Chapter 30 — Key Takeaways

A one-page reference card. Keep it next to your monthly cost report and your float report — you need all three on the desk at the same time.

The whole instrument on one card

Term Formula Reads as Watch out for
PV Budgeted cost of work scheduled What you planned to have built Needs a cost-loaded schedule; no schedule, no PV
EV Budgeted cost of work performed What you built, at budget rates Never at actual cost
AC Actual cost of work performed What it cost Must include accruals
BAC Budget at completion The budget for whatever you are measuring Moves only on executed change orders
CV EV − AC Dollars over or under Cannot be recovered, only stopped
SV EV − PV Dollars of work ahead or behind Dollars, not days. Goes to zero at completion
CPI EV ÷ AC Dollars of work per dollar spent Stabilizes early; believe it
SPI EV ÷ PV Work volume against plan Goes to 1.000 at completion; is not schedule
EAC #1 BAC ÷ CPI Current performance continues The default and usually the honest one
EAC #2 AC + (BAC − EV) One-time event, rest to plan Can never forecast an unrealized problem
EAC #3 AC + [(BAC − EV) ÷ (CPI × SPI)] Both pressures continue Right when you will have to compress
EAC #4 AC + bottom-up ETC Re-estimated remainder Most work, most accurate; do it quarterly
ETC EAC − AC Cost of the rest Reconcile against the Chapter 28 cost-to-complete
VAC BAC − EAC Forecast overrun Compare it to contingency, not to zero
TCPI (BAC − EV) ÷ (target − AC) Efficiency the remainder must achieve The credibility test on any forecast

Key Takeaways

  • Earned value is measured at budget rates, not at what the work cost you. If actual cost leaks into EV, your CPI is exactly 1.000 forever and you have built an expensive way to learn nothing.
  • "Under budget" on a cost report is PV − AC and it has a hole in it the size of the building. Northgate looked $118,000 under budget at month 11 while it was actually $672,000 over cost and $790,000 of work behind. Being behind looks exactly like being under budget until the month you catch up.
  • SV and SPI are denominated in dollars of work volume, not days. They converge to zero and to 1.000 at completion on every project, including one that finishes a year late, because EV and PV both must end at BAC. That is a property of the metric, not a flaw in your implementation.
  • A job can post an SPI of 1.02 while its critical path is thirty-four days late — non-critical work completed early, a front-loaded schedule of values, and level-of-effort dilution, all three perfectly ordinary. There is no arithmetic on PV, EV, and AC that produces the number 34.
  • Read three percentages in this order: time elapsed, EV ÷ BAC, AC ÷ BAC. On Northgate at month 11: 59.1, 62.4, 64.1. The gap between the second and third is your entire cost problem.
  • An early bad CPI is probably the number. Nothing about reaching the 40 percent mark makes your drawings better, your subcontracts cheaper, or your site easier. Treat the CPI-stabilizes-early observation as a widely reported practitioner finding rather than a law — and act on it anyway, because the mechanism is plausible and the failure mode is expensive.
  • Trend beats level. A CPI of 0.96 that has been 0.96 for six months is a priced, known condition. A CPI of 0.99 that was 1.03 two months ago is an emergency in progress.
  • Small denominators lie. Below roughly 20 percent complete or $250,000 of EV, look at the package and do not forecast from it. Northgate's specialties package at 1.9 percent complete contributed $323,478 of pure extrapolation noise.
  • The choice of EAC formula is professional judgment; the arithmetic is just arithmetic. On one job in one month it was worth $223,203 to the contractor through a 75/25 contingency share. Which is why the person whose performance depends on the answer should not pick the formula.
  • TCPI turns a forecast into a testable claim. 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.
  • On subcontracted work, CPI measures your pay application review, not the sub's productivity — EV and AC both come from their pay application. The SPI on a subcontract package is still real, because PV comes from your schedule.
  • Earned value is a cost-performance instrument. Use it alongside the CPM, never instead of it. Report cost performance from earned value. Report schedule from the critical path.

Action Items — do these on your job this week

  1. Compute your three percentages. Time elapsed, EV ÷ BAC, AC ÷ BAC. If you cannot produce EV, that is this week's real finding, and the cause is almost always that the budget lives in cost codes and never touched a schedule activity.
  2. Foot your report. BAC, PV, EV, AC must each sum from the packages to the project row. CV and SV must reconcile. A report that does not foot is a report nobody believes.
  3. Sort by cost variance at the package level and read the top three. Aggregates hide everything: a project CPI of 0.974 concealed a curtain wall at 0.857.
  4. Run TCPI against the EAC you are about to report. Compare it to the best month this job has ever posted — period and cumulative. If you cannot name the physical thing that changed, change the forecast, not the sentence.
  5. Pick one subcontract package and physically verify the billed percentage. Send a field engineer with a tape and a count sheet. On Northgate that walk found $275,400 of overstated earned value in a day and a half.
  6. Write down the EV measurement method for every package, today, in a controlled document — units complete, weighted milestones, fixed formula, judgment, or level of effort. A method chosen after you see the numbers is not a measurement method; it is a negotiating position.
  7. Move level of effort into its own row, outside every index you forecast from.
  8. Print the float report and put it next to the EVM report. If you leave a meeting having answered only the money question, you did half the job.

Common Mistakes and the Fix

Mistake What it costs The fix
Reporting "under budget" from PV − AC You believe you have money you have not earned; the truth arrives in the month you catch up Compute EV. Report CV and SV separately and never add them
Actual cost inside EV CPI is exactly 1.000 forever; the report measures nothing EV = installed quantity × budget unit rate, always
Not accruing uninvoiced subcontractor work AC understated, CPI inflated — the most common way a report lies by accident Accrue work performed before the data date, with a source, a basis, and a reversal period
Quoting SPI as a schedule position An owner is told 1.02 while the building will not be dry in March Report schedule from the CPM; if you convert SV to days, label the approximation loudly
Extrapolating a package at 2 percent complete $323,478 of forecast noise in a single line | No forecast below ~20% complete or $250,000 of EV; carry at budget or bottom-up
Defaulting to AC + (BAC − EV) A forecast that structurally cannot show a problem you have not already had Use #1 as baseline, #4 as the check, #2 only with a written, named, closed event
Accepting a subcontractor's billed percentage as EV You report their optimism as your measurement, and you pay for it less retention Physically verify quantities; negotiate units-complete SOVs at buyout
Folding general conditions into the project index SPI and CPI both move for reasons that never happened in the field LOE in its own bucket, reported separately
Letting BAC drift out of date Indices become meaningless within about four months Executed changes only, each with a log entry, a date, an amount, and an authorizing document
Front-loading the schedule of values Borrowing from the owner without asking — and losing the ability to see your own project Price the SOV to cost plus fair markup; disclose genuine mobilization as mobilization
A report with no data date It is not a report. It is a rumor with a table in it Fix the data date, fix the issue date, publish on the date whether or not it is pretty

Decision Framework — reading any EVM report in five moves

  1. Check the data date and the footing. No date, no report. Does not foot, no report.
  2. Line up the three percentages — time elapsed, EV ÷ BAC, AC ÷ BAC. The gap between the last two is the cost problem, and you can state its dollar size before you read another line.
  3. Go to the package level and sort by cost variance. Then sort each package into one of three bins, because they are three different management problems: a discovery (real, caused, probably permanent — contain it), an opportunity (early, small, fixable — spend your attention here), or a measurement failure (the number was never a measurement — go count something).
  4. Test the forecast with TCPI. Required to-complete performance against demonstrated performance. If the gap is large, ask what specific, physical thing changed. "We'll get it back" is not a plan; it is a mood.
  5. Put the float report next to it. EVM tells you about money. Only the CPM tells you about time.

When somebody hands you a forecast

Ask four questions, in this order, and do not skip to the fourth:

Which formula? → What does it assume about the remaining work? → What TCPI does it require? → When did this job last perform at that level?

If the answer to the fourth is "never," you have your answer, and the conversation you now need is about what physically changes — a detail revised by RFI, a crew added whose production rate is known, a one-time learning curve that will not repeat and that somebody has quantified. Not effort. Effort is not a change.