Schedule compression
Last reviewed 9 September 20262,320 words11 min read
Crashing vs fast-tracking
🟩 In one line: Compression shortens the critical path either by overlapping work that was sequential (fast-tracking — cheap, adds rework risk) or by adding resource to shorten durations (crashing — costs money, hits diminishing returns fast); do it on the critical path only, soft logic first, and re-check the near-critical paths after every change because one of them is about to become critical.
👤 Who this is for: J/M. You should understand hard vs soft logic, near-critical paths and overtime arithmetic. Senior planners: the ⚖️ section on who pays is the part that matters.
First, let's be honest about why this page exists
Compression is asked for in a meeting and expected by the next morning. The project manager wants 20 days back; the planner's instinct is to open the live schedule and start shortening durations until the milestone turns green. That produces a programme that fits the date and cannot be built.
Real compression is a set of specific decisions — this overlap, this second crew, this shift pattern — each with a cost, a risk and a days-saved figure, tested in a Reflection and recorded so that when someone asks in six months why the façade cost 400,000 more than tendered, the answer is in the file.
🟨 The standard — what "good" looks like
| Source | What it says (paraphrased) |
|---|---|
| PMI Practice Standard for Scheduling, 3rd ed. / PMBOK Guide | Crashing shortens duration by adding resource at increased cost; fast-tracking overlaps phases or activities normally done in sequence, at increased risk of rework. Both apply to the critical path. |
| SCL Protocol, 2nd ed. — Core Principle 15 (mitigation) | The Contractor must take reasonable steps to mitigate delay, but is not obliged to add significant resource or incur substantial cost to do so unless instructed and paid. |
| SCL Protocol, 2nd ed. — Core Principle 16 (acceleration) | Acceleration should be instructed and priced before it starts; where the Contractor accelerates to avoid delay damages for a delay that was the Employer's, a constructive acceleration argument arises. |
| FIDIC 1999 Cl 8.6 / FIDIC 2017 Cl 8.7 (rate of progress) | If progress is too slow for reasons not attributable to the Employer, the Engineer may require revised methods to expedite, at the Contractor's cost. |
| NEC3 / NEC4 ECC Cl 36 | Acceleration is by quotation and acceptance; the Contractor is not obliged to accelerate without an accepted quotation. |
| DCMA 14-Point — Metrics 2 (leads) and 3 (lags) | Zero leads; lags on no more than 5% of relationships. Compression done with negative lags will fail the next review. |
| AACE International RP 25R-03, Estimating Lost Labor Productivity in Construction Claims | Extended overtime and crew stacking reduce productivity; the loss compounds with duration and intensity. |
🟢 Rule: compress the critical path only, soft logic before money, one change at a time in a Reflection, and re-run the near-critical filter after each.
How it actually works
1. Two levers, and which to pull first.
| Fast-tracking | Crashing | |
|---|---|---|
| What changes | Logic — sequential work overlaps | Durations — more crew, more hours, more shifts |
| Direct cost | Usually none | Always: premium labour, extra plant, supervision |
| Risk | Rework if the upstream work changes; congestion; quality | Productivity loss, safety exposure, supply chain strain |
| Where it works | Soft preference logic; repetitive work across zones or floors | Labour-driven activities with space to add crews |
| Where it fails | Hard logic — you cannot plaster a wall that isn't built | Space-limited, plant-limited, or already multi-shift work |
| Order | First | Second |
2. Fast-tracking properly. Find the soft preference logic on the critical path — the links tagged as such in the relationship Comments. Convert FS to SS with a lag that reflects the real minimum offset, or better, split the work by zone so that each zone has its own hard FS chain and zones overlap naturally. Do not use negative lags: a lead of −5 days is a lag you can't explain in six months and a DCMA failure now. If the overlap changes the method — plastering above while blockwork continues below — the method statement changes with it, and so does the safety plan.
