Earned Schedule
Last reviewed 9 September 20261,254 words6 min read
SPI in time units — so it stops lying to you after 70 %
🟩 In one line: Earned schedule asks "on what date was the baseline supposed to have earned what we have earned today?" — the gap between that date and today is the schedule variance in weeks, and unlike SPI it does not drift back to 1.0 as the job ends.
👤 Who this is for: Mid-level planners already reporting SPI; PMOs who want one time-based trend. Prerequisites: Earned-value-basics-spi-cpi, S-curves-early-late-actual.
First, let's be honest about why this page exists
Classic SPI has a built-in defect: PV stops at BAC when the baseline finishes, EV catches up eventually, and SPI returns to 1.00 on every project regardless of how late it is. Earned schedule (Lipke, 2003) fixes it by reading the S-curve horizontally instead of vertically. It is a ten-minute spreadsheet calculation from data you already have, and it produces a variance in weeks the PMO can trend.
It is still not a critical-path measure. It says how far behind the cumulative output plan you are, in time. The longest path still says the date.
🟨 The standard — what "good" looks like
| Source | What it says (paraphrased) | Use it for |
|---|---|---|
| Lipke, W., "Schedule is Different", The Measurable News, 2003 | Introduces ES: the point on the PV curve where cumulative PV equals current EV; SV(t) and SPI(t) in time units | Origin and definitions (Check the citations in a law report before relying on them.) |
| PMI Standard for Earned Value Management (2019) | Earned schedule recognised as an extension of EVM with time-based indicators (check against your copy) | Standing |
| AACE RP 82R-13 | EVM practices including time-based measures (check against your copy) | Cross-reference |
| Hub convention | ES computed in the workbook from the monthly PV table (Project baseline) and cumulative EV; reported beside SPI and longest-path variance; used from 30 % complete onward, preferred to SPI after 70 % | Your defaults |
🟢 Rule: ES is read from the baseline PV curve, in the same units the curve was built in (hours or cost), and it goes in the table next to the longest-path variance, not instead of it.
How it actually works
Definitions
| Term | Meaning |
|---|---|
| AT | Actual time — periods elapsed from baseline start to the data date (months or weeks) |
| ES | Earned schedule — the time at which cumulative PV on the baseline curve equalled today's cumulative EV |
| SV(t) | ES − AT (negative = behind, in periods) |
| SPI(t) | ES ÷ AT |
| PD | Planned duration of the baseline (periods) |
| IEAC(t) | Independent estimate at completion in time = PD ÷ SPI(t) (trend assumption) |
Finding ES. From the monthly cumulative PV table (Project baseline, exported from the Activity Usage Spreadsheet), find the last period n where PV(n) ≤ EV now. Then ES = n + (EV − PV(n)) ÷ (PV(n+1) − PV(n)). Linear interpolation inside the period is standard.
Workbook layout. One sheet: month number, DD, cumulative PV, cumulative EV, AT, ES, SV(t), SPI(t), IEAC(t), longest-path variance, SPI. One row per DD, never reset; baseline revision = flagged row, PV column re-pulled from the new revision from that row on.
Reading it
| SPI(t) | Meaning |
|---|---|
| ≥ 0.95 | Output pace within tolerance |
| 0.90–0.95 | Behind; check whether the shortfall is critical |
| < 0.90 | Sustained underperformance; IEAC(t) worth stating |
IEAC(t) assumes the past pace continues. It usually over-estimates the delay early (mobilisation drag) and under-estimates it late (T&C gating). It is a trend, reported as one.
Relation to the schedule. SV(t) and the longest-path variance should move together. When SV(t) is −6 weeks and the longest path is −2 weeks, non-critical work is behind — good news for the date, bad news for the peak later. When the longest path is −8 weeks and SV(t) is −2, a small critical chain is driving the delay and total output is fine — the recovery lever is sequence, not manpower.
🟥 Where people go wrong
- PV curve from the current schedule. ES then equals AT every month. PV is the Project baseline curve, exported once per accepted revision and stored.
- Mixing units. EV in cost read against a PV curve in hours. Same units, same curve.
- IEAC(t) presented as the forecast completion. It is a pace extrapolation; the forecast date is the longest path's. Put IEAC(t) in the trend table with the word "trend".
- Using ES before there is a curve to read. Under about 30 % complete, the interpolation sits on the flat start of the S-curve and swings wildly. Start reporting at 30 %.
- Re-basing the series at each recovery plan. Recovery schedules are not baselines; PV stays on the accepted revision until the register says otherwise.
⚖️ When you're challenged
"Why do we need this if we already have SPI?" SPI is 0.97 this month and was 0.91 six months ago — it's improving because the baseline is running out of planned work, not because we've caught up. Earned schedule says we're still 7.4 weeks behind the output plan, which agrees with the longest path at 6 weeks. That's the number to trend.
"IEAC(t) says 41 months. The programme says 37. Which is right?" The programme. IEAC(t) assumes the last 15 months' average pace continues to the end; the recovery plan changes the pace on the driving chain, and the longest path shows the effect of that. If the recovery doesn't happen, 41 is where the trend points.
📄 Related pages
- Earned Value Basics: SPI and CPI — the inputs and SPI's late-project drift
- S-Curves: Early, Late and Actual — the curve ES is read from
- Slippage and Trend Analysis — the trend register ES joins
- CPLI and BEI — the schedule-side ratios reported alongside
- Forecasting honestly — why IEAC(t) is a trend, not the forecast
✏️ Worked example — Sharjah mid-rise, DD month 18 of 24
Baseline BL0, labour BAC 640,000 mh. Cumulative PV table (kmh): month 12 = 358; month 13 = 392; month 14 = 428; month 15 = 462; month 16 = 494; month 17 = 524; month 18 = 552.
EV at DD (month 18) = 441 kmh. PV(18) = 552 → SPI = 0.80.
ES: PV(14) = 428 ≤ 441 < PV(15) = 462 → ES = 14 + (441 − 428) ÷ (462 − 428) = 14 + 0.38 = 14.38 months.
| Measure | Value |
|---|---|
| AT | 18.0 |
| ES | 14.38 |
| SV(t) | −3.6 months |
| SPI(t) | 0.80 |
| IEAC(t) | 24 ÷ 0.80 = 30.0 months (trend) |
| Longest-path variance | −68 working days ≈ −2.9 months |
| Forecast completion (longest path) | Month 27.1 |
Read: Output pace and the longest path agree within a month; the recovery plan on the finishes chain targets month 25.5, and next month's SV(t) is the test of whether the pace actually changed. Series row added to the trend register.
📖 References
- PMI, The Standard for Earned Value Management (2019) (check against your copy)
- AACE International RP 82R-13, Earned Value Management Overview (check against your copy)
- Oracle Primavera P6 Professional User Guide — Activity Usage Spreadsheet (check against your P6 version)
From the field
Experience from working planners. Unreviewed — read it as experience, not guidance.
Add what you know about earned schedule. What worked, what the consultant pushed back on, what you would do differently next time. A paragraph is plenty.
Contributors get their name and one link on the site — your own templates, course or consultancy. We take nothing and hold nothing.
Add your experience