

TCPI — the credibility test

The to-complete performance index (TCPI) is the efficiency the remaining work must achieve in order to hit a stated target.

TCPI to BAC = (BAC − EV) ÷ (BAC − AC) TCPI to EAC = (BAC − EV) ÷ (EAC − AC)

Remaining work to do, divided by remaining money to do it with. On Northgate, if you want to finish at the $40,000,000 budget:

TCPI = ($40,000,000 − $24,970,000) ÷ ($40,000,000 − $25,642,000) = $15,030,000 ÷ $14,358,000 = 1.047

The remaining $15,030,000 of work must be executed at 1.047 — about 7.5 percent better than budget — by a team that has run at 0.974 for eleven consecutive months and has never posted a cumulative month above 1.000.

This is the most useful thing TCPI does. It converts a forecast into a testable claim. Run it against each of the four EACs and the pattern is immediate:

Note that the TCPI for formula #1 is always exactly the current CPI, and for formula #2 it is always exactly 1.000. Those are useful sanity checks that your spreadsheet is right.

And here is the rule to carry out of this section, which you can apply to any forecast anybody hands you for the rest of your career:

If a reported EAC requires a to-complete performance index of 1.18 from a team that has been running at 0.94, that is not a forecast. It is a wish with a decimal point.

📋 Try it — run the full metric set on one work package.

Unit masonry, cost code 04-200, on a mid-size commercial job. Data date is the end of month 8.

Compute CV, SV, CPI, SPI, all four EACs, VAC, and TCPI to BAC. Then answer the two questions that actually matter: (a) is this package in trouble, and (b) what single additional piece of information would you need in order to know whether the project is late?

For EAC #4, assume the mason's own bottom-up re-estimate of the remaining work is $402,000.

Work it with a pencil before you open the answer. Twelve minutes.

30.6 The Full Northgate Report at Month 11

This is the centerpiece. Everything above, applied to a real job, at a real data date, with the project rolling up to the canonical $40,000,000 cost of the work inside the $47,500,000 GMP.

Data date: January 31, Year 2. Day 334 of 565 — 59.1 percent of contract time elapsed.

Check the footing before you read the interpretation, because a report that does not foot is a report nobody believes: BAC sums to $40,000,000; PV to $25,760,000; EV to $24,970,000; AC to $25,642,000. CV totals −$672,000 and SV totals −$790,000, both of which reconcile to the project row exactly.

Percent complete by package (EV ÷ BAC), because you cannot read the indices without it:

Reading it in prose

Time elapsed 59.1 percent. Work earned 62.4 percent. Money spent 64.1 percent. Those three percentages, in that order, are the fastest possible read of a job, and this one says: we are producing slightly faster than the calendar, and spending faster than we are producing. The gap between the second and third numbers is the entire problem, and it is $672,000 wide.

The healthy packages. Sitework finished at a CPI of 1.021 — Kestrel's earthwork subcontractor found a closer fill placement site than the estimate assumed and the 32,000 CY of net export cost less per yard to haul. That is a real $48,000 and I would like more of them. Precast at 1.024 is going well because the crane sequence and the panel delivery matched the plan. Roofing at 1.016 is unremarkable and unremarkable is the goal.

The known, priced problem. Steel and deck closed at a CPI of 0.974, which is a $168,000 cost variance. That number should look familiar: it is exactly what the acceleration cost. When the anchor-bolt and embed submittal sat in our office eleven days, Caldwell Structural took its full fourteen-day review, Ironbridge missed its mill slot, and erection start slipped from August 4 to August 27 — twenty-three calendar days. Kestrel bought back seventeen of them with a second erection crew, premium time, and a resequenced enclosure for $168,000. That decision was made deliberately, in Nadia's office, with the arithmetic on the table (Chapter 29). It is the only line in this report that is a decision rather than a discovery, and that distinction is worth noticing: an EVM report is mostly a list of things you found out.

Trouble package #1 — curtain wall, CPI 0.857. This is the one that will hurt. At 54 percent complete, running at 85.7 cents of work per dollar, the CPI extrapolation forecasts $5,750,426 against a $4,930,000 budget — $820,426 over. That is 122 percent of the entire construction contingency, sitting in one package. The cause is not mysterious: the deck-edge geometry moved during the steel acceleration, the unitized panels are meeting embeds that are not where the model said they were, and every unit is getting field-shimmed and, on the north elevation, field-modified. This is the delayed bill for the acceleration, arriving five months later in a different cost code. Because it is 54 percent complete and the cause is still present, formula #1 is the honest formula here.

Trouble package #2 — the one that looks fine. Mechanical shows CPI 0.997 and SPI 1.000. Perfect. Look at how those numbers were made: EV was taken as Cardinal Mechanical's billed percentage on their schedule of values, and AC is what we approved for payment. When EV and AC come from the same document, CPI is 1.000 by construction and measures nothing at all. The 0.997 is a rounding artifact from a small backcharge. This package's indices are not a measurement; they are a restatement of Cardinal's pay application. We will come back to this in 30.7 and again in Chapter 32, and Case Study 1 shows what happened when Wei went out and physically verified it.

Trouble package #3 — framing and drywall, CPI 0.920, 15 percent complete. Small dollars: the cost variance is only $45,000. Everybody's instinct is to ignore it because it is early. Do not ignore it. Fifteen percent complete on a $3,450,000 package with a CPI of 0.920 is the single most actionable line on this page, because it is the only trouble you have found early enough to fix cheaply. The recovery required to hit budget is (3,450,000 − 517,000) ÷ (3,450,000 − 562,000) = 2,933,000 ÷ 2,888,000 = 1.016 — the remaining 85 percent must run 1.6 percent better than budget. That is achievable, unlike the curtain wall. This is the package where a crew change, a layout change, or a material-staging fix actually returns the money.

The noise line. Specialties at CPI 0.793 is 1.9 percent complete. Ignore the index, note the package, forecast it at budget.

⚠️ Safety alert. A report like this becomes dangerous at the moment it leaves the trailer without a system attached to it. If page one lands on Margo Deacon's desk as "you are 0.857 on curtain wall and 0.749 on drywall" and the only response is pressure, you have just built the third finding of the week-34 scaffold investigation — a crew running behind under an unwritten make it up expectation. Bea Salgado's whole argument in Chapter 24 is that schedule pressure is a hazard exactly like an unguarded edge, and a bad index communicated as a demand rather than as a problem to be solved is how that hazard gets installed. Every metric you publish comes with an obligation to publish the system change alongside it.

The forecast, and the EAC column that does not foot

Sum the EAC column and you get $41,623,447. The project row says $41,076,492. They are $546,955 apart and neither one is a spreadsheet error.

Package EACs do not sum to the project EAC because CPI is a weighted ratio — the project CPI weights each package by its actual cost, while summing package EACs weights each package by its own extrapolation, including the ridiculous ones. Two packages are doing all the damage: specialties, at 1.9 percent complete, contributes $323,478 of pure extrapolation noise, and framing and drywall, at 15 percent, contributes $300,290.

Apply the discipline from 30.4 — do not extrapolate a package below 20 percent complete; carry it at budget or bottom-up — and the picture changes:

The disciplined package sum and the project rollup land $76,813 apart on a $41 million forecast — about two-tenths of one percent. That agreement is the most valuable thing in this report. Two methods computed different ways converging is evidence. One method by itself is an opinion.

Wei's bottom-up estimate to complete, prepared package by package with Margo and the trade superintendents, came in at $15,598,000 against $15,030,000 of remaining budget — implying the remaining work runs at 15,030,000 ÷ 15,598,000 = 0.964, slightly worse than to date. EAC #4 = $25,642,000 + $15,598,000 = $41,240,000.

One detail worth stopping on. Wei's bottom-up for curtain wall was $2,645,000. The CPI extrapolation for the same package is $5,750,426 − $3,105,000 = $2,645,426. They agree within $426. That is not luck: the CPI extrapolation is correct precisely when the cause of the overrun is still present, and on the curtain wall the deck edge is still where it is.

💰 Money check: what the month-11 EAC does to Kestrel's fee

The GMP build-up carries $1,320,000 of construction contingency and $575,200 of escalation allowance, of which $310,000 has already been drawn for steel and copper. Available buffer against a cost-of-work overrun: $1,320,000 + $265,200 = $1,585,200.

Kestrel's fee is $1,804,800 — 4.0 percent of the $45,120,000 subtotal. And unused contingency is shared 75 percent owner / 25 percent Kestrel, so returned contingency is real money to Kestrel.

The choice of formula is worth $223,203 to Kestrel — the difference between the top and bottom rows. Which is exactly why the person whose bonus depends on the answer should not be the person who picks the formula, and why Nadia asks Wei for the EAC and not me.

And a number that appears nowhere on any of these tables: the nine-day slip carries 9 CD × $10,650/CD = $95,850 of extended general conditions and liquidated damages. Earned value does not know about liquidated damages. BAC is the budget for the work; it has no line for being late. That is the first of several things this instrument cannot see, and 30.9 lists the rest.

30.7 The S-Curve: Three Lines That Tell You Everything

📊 Diagram (described). Plot cumulative dollars on the vertical axis against time on the horizontal, and draw three curves: PV (the baseline — the cost-loaded schedule), EV (what you have built, at budget prices), and AC (what you have spent). All three start at the origin. PV and EV both must end at BAC. AC ends wherever the job actually costs. The vertical gap between PV and EV, read horizontally, is your schedule problem; the vertical gap between EV and AC is your cost problem. It is the same S-curve you built in Chapter 14 §14.9, with two more lines on it.

Northgate, monthly, on the $40,000,000 cost of the work:

Read the CPI column top to bottom. At month 3 it was 0.978. At month 11, after an acceleration, a change of superintendent on the enclosure, and eight more months of effort, it is 0.974. It never got better. The EAC column has been sitting between $40.7 million and $41.4 million since month three — nine months before anyone in this story asked the question.

Now the same data as a picture. One character equals roughly one percent of the $40,000,000 BAC.

    Cumulative % of the $40,000,000 cost of the work
    0%        10%       20%       30%       40%       50%       60%       70%
    |---------|---------|---------|---------|---------|---------|---------|
M2  PV   2.9% ###
    EV   2.8% ###
    AC   2.9% ###

M4  PV   9.5% ##########
    EV   9.4% #########
    AC   9.6% ##########

M6  PV  20.4% ####################
    EV  19.2% ###################
    AC  19.6% ####################

M8  PV  35.6% ####################################
    EV  33.2% #################################
    AC  34.0% ##################################

M10 PV  54.9% #######################################################
    EV  53.1% #####################################################
    AC  54.7% #######################################################

M11 PV  64.4% ################################################################
    EV  62.4% ##############################################################
    AC  64.1% ################################################################

Watch the EV bar. It tracks PV closely through month 4, falls away sharply at months 6 through 8 — that is the steel delay landing — recovers at 9 and 10 as the acceleration bought work back, and drifts out again at 11 as the curtain wall bogs down. The picture tells the project's history without a single word of narrative, which is why it belongs on page one of every monthly report you ever issue.

The four positions, and what each one means

That third row deserves a second read. A project that is behind schedule frequently shows a good CPI for a while, because the work not yet started cannot overrun. Then it staffs up to recover, adds overtime and a second shift, and the CPI collapses in two months. A rising CPI on a falling SPI is not two pieces of news. It is one piece of news arriving in the wrong order.

🔄 Check your understanding. At month 7, Northgate's SV was −$720,000 and its CV was −$210,000. At month 11, SV is −$790,000 and CV is −$672,000. Which of those two trends is worse, and why?

30.8 Making It Work on a Real Construction Project

Everything to this point is arithmetic, and the arithmetic is the easy part. Most earned-value implementations on construction projects fail for four specific reasons, none of which is mathematical.

1. The budget must be loaded onto schedule activities, not just cost codes

PV is the budgeted cost of work scheduled. If your budget lives only in a cost-code structure, there is no schedule in it, and PV cannot exist. You need a cost-loaded schedule: every activity in the CPM carrying its share of the budget, summing exactly to BAC.

That is real work — on Northgate, mapping a $40,000,000 budget across 1,340 activities took Wei and Lorena about three weeks, and it is a mapping that must be maintained through every change order for the life of the job. Two practical rules make it survivable:

• Do not cost-load every activity. Load at the level you will actually report at — Northgate loads about 260 of the 1,340 activities and the rest carry zero. The zero-dollar activities still drive logic and float; they just do not earn.
• The mapping is many-to-many and you must own it. One cost code (03-330 slab on grade) spans six pour activities; one activity (level-3 overhead rough-in) spans four cost codes. Build the crosswalk once, in writing, and make it a controlled document.

2. Measuring EV honestly — pick the method before the work starts

There are several legitimate ways to claim earned value, and each carries a built-in bias.

The rule that makes all of this work: the measurement method is chosen and written down before the work starts, activity by activity, and it does not change afterward.

A method chosen after you see the numbers is not a measurement method. It is a negotiating position. Somebody who is behind will discover a preference for 50/50 in exactly the month they need one, and because the choice is defensible in isolation, nobody will catch it.

3. The subcontracted-work problem

Northgate is roughly 85 percent subcontracted, and this changes what earned value can tell you.

For a subcontract package, EV is almost always taken as the percentage on the subcontractor's pay application, and AC is what you approved for payment. Those two come from the same document, so:

CPI on a subcontract package is approximately 1.000 by construction, and it measures your pay application review — not the subcontractor's efficiency.

Look at the Northgate table again with that in mind. Mechanical: 0.997. Electrical: 0.999. Elevators: 0.985. Those are not measurements of Cardinal's or Halcyon's productivity. Kestrel does not have Cardinal's labor hours and never will. Those numbers say only that Kestrel approved roughly what Cardinal billed.

This is not a reason to abandon earned value on subcontracted work. It is a reason to understand what you are actually measuring, and to fix the input:

• The SPI on a subcontract package is still real and still useful, because PV comes from your schedule and EV comes from their billing. A subcontract package with an SPI of 0.936 (Halcyon) is genuinely behind your plan, whatever the CPI says.
• Verify the billed percentage physically before you accept it as EV. Send your field engineer with a count: terminal units set, valves installed, LF of duct hung, panels terminated. This is the same walk that protects your pay application review in Chapter 32, and it is the only thing standing between your EVM report and a subcontractor's optimism.
• Where the subcontract is large enough, negotiate a units-complete schedule of values at buyout (Chapter 16). It costs you an argument in month one and saves you an argument in month fourteen.

Your EVM is only as honest as your pay application review. Write that on something.

4. Data timeliness

An earned-value report on month-old data is a history lesson. The value of the whole instrument is that it gives you a signal early enough to act, and every week of lag burns some of that value.

Three lag sources, and what to do about each:

Kestrel's rule on Northgate: data date is the last calendar day of the month; the report is on Nadia's desk on the fifth working day. Five working days of lag on a job that costs $10,650 a day to be wrong about is an acceptable trade. Twenty is not.

⚖️ What the contract says. Earned-value reporting shows up in construction contracts in three distinct ways, and they carry different obligations.

Federal work. Federal acquisition regulations require an earned value management system on certain major acquisitions, and agencies commonly reference the industry EVM system guideline published as ANSI/EIA-748. For most federal construction contracts, though, the requirement is narrower than a formally validated EVMS: what you will typically see is a specification requiring a cost-loaded CPM schedule, monthly updates on a stated data date, and progress payments computed from the earned value of scheduled activities. The applicability thresholds and the exact wording change; read the solicitation, not your memory, and not this book.

Large institutional and healthcare owners. Universities, health systems, and school districts increasingly require monthly cost-loaded schedule updates with earned-value summaries as a condition of payment. That clause is worth finding before you bid, because it obligates you to build and maintain the cost-loaded schedule described above, and that is a staffing decision.

Open-book GMP — what the owner is entitled to see. Under a CM-at-Risk GMP like Meridian's, the owner has audit rights over the cost of the work: the actual costs, the payrolls, the subcontract values, the invoices. Because unused contingency is shared 75/25, the contingency log and its current status are the owner's business too — Meridian is a stakeholder in that number. What is arguably not the owner's business is Kestrel's internal forecast of its own fee, its bonus accruals, and its home-office margin analysis.

Two lines you should not cross in either direction. Reporting an EAC you do not believe, in a contract where the owner is relying on your open-book reporting, is a misstatement to a party that has a contractual right to accurate cost information — it is not optimism, and "we thought we'd get it back" is a poor defense in a room with an auditor in it. Equally: an EAC is not a claim. Reporting a forecast cost overrun does not create entitlement, does not constitute notice of a claim, and does not obligate the owner to anything. If you want relief, the notice provisions in Chapter 31 and Chapter 33 are how you get it, and they run on their own clocks. Many contracts also require prompt written notice of a forecast overrun; check yours, because those clauses vary by contract and by jurisdiction.

30.9 Where Earned Value Misleads on a Construction Project

This is the section the certification courses skip, and it is the reason this chapter exists in a construction textbook rather than a project-management one.

SV and SPI are not schedule, and the counterexample proves it

A project can show an SPI of 1.02 while its critical path is thirty-four days late. Not theoretically. Routinely. Here is exactly how, using Curtis Boone's job — Rivermont Elementary School

12, $22,400,000 hard-bid lump sum, 610 calendar days, at a month-9 data date.

The project reports an SPI of 1.020 — ahead of schedule. The gymnasium, which is the critical path, is running at 0.845, and the CPM update shows the gym chain at minus 34 days of total float. Substantial completion has moved thirty-four calendar days.

Three mechanisms produced this, and they compound:

1. Non-critical work was completed early. Curtis's crews had float in the sitework and the classroom wing, and when the gym slowed down they moved to work that was available. That is reasonable field management. It also earns value — $1,174,000 of it — which is why the aggregate number went positive.
2. The schedule of values is front-loaded. Mobilization, general conditions, and early trades were priced high in the SOV so that Curtis's cash flow would be positive early (Chapter 32). Because EV on this job is computed from the SOV percentage, front-loading the SOV directly inflates EV. The project earns value faster than it builds building.
3. Level-of-effort dilution. $1,900,000 of general conditions earns exactly its plan every single month, by definition. That is 8.5 percent of the BAC permanently pinned at SPI 1.000, dragging the aggregate toward the middle.

What the CPM said that the EVM could not: that the gym roof structure, the gym overhead MEP, and the gym finishes form a single chain with no float, that the chain is 34 days late, and that 34 CD × ($2,100 extended general conditions + $2,500 liquidated damages) = $156,400 of exposure is now real. There is no arithmetic operation you can perform on PV, EV, and AC that will produce the number 34. The only instrument that produces it is the forward and backward pass.

Everything else EVM cannot see

Float, logic, and the critical path. EVM has no concept of a predecessor. It cannot tell you that finishing $500,000 of drywall on level 2 is worth nothing if the level-2 ductwork is not in yet. Two packages with identical CV, SV, CPI, and SPI can have wildly different consequences, and the difference is total float. Read the EVM report and the float report side by side or do not read either.

Front-loaded schedules of values. Covered above and worth repeating because it is also an ethics line. Modest front-end loading to cover genuine mobilization cost is normal and defensible. Loading the SOV so that the job bills ahead of its progress is borrowing from the owner without asking, and it corrupts your own EVM in the process — you lose the ability to see your own project. The contractor who front-loads is the contractor most likely to be surprised in month fourteen.

Approved change orders move BAC. Every executed change order changes the budget at completion and must be added to the performance measurement baseline, along with its PV in the right months. If the baseline is not maintained, the indices become meaningless within about four months. Two rules: pending changes stay out of BAC until executed, and every baseline change gets a log entry with a date, an amount, and an authorizing document. A baseline you can silently move is a baseline you can always hit.

Level-of-effort dilution. Northgate's general conditions are $2,900,000 and are pure LOE. At month 11, GC PV and EV are both $1,714,000 (334 of 565 days), while AC is $1,802,000 — staff overtime and extra temporary heat during the acceleration. Fold that line into the project and watch what happens:

SPI improved and CPI worsened, and neither movement reflects one thing that happened in the field. The SPI improved because an LOE line earns exactly its plan and therefore pulls the aggregate toward 1.000. The CPI worsened because general conditions happen to be running over. Keep level-of-effort work in its own bucket, report it separately, and never let it into the index you use to forecast.

Liquidated damages, extended general conditions, and acceleration. BAC is the budget for the work. It contains no line for being late. Northgate's nine-day slip is $95,850 of exposure that appears nowhere in the EVM report.

The conclusion this chapter is driving at

Earned value management is a cost-performance instrument. It gives you an early, honest read on efficiency and a defensible, testable forecast of final cost. It is a poor schedule instrument. Use it alongside the CPM, never instead of it.

The failure mode this whole section is trying to prevent is a project manager standing in an owner's meeting saying "our SPI is 1.02, we are ahead of schedule" while the superintendent knows the building will not be dry in March. Do not be that person. Report cost performance from earned value. Report schedule from the critical path. When somebody asks for one number, tell them there are two, and tell them why.

🪞 Learning check-in. Go back to the four Northgate EACs in 30.5 — $40,672,000, $41,076,492, $41,240,000, and $41,564,812.

Which one did you want to be true?

Be honest with yourself, because everyone has an answer and most people's answer is the second one. Then sit with three questions:

1. When you looked at the four formulas, did you evaluate the assumptions, or did you scan for the number and then find the assumption that justified it? Both feel identical from the inside. The only difference is the order.
2. What is your personal default when a metric is ambiguous — optimism or pessimism? Neither is a virtue. Both are a bias, and knowing which one is yours is worth more than any formula in this chapter, because you can correct for a bias you have named.
3. Think of a project, a class, or a piece of work where you were behind and told yourself you would make it up. Did you? Run the TCPI on it in retrospect: what improvement rate were you actually assuming, and had you ever demonstrated it? That is the same test, applied to you.

The technical skill in this chapter takes an afternoon. The skill of not flinching when the number is bad takes a career, and it is the one that separates the project managers who get told the truth by their teams from the ones who do not.

30.10 Earned Schedule, and Earned Value in Man-Hours

Two extensions, one of which fixes the SV problem and one of which is what most contractors actually use in the field.

Earned schedule: converting the variance into time

Earned schedule (ES) answers the question SV cannot: at what point on the baseline should we have earned the value we have actually earned? You find today's EV on the PV curve and read off the time.

Northgate at month 11. EV is $24,970,000. On the baseline, month 10 cumulative PV is $21,960,000 and month 11 is $25,760,000, so the EV falls between them:

ES = 10 + [($24,970,000 − $21,960,000) ÷ ($25,760,000 − $21,960,000)] = 10 + ($3,010,000 ÷ $3,800,000) = 10 + 0.792 = 10.79 months

Actual time (AT) is 11.00 months. So:

SV(t) = ES − AT = 10.79 − 11.00 = −0.21 months ≈ −6.4 calendar days SPI(t) = ES ÷ AT = 10.79 ÷ 11.00 = 0.981 Forecast duration = 565 CD ÷ 0.981 = 576 CD — about 11 calendar days late

Three things to notice. First, SPI(t) of 0.981 is not the same as SPI($) of 0.969, and the time-based one is the more meaningful of the two. Second, earned schedule does not converge to 1.000 at completion; if the job finishes at day 576, ES stops at 565 while AT keeps running, and the index correctly reports lateness. That is the whole point of the extension. Third, its forecast of 11 days sits near the CPM's 9 days — close, and still not the same number, and the CPM is the one you defend in a claim because it is built from logic rather than from a curve.

Be honest about adoption: earned schedule is well established in the project-controls literature and in the aerospace and defense world, and it is not widely used in building construction. Most owners do not ask for it and most contractors do not produce it. Learn it because it is genuinely better than SV and because it takes about four minutes to compute; do not expect to find it in your next specification.

Earned value in man-hours — closing the loop with Chapter 20

Here is the thing that should have been nagging at you since 30.2. In Chapter 20 you learned the productivity factor:

PF = earned man-hours ÷ actual man-hours, where earned MH = quantity installed × budgeted unit rate

Read that next to the definition of CPI and there is nothing to distinguish them but the unit. Earned man-hours are earned value denominated in hours. The productivity factor is a cost performance index computed in a currency a foreman can actually count.

This is why many contractors run earned value on man-hours rather than dollars for self-perform work: a superintendent can count 61 cubic yards and 100 man-hours. Nobody in the field can see a dollar.

Jamal Foster's self-perform concrete at Northgate, at package completion:

Actual man-hours charged: 7,570.

PF = 7,363 ÷ 7,570 = 0.973

Now convert to dollars at the budgeted burdened rate of $54.12 per hour, and use the actual average rate of $55.40 (Saturday premium time during the acceleration):

EV = 7,363 MH × $54.12 = $398,486 AC = 7,570 MH × $55.40 = $419,378 CPI (labor) = $398,486 ÷ $419,378 = 0.950

PF is 0.973 but the labor CPI is 0.950, and the gap is entirely the rate. Decompose it exactly the way Chapter 20 taught:

Cross-check: EV − AC = $398,486 − $419,378 = −$20,892, agreeing to a dollar of rounding. The crew was 2.7 percent less productive than the estimate and got paid 2.4 percent more per hour than the estimate assumed, and those are two different management problems with two different owners.

One last observation from the Northgate table that this explains. The concrete package CPI is 0.973 while the concrete labor CPI is 0.950. The package looks better than the labor because concrete material was bought below the estimate, offsetting part of the labor loss. A package index is a blend, and the blend can hide the thing you can actually manage. You manage labor. You do not manage the price of ready-mix after the purchase order is signed. Report both.

🔄 Check your understanding. Why can a contractor run meaningful earned value on man-hours for self-perform work but not for subcontracted work?

Spaced Review

Recall first, then read. Cover the answers.

1. From Chapter 29 — what is the earliest reliable warning that a schedule is in trouble, and why is it earlier than a missed milestone?

Recall. — Float erosion. Total float on a chain of activities shrinking update over update is a warning weeks or months before anything is actually late, because a chain consumes its float before it consumes the finish date. A missed milestone tells you the problem has already happened; float erosion tells you it is happening. Notice how this pairs with what you just learned: EVM has no concept of float at all. Northgate's SPI of 0.969 and its nine-day critical-path slip are two independent findings, and only one of them can move the substantial completion date.

2. From Chapter 29 — the delay taxonomy. Name the two questions you ask about any delay, and what each answers.

Recall. — Is it excusable? — meaning does the contractor get time. And is it compensable? — meaning does the contractor get money. The two are independent, which produces four categories: excusable and compensable (owner-caused), excusable but non-compensable (weather, most force majeure), non-excusable (contractor-caused — no time, no money, and liquidated damages run), and concurrent delay, which is where the arguments live. Northgate's steel slip was non-excusable: the submittal sat in Kestrel's office for eleven days. That is why the $168,000 acceleration shows up as a cost variance on the steel package instead of as a change order.

3. From Chapter 28 — name the five columns of a cost report, and say which one actually tells you whether you are making money.

Recall. — Budget, committed, cost to date, cost to complete, forecast at completion. The one that matters is cost to complete — the forecast of the remaining work — because cost-to-date is history and a job that is 60 percent billed and 75 percent spent is already in trouble while every column looks defensible. This chapter gave you a second, independent way to produce that forecast: ETC = EAC − AC. When the cost report's bottom-up cost-to-complete and the EVM's CPI-based ETC agree, you can defend the number. When they disagree by more than a few percent, one of them is wrong and finding out which is this month's most valuable work.

4. Deep callback to Chapter 20 — what is the productivity factor, and what did you just learn that it actually is?

Recall. — PF = earned man-hours ÷ actual man-hours, where earned hours are installed quantity times the budgeted unit rate. What you now know: it is a cost performance index in hours. You have been computing earned value since Chapter 20 without calling it that. The difference between PF (0.973 on Northgate's concrete) and the labor CPI (0.950) is entirely the wage rate variance, and separating them tells you whether you have a production problem, a pay problem, or both.

Project Checkpoint: The Willow Street Earned Value Report

In Chapter 28 you built a cost-code structure, loaded the budget, and produced a month-6 cost report with a cost-to-complete forecast. In Chapter 29 you updated the CPM schedule at the same data date with as-built logic and priced your recovery options. This checkpoint joins them. Same data date, same job, one page that reconciles both.

Your deliverable is a complete earned value report for the Willow Street Community Center at the month-6 data date. Six parts.

1. Cost-load the schedule. Take the CPM schedule you built in Chapter 14 and assign budget dollars to activities so they sum exactly to your BAC. Use the cost of the work from your Chapter 13 estimate — not the contract sum. On a $6,800,000 lump-sum project, expect a cost of work somewhere near $5,600,000–$5,800,000 after general conditions, insurance, bonds, contingency, and fee are stripped out; use your own number and state it. Do not load every activity — pick the 60 to 90 activities you would actually report on. Write your BAC at the top of the page.

2. State the data date and the three measures. Month 6 of the 425-calendar-day contract. Compute PV from your cost-loaded baseline, EV from the work actually in place at budget rates, and AC from your Chapter 28 cost report. For a 14-month job at month 6, a normal S-curve puts PV somewhere around 38 to 42 percent of BAC. Your EV and AC must be consistent with the story you told in Chapters 28 and 29 — if your schedule update said you were behind, your EV had better be below your PV.

3. Build the work-package table. Eight to twelve packages, each with BAC, PV, EV, AC, CV, SV, CPI, SPI, percent complete, and EAC, rolling up to a project row that foots exactly. Check the footing. State the EV measurement method for every package — units complete, weighted milestones, fixed formula, judgment, or level of effort — and put general conditions in its own row, outside the index.

4. Forecast, and say which formula you used and why. Compute all four EACs plus VAC and TCPI to BAC. Then pick one to report and write the one-sentence justification. If you choose AC + (BAC − EV), name the closed, non-recurring event. If you cannot name it, choose a different formula.

5. Plot the three-curve S-curve. Months 1 through 6, cumulative PV, EV, and AC, as a table and as an ASCII chart. Mark which of the four positions from 30.7 you are in.

6. Write the two-paragraph interpretation you would give the City. First paragraph: cost performance — what the CPI is, what it forecasts, what you are doing about the worst package. Second paragraph, and this is the one that matters: what earned value does not tell the City about your completion date. State plainly that SV and SPI are measured in dollars of work volume, that the completion date comes from the CPM, and give them the actual forecast substantial completion date and the total float on the controlling path. A municipal owner with $1,200 a day of liquidated damages deserves that sentence in writing, and writing it is also how you protect yourself from having said "we're at 1.02, we're fine."

Next: Chapter 31 puts three priced change orders into the notebook — and every one of them will move the BAC you just built, which is the maintenance problem 30.9 warned you about.

Chapter Summary

The reference card.

The decision framework — reading any EVM report in five moves.

1. Check the data date and the footing. No date, no report. Doesn't foot, no report.
2. Line up three percentages: time elapsed, EV ÷ BAC, AC ÷ BAC. On Northgate: 59.1, 62.4, 64.1. The gap between the last two is your whole cost problem.
3. Go to the package level and sort by cost variance. Aggregates hide everything. The project CPI of 0.974 concealed a curtain wall at 0.857.
4. Test the forecast with TCPI. If the required to-complete performance exceeds anything the job has demonstrated, the forecast is a wish. Ask what specific, physical thing changed.
5. Put the float report next to it. EVM tells you about money. Only the CPM tells you about time. If you leave a meeting having answered only one of those two questions, you did half the job.

Five things to carry out of this chapter.

• Earned value is measured at budget rates. If actual cost leaks into EV, your CPI is 1.000 forever and you have built an expensive way to learn nothing.
• SPI is not schedule. A job can post 1.02 and be thirty-four days late, and the mechanism — non-critical work completed early, a front-loaded SOV, and level-of-effort dilution — is completely ordinary.
• An early bad CPI is probably the number. Nothing about reaching the 40 percent mark improves your drawings, your subcontract prices, or your site.
• On subcontracted work, CPI measures your pay application review. Verify billed percentages physically or accept that you are reporting somebody else's opinion.
• The choice of EAC formula was worth $223,203 to Kestrel on one job in one month. That is why the assumption behind the formula is the professional judgment, and the arithmetic is just arithmetic.

What's Next

Every approved change order moves BAC, moves PV, and resets the baseline you just learned to measure against — and a change order priced without its impact costs are a forecast that was wrong the day it was signed. Chapter 31 takes changes apart: identifying them, pricing them completely, negotiating them, and documenting them well enough to survive. It is also where CO #14 finally gets dissected — the change that got built on a Thursday verbal and cost Kestrel $43,650 it could not substantiate.