Two Delay Analysis Methods, One Set of Facts, Opposite Answers
A worked construction example where two delay analysis methods invert which event caused the delay, plus what AACE, the SCL Protocol and CIOB require.
Every delay report picks an approach, usually in a methodology section near the front, often in a single sentence. The received wisdom is that the answer is settled: prospective for extensions of time, retrospective for damages, pick one and defend it.
What that framing hides is how much the choice can change the number. Not at the margins. This article works through a project where a prospective analysis and a retrospective one disagree about which event caused the delay at all, and neither is doing anything wrong. The choice sits inside a wider construction schedule analysis, and it is the one practitioners most often make by reflex.
What MIP stands for
MIP means Method Implementation Protocol. The term comes from AACE International Recommended Practice 29R-03, Forensic Schedule Analysis (2011), whose Section 3 sets out nine of them and describes their purpose as being to describe each forensic schedule analysis method in a consistent structure.
A MIP is not a standard you comply with. It is a catalogue entry: a numbered, neutral description of one way of doing the analysis. AACE avoids the industry’s marketing names in the numbering precisely because those names are used inconsistently, with the same phrase meaning different things to different experts.
The numbering matters when you cite it. Naming a MIP in a methodology section is a specific claim about what you did. If the number does not match the method, you have handed an opposing expert a free opening point before they reach any of your analysis.
Every method answers two questions
Strip away the names and each method is defined by two choices, which is how the SCL Protocol organises its own comparison at §11.5.
When is the critical path determined? From the baseline as planned (prospectively), from the updates as they stood at the time (contemporaneously), or from what actually happened (retrospectively).
When is the delay impact determined? By modelling forward from the moment of the event (prospectively), or by measuring what in fact occurred (retrospectively).
Two mechanical choices sit underneath, and AACE’s taxonomy turns on them:
- Additive or subtractive. Additive inserts the delay into a schedule and sees what moves. Subtractive removes it from a schedule and sees what collapses back. In AACE’s numbering these are different families: 3.6 and 3.7 are the additive protocols, 3.8 and 3.9 the subtractive ones. They are not variants of one method.
- Single base or multiple base. A “base” is the schedule you do the arithmetic on. Single base measures every event against one schedule. Multiple base gives each event its own, normally the update current immediately before it happened.
Base is not baseline. Base is the schedule you run the calculation on. Baseline is the plan of record, typically a contract term such as the Contract Programme, formally approved, with updates measured against it and re-baselining requiring approval. A method can use the baseline as its base. Most do not. Keeping the two words apart avoids most of the confusion in this area.
The same project, three different answers
This is an illustration. The project below is constructed for teaching. It is not taken from any judgment and not from any real project. Nothing in it should be cited as authority.
The baseline: 20 working days
Two paths. Path A runs foundations into steel and drives completion at day 20. Path B runs services design into services installation, finishing at day 14 with six days of total float.
What actually happened
Three things went wrong, and one went right:
- E1, ground conditions (employer risk). Foundations take 15 days instead of 10, discovered in the first week.
- E3, labour shortage (contractor risk). Services design takes 10 days instead of 6.
- E2, variation (employer risk), instructed at day 12. Services installation grows from 8 days to 14.
- Acceleration (contractor). Steel is resourced up and takes 6 days rather than 10.
Completion lands on day 24, four days late, and it is path B that finishes last, not the path that was critical in the baseline. That switch is what makes the choice of method bite.
The update at day 10, which is the base a prospective method uses
At the day 10 data date, foundations are still running and forecast to finish day 15. Services design finished four days late at day 10. Steel is still forecast at 10 days. Nobody has yet decided to accelerate, and the variation has not been instructed.
The services path now carries seven days of total float: the day 25 forecast completion less its day 18 finish.
Method 1: time impact analysis, prospectively
A true time impact analysis is additive and multiple base. Each event is inserted into the schedule that was current immediately before it, and you record how far the then-forecast completion date moves.
E1, ground conditions. The base is the baseline, before anything has happened. Insert five days on foundations and completion moves from day 20 to day 25. Impact: five days.
E2, the variation. The base is the day 10 update above, forecasting completion at day 25 with seven days of float on the services path. Insert six days on services installation. The variation consumes six of the seven available days of float and the completion date does not move. Impact: zero days.
Prospective total: five days, all of it E1.
Method 2: collapsed as-built, retrospectively
A collapsed as-built is subtractive. You take the as-built, remove the event, and see where completion collapses back to. The base is what actually happened, day 24.
But-for E2. Return services installation to eight days: path B finishes day 18, path A finishes day 21, and completion collapses to day 21. Impact: three days, being 24 less 21.
But-for E1. Return foundations to 10 days: steel runs days 11 to 16, path A finishes day 16. Path B still finishes day 24, so completion does not move at all. Impact: zero days.
Retrospective total: three days, all of it E2.
The reversal. Prospectively E1 is worth five days and E2 is worth nothing. Retrospectively E1 is worth nothing and E2 is worth three days. The two methods invert which event mattered, on identical facts, honestly applied.
Method 3: time slice windows, contemporaneously
A windows analysis does not model anything. It observes what the forecast completion date did in each period and asks what was driving the longest path at the time.
| Window | Forecast completion | Movement | What was driving it |
|---|---|---|---|
| Baseline to day 10 | day 20 to day 25 | +5 days | Foundations, E1 (employer risk) |
| Day 10 to completion | day 25 to day 24 | -1 day | Path switches to services. Acceleration on steel pulls four days back; the variation pushes services out six days and it becomes the driving path |
| Net | day 20 to day 24 | +4 days | Reconciles exactly to the outturn |
The windows analysis is the only one of the three that ties back to the four days actually lost, and the only one that surfaces the thing a tribunal would most want to know: the variation only ever mattered because the contractor’s own acceleration removed the cushion protecting it.
Side by side
| Method | Mechanics | E1 ground | E2 variation | Total |
|---|---|---|---|---|
| Time impact analysis (prospective) | Additive, multiple base | 5 days | 0 days | 5 days |
| Collapsed as-built (retrospective) | Subtractive, as-built base | 0 days | 3 days | 3 days |
| Time slice windows (contemporaneous) | Observational, per period | 5 days | see note | 4 days |
Actual overrun: four days. No modelled method reproduces it, and neither should be expected to, because isolating events from one another discards the interaction between them. In the windows analysis, E2’s contribution cannot be separated from the acceleration in the same period without a further attribution step.
What all three standards require before you call it a TIA
Three documents define the method, and on the points that matter they agree. Each is cited so you can check it in your own copy rather than take this on trust.
It must be additive. CIOB §5.8.36 describes calculating the effect of an intervening event by “adding the fragnet to the working schedule”, connecting it logically and recalculating the critical path. AACE lists time impact analysis among the common names for its additive protocols, MIP 3.6 (single base) and MIP 3.7 (multiple base), and for neither of the subtractive protocols.
The base must be the schedule current when the event began. CIOB §5.8.37 says the impact “is calculated by reference to its effect upon the working schedule at the time of its initiation”. AACE’s MIP 3.7 is a multiple-base method for that reason: each base model creates a period of analysis that confines the quantification within it.
It determines impact prospectively. CIOB §5.8.40 states it outright: “It is a prospective method, which predicts the likely effect of the event on the planned sequence at the time of initiation.” The SCL Protocol’s §11.5 comparison classifies time impact analysis as determining the critical path contemporaneously and the delay impact prospectively, which is what separates it from a time slice windows analysis, where the impact is determined retrospectively. One cell of difference, and it is the whole argument.
Events are impacted in sequence, not in isolation. CIOB §5.8.40 again: events are impacted “in chronological order of initiation date, and the critical path recalculated after each event”, so that each is assessed in the light of those before it.
The practical test. If your analysis removes a change from a later schedule and measures the collapse, it is a collapsed as-built, AACE’s subtractive family, whatever your column is headed. If it isolates each event from every other rather than impacting them in sequence, it is measuring each event’s isolated contribution, which is a useful screening figure and not a quantum.
What English courts have said about the choice
Quotations below are short and pinpointed, and each was checked verbatim against the judgment text at the paragraph cited.
The approaches need not agree. In Fluor v Shanghai Zhenhua Heavy Industry [