3. Crashing properly. For each candidate activity, establish the cost per day saved and rank from cheapest. Crash in that order until the target is met or the next option is not worth it. Three limits apply and none of them is negotiable:
| Limit | Rule of thumb |
|---|---|
| Overtime | 50-hour weeks for more than three to four weeks lose 10–15% productivity; 60-hour weeks lose 20–30%. The second week of a six-day 12-hour pattern already yields less than the first. See overtime-and-shift-arithmetic. |
| Crew stacking | A second crew in the same space yields perhaps 60–70% of a first crew; a third often yields nothing. |
| Summer | Between mid-June and mid-September the midday break caps the working day; night shift is the only real lever on external work. |
4. Do it in a Reflection. Project window → right-click → Create Reflection. Make every change there — one at a time — and reschedule after each. When done, Tools → Schedule Comparison against the live schedule to list exactly what changed and what each change bought. Only then merge, or more usually, submit the Reflection as the recovery plan and leave the live schedule as the record of the current position.
5. Re-check the near-critical paths every time. Compressing the critical path by 20 days makes any path with less than 20 days float the new critical path. This is not a possibility; it is arithmetic. Run the near-critical filter after every change, and expect the achievable saving to be less than the sum of the individual options.
6. Record it. Each compression measure goes into a table: activity IDs, what changed, days saved, cost, risk, who approved it. This table is the backbone of the recovery narrative and — if the delay turns out to be the employer's — the acceleration claim.
📜 Who pays
| Situation | Contract position |
|---|---|
| Delay is the Contractor's | Contractor compresses at its own cost. FIDIC 1999 Cl 8.6 / 2017 Cl 8.7 lets the Engineer demand it. |
| Delay is the Employer's, acceleration instructed | Priced and agreed before starting. NEC Cl 36 quotation; under FIDIC via a variation under Cl 13. |
| Delay is the Employer's, EOT refused or delayed, Contractor compresses to avoid LDs | Constructive acceleration. Arguable, expensive to prove, needs a notice trail: EOT claimed, refused or ignored, acceleration notified as a consequence, costs recorded separately. See acceleration-and-constructive-acceleration. |
| Delay concurrent | Contractor mitigates but is not obliged to spend significantly (SCL Core Principle 15); anything beyond that should be instructed. |
The planner's part: keep the compression measures itemised with cost and dates so that each can be tied to the delay it responds to.
🟥 Where people go wrong
- Shortening durations on activities that aren't critical. It buys nothing and it looks like padding removal to the reviewer. Compress the longest path, then the next one that becomes critical.
- Ignoring the near-critical paths. Twenty days is recovered on paper; the second path was at 12 days float and is now 8 days negative. Net saving: 12. The narrative should have said 12 from the start.
- Using negative lags. A lead is a hidden overlap with no logic basis. Split the activity by zone or use SS with a positive lag, and write the basis in the Comments.
- Breaking hard logic. Overlapping curing with loading, or second-fix with wet trades, isn't compression — it's a defect scheduled in advance.
- Assuming overtime is linear. Twelve hours doesn't buy 50% more than eight. By week four it may buy 15%. Use the productivity factors, not the hour count.
- Compressing the live schedule. Every trial change overwrites the record of the current position. Reflection first, always.
- Treating it as free. Someone pays for every day recovered — in cash, in rework, or in risk. If the recovery plan has no cost column, it isn't a plan.
⚖️ When you're challenged
"Just take 20 days out of the finishes. They're always padded." Finishes on the critical path are 84 days; the durations came from the SBM at the crew sizes we have. I can get 12 days by starting plaster on Levels 1 to 3 while blockwork finishes above — that's a logic change and costs nothing. Another 6 needs a second plaster crew at around 180,000. Beyond that we're into overtime that stops paying back after three weeks.
"Why is the recovery only 19 days when the options add up to 23?" Because once the finishes path came in by 20, the façade path — which had 16 days float — became critical at 4 days over. The options table shows each measure; the net figure is what the network gives after all of them.
"Put a lead on the plaster start, it's quicker than re-sequencing." A negative lag says plaster starts five days before blockwork finishes without saying which walls. Splitting by level says exactly which walls. The second one survives a review and a claim; the first doesn't.
"We're accelerating because the Engineer hasn't answered the EOT. Do we need to say so?" Yes, in writing, before we spend anything — that the acceleration is a consequence of the unanswered claim and costs will be recorded separately. Without that letter it's our cost.
📄 Related pages
- Building a Recovery Plan — the document this arithmetic feeds.
- Near-Critical Paths — why the net saving is always less than the sum.
- Hard Logic vs Soft Logic — which links you're allowed to overlap.
- Overtime and Shift Arithmetic — why twelve hours doesn't buy 50%.
- Productivity Adjustment Factors — the loss factors for stacked crews and extended hours.
- P6 Reflections and What-If — testing without wrecking the live schedule.
- Acceleration and constructive acceleration — who pays when the delay wasn't yours.
- Revised Baseline vs Recovery Schedule — the compressed programme is a recovery plan, not a new baseline.
✏️ Worked example
A mid-rise residential block in Sharjah, 6-day calendar, Update 11 shows completion 24 working days late. Driver: blockwork and finishes on Levels 4 to 8. Second path (façade) has 16 days float; third path (lift installation) has 22.
Options tested, each in its own Reflection, against the live schedule
| # | Measure | Type | Logic or duration change | Days saved (alone) | Direct cost (AED) | Cost per day | Risk |
|---|---|---|---|---|---|---|---|
| 1 | Start plaster L1–L3 while blockwork continues L4–L8 | Fast-track | Plaster per level FS from blockwork that level instead of FS from blockwork all levels | 12 | 0 | 0 | Low — separate floors, separate access |
| 2 | Second plaster crew L4–L8 | Crash | Plaster durations from 8d to 5d per level | 6 | 180,000 | 30,000 | Low — space available |
| 3 | Façade double shift | Crash | Unitised panel install 24d to 16d | 8 | 420,000 | 52,500 | Medium — night lifting, summer |
| 4 | Ceiling framing SS with MEP second fix | Fast-track | FS to SS + 3d, zone-split | 5 | 0 | 0 | Medium — rework if MEP routes change |
Selection: 1, 2 and 4. Sum of individual savings: 23 days. Option 3 held in reserve — expensive and not needed on the finishes path.
After combining 1, 2 and 4 in one Reflection and rescheduling:
| Path | Float before | Float after | Comment |
|---|---|---|---|
| Finishes L4–L8 | −24 | −1 | Compressed by 23 |
| Façade | +16 | −5 | Now critical: the finishes path no longer masks it |
| Lift installation | +22 | +1 | Near-critical, watch |
Net recovery: 19 days, not 23. Completion forecast 5 working days late. To close the last 5, option 3 is the only lever on the façade path — 5 days at 52,500 per day is 262,500, or accept 5 days of delay damages at the contract rate and compare. That comparison, not the planner's opinion, is what went to the commercial manager.
Recorded: the options table above, the Schedule Comparison report for the combined Reflection, the revised method statement for plaster-over-blockwork, and a note that options 1 and 4 are soft-logic reversals of the baseline sequence, tagged in the relationship Comments as "compression — Update 11 recovery".
📖 References
- PMI, Practice Standard for Scheduling, 3rd ed.; PMBOK Guide — schedule compression techniques.
- Society of Construction Law, Delay and Disruption Protocol, 2nd ed. (2017) — Core Principles 15 and 16.
- FIDIC Conditions of Contract for Construction, 1999 — Cl 8.6, 13; 2017 — Cl 8.7, 13 (check the edition in your contract).
- NEC3 ECC / NEC4 ECC — Cl 36 (check the edition in your contract).
- DCMA 14-Point Schedule Assessment — Metrics 2 and 3.
- AACE International RP 25R-03, Estimating Lost Labor Productivity in Construction Claims.
- Oracle Primavera P6 Professional User Guide — Reflections; Schedule Comparison.
From the field
Experience from working planners. Unreviewed — read it as experience, not guidance.
Add what you know about schedule compression. 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