Forecasting honestly
Last reviewed 9 September 20262,482 words11 min read
Remaining duration, trend
🟩 In one line: The forecast completion date is only as good as the remaining durations behind it, and a remaining duration is only honest if it comes from remaining quantity divided by the rate actually being achieved — not from the original duration minus time elapsed, and not from a promise to go faster next month.
👤 Who this is for: M/S, and every junior who has ever been told "just leave the remaining duration as it is". Read quantity-tracking-and-physical-progress and slippage-and-trend-analysis alongside.
First, let's be honest about why this page exists
Most schedules forecast the same completion date for eight months and then lose three months in one update. That isn't the site suddenly collapsing; it's eight months of remaining durations that were never re-forecast, held up by the assumption that the crew would catch up. The catch-up never came, and when the last activities couldn't be squeezed any further the truth arrived all at once.
An honest forecast moves a little every month. It is usually bad news in small doses, which is uncomfortable, and it is the only forecast anyone can act on. The planner who reports the same date until the cliff has not protected the project; they have removed its chance to recover.
🟨 The standard — what "good" looks like
| Source | What it says (paraphrased) |
|---|---|
| FIDIC 1999 Cl 8.3 / 2017 Cl 8.3 | Revised programme whenever the current one is inconsistent with actual progress; the programme must show the Contractor's genuine intended sequence and timing for remaining work. |
| FIDIC 1999 Cl 4.21 / 2017 Cl 4.20 | Monthly report to include comparison of actual and planned progress and any events which may jeopardise completion, with measures being adopted. |
| NEC3 / NEC4 ECC Cl 32.1 | Each revised programme shows actual progress and its effect on the timing of remaining work — the effect, not the original plan re-dated. |
| SCL Protocol, 2nd ed. — Core Principle 1 and Part B | Updated programme should be a realistic forecast; an update that does not reflect actual productivity is not a reliable record for later analysis. |
| AACE International RP 53R-06, Schedule Update Review | Remaining durations should be reviewed for consistency with achieved progress; unchanged remaining durations on activities behind plan are a red flag. |
| AACE International RP 55R-09, Analyzing S-Curves | Forecast at completion should be derived from trend, not plan; a late-stage vertical S-curve is a symptom of un-forecast remaining work. |
| GAO Schedule Assessment Guide — Best Practice 9 | Updates must reflect realistic remaining durations based on progress to date; "planned finish" is not "forecast finish". |
| PMI Practice Standard for Scheduling, 3rd ed. | Remaining duration is an independent estimate at each status; it is not derived from percent complete. |
| DCMA 14-Point — Metric 9 (Invalid Dates); Metric 12 (Critical Path Test) | Forecast dates must be after the data date and the completion must respond to remaining-duration changes. |
🟢 Rule: every in-progress activity gets a fresh remaining duration at every update, derived from remaining work ÷ achieved rate, adjusted only for changes in resource that are confirmed in writing — and the forecast completion is whatever the network then says.
How it actually works
1. Three numbers that get confused.
| Number | What it is | Who sets it |
|---|---|---|
| Original duration | Baseline estimate from the SBM | Fixed at baseline |
| Remaining duration | Fresh estimate of working time to finish, from the data date | Planner, from site data, every update |
| Percent complete | Physical work done ÷ total, from the rule of credit | Quantity sheet |
P6 will happily let you leave remaining duration = original − elapsed, and if Percent Complete Type is Duration it does exactly that automatically. Neither number has anything to do with how fast the work is going.
2. The formula.
| Step | Calculation | Example |
|---|---|---|
| Remaining quantity | Total − installed | 12,000 − 7,500 = 4,500 m² |
| Achieved rate | Installed over last 2–3 periods ÷ working days in those periods | 3,600 m² ÷ 48 days = 75 m²/day |
| Remaining duration | Remaining quantity ÷ achieved rate, round up to whole shifts | 4,500 ÷ 75 = 60 days |
| Baseline said | Original duration − elapsed | 96 − 48 = 48 days |
| Variance | Forecast − baseline remaining | +12 days |
Use the last two or three periods, not the whole history — the first month includes mobilisation and the rate has usually settled. Use the SBM rate only if the achieved rate isn't yet measurable (activity just started).
3. When to depart from the achieved rate. Only when something has changed and you can name it: a second crew mobilised (delivery note, gate pass, foreman's name), a shift added (approved overtime instruction), a constraint removed (drawing approved, area released). Apply the productivity factors from productivity-adjustment-factors to the added resource — a second crew does not double the rate. Write the assumption into the Notebook topic "Forecast Basis" on the activity. If the change is promised but not evidenced, the forecast uses the achieved rate and the narrative records the promise.
4. Activities not yet started. Original duration stands unless a comparable activity has already shown a different rate. If plaster on Levels 3–5 ran 20% slower than the SBM, plaster on Levels 6–20 should be re-forecast at the achieved rate — a change table entry, a reason, and probably a recovery plan. Holding L6–L20 at the SBM rate while L3–L5 proved it wrong is the commonest way the cliff is built.
5. Expected Finish. P6's Expected Finish field (activity Status tab) sets a date and back-calculates remaining duration on scheduling, if "Use Expected Finish Dates" is ticked in the options. Useful when the site gives a date rather than a duration — "pour booked 28 May" — and only for that. It is a forecast input, not a constraint, and it should be cleared when the activity finishes.
6. The forecast at project level. Three cross-checks each month before the narrative goes out:
| Check | Method | If it fails |
|---|---|---|
| Network forecast | Reschedule; read completion milestone | This is the forecast; the others test it |
| Trend | Plot forecast completion for the last six updates; extrapolate | Trend points later than the network → remaining durations are still optimistic |
| Rate-based | Total remaining manhours (or quantities) ÷ achieved monthly output → months remaining | Longer than network → same conclusion |
If all three agree within a couple of weeks the forecast is honest. If the network is earliest of the three by a month, go back to the remaining durations.
7. Reporting it. The narrative gives forecast completion, variance against contract and against last update, and the reason for this month's movement in one sentence. Movement of zero for the third month running on a project with falling productivity is itself a finding. Report the forecast basis table — achieved rates by trade against SBM rates — in an appendix.
📜 Forecasts and the contract
An honest forecast that shows delay is not an admission of culpability. Under FIDIC the revised programme (Cl 8.3) and the progress report (Cl 4.21 / 4.20) are obligations; a programme that hides delay breaches them and, worse, destroys the record you need for the EOT. Under NEC a revised programme that does not show the real effect on remaining work is grounds for the Project Manager not to accept it (Cl 31.3), and an un-accepted programme weakens every compensation event assessment that follows. Delay analysts on both sides prefer a contractor whose updates showed the slip developing, month by month, with causes recorded, to one whose updates were green until the last quarter. The forecast is evidence; keep it truthful and keep the causes next to it.
🟥 Where people go wrong
- Remaining duration = original − elapsed. The default, and the lie. It assumes the plan rate regardless of what's happened.
- Deriving remaining duration from percent. 60% done in 12 days therefore 8 days left. Only true if the rate has been constant and stays constant. Check it against quantity.
- Forecasting the catch-up. "We'll double the crew next month." Not until the crew is on site with names. The achieved rate stands.
- Re-forecasting only the activities in progress. The 14 unstarted floors will run at the rate the first three floors proved. Change them, record it, and start the recovery plan now.
- Holding the date because the PM asked. The date is an output of the network. If the PM wants it earlier, the recovery plan changes the inputs — logic, resource, shifts — and records what each costs. See planner-ethics-and-pressure.
- Duration % Complete on production work. The percent advances with the calendar and the remaining duration shrinks with it. Physical or Units.
- No forecast basis recorded. Six months later nobody knows whether the 60 days was measured or invented. Notebook topic "Forecast Basis" on every re-forecast activity; achieved-rate table in the narrative appendix.
⚖️ When you're challenged
"Why has the forecast moved out two weeks? Nothing happened this month." Nothing happened is the point. Blockwork has run at 78 m²/day for three periods against 97 in the SBM. Last month I carried the SBM rate for the unstarted levels on the basis that a second crew was coming; it hasn't, so this month the remaining fourteen levels are forecast at the achieved rate. That's the two weeks. The forecast basis table is in Appendix C.
"Leave the remaining duration alone — the subcontractor says they'll catch up." If they mobilise the second crew, I'll re-forecast the day the gate passes are issued and the date will come back in. Until then the schedule shows what the achieved rate gives. If I forecast the catch-up now and it doesn't happen, we'll have told the Engineer a date we knew was unsupported.
"The forecast has said 14 March for seven updates. Is that credible?" No, and it should have been questioned earlier. The S-curve has been running 4–6% behind plan since Update 04 while the completion date didn't move, which means the remaining durations weren't being re-forecast. This update they have been; the forecast is now 2 May, and the change table shows every activity revised and why.
"Can we show two dates — the forecast and the target?" Yes, and we should. The forecast is what the network gives on current rates. The target is what the recovery plan gives if its measures are implemented, with the measures and costs listed. Two dates with a clear basis each are honest; one date that's secretly the target is not.
📄 Related pages
- Quantity tracking and physical progress — the remaining quantity and achieved rate.
- Slippage and Trend Analysis — the trend cross-check across six updates.
- Productivity Adjustment Factors — what a second crew or a shift really adds.
- Statusing a schedule — entering remaining durations.
- Percent Complete Types — why Duration % is wrong for production work.
- Building a Recovery Plan — the target date and what it costs.
- Planner ethics and pressure — when you're asked to hold the date.
- Schedule Narrative — reporting the forecast and its basis.
✏️ Worked example
A residential tower in Sharjah, 26 floors, blockwork one activity per floor, 1,860 m² per floor, SBM rate 97 m²/day with one crew (≈19 working days per floor). Update 07, data date 25 May 2025. Floors 1–5 complete, floor 6 in progress, floors 7–26 not started.
Achieved rate, Updates 05–07: 5,580 m² installed over 72 working days = 77.5 m²/day, 20% below SBM.
Floor 6 remaining duration:
| Item | Value |
|---|---|
| Total | 1,860 m² |
| Installed | 1,120 m² (60%) |
| Remaining | 740 m² |
| At achieved rate | 740 ÷ 77.5 = 9.5 → 10 days |
| P6 default (original − elapsed) | 19 − 14 = 5 days |
Floors 7–26, not started: original duration 19 days each. Re-forecast at achieved rate: 1,860 ÷ 77.5 = 24 days each. Twenty floors × 5 extra days on a partially overlapped chain: the blockwork path lengthened by 62 working days after rescheduling; the finishes chain following it moved with it. Forecast completion moved from 14 March 2026 (unchanged since Update 02) to 8 June 2026.
Three-way check:
| Method | Completion |
|---|---|
| Network (after re-forecast) | 8 Jun 26 |
| Trend of S-curve slip (4–6% behind, steady) extrapolated | ≈ late May 26 |
| Remaining manhours ÷ achieved monthly manhours | 11.4 months from DD ≈ mid-May 26 |
Agreement within three weeks; the network forecast was accepted as the honest date.
What went in the narrative: forecast 8 June 2026, variance +84 working days against contract, +58 against Update 06; driver blockwork productivity at 80% of SBM; 21 activities re-forecast, listed in the change table with the rate basis; recovery plan issued separately proposing a second blockwork crew from floor 9 (target 18 April 2026, cost tabled, not yet confirmed). Notebook "Forecast Basis" on each blockwork activity: "Re-forecast Upd 07 at achieved 77.5 m²/day (Upd 05–07 avg); SBM 97."
The second crew mobilised three weeks later. Update 08 re-forecast floors 10–26 at a combined rate of 135 m²/day (second crew at 75% effectiveness, congestion) and the forecast came back to 3 May 2026 — a date with a crew behind it.
📖 References
- FIDIC Conditions of Contract for Construction, 1999 — Cl 4.21, 8.3; 2017 — Cl 4.20, 8.3 (check the edition in your contract).
- NEC3 ECC / NEC4 ECC — Cl 31.3, 32.1 (check the edition in your contract).
- Society of Construction Law, Delay and Disruption Protocol, 2nd ed. (2017) — Core Principle 1; Part B.
- AACE International RP 53R-06, Schedule Update Review — As Applied in Engineering, Procurement, and Construction.
- AACE International RP 55R-09, Analyzing S-Curves.
- GAO-16-89G, Schedule Assessment Guide — Best Practice 9.
- PMI, Practice Standard for Scheduling, 3rd ed.
- DCMA 14-Point Schedule Assessment — Metrics 9 and 12.
- Oracle Primavera P6 Professional User Guide — Activity Status tab (Remaining Duration, Expected Finish); Schedule Options; Notebook topics.
From the field
Experience from working planners. Unreviewed — read it as experience, not guidance.
Add what you know about forecasting honestly. 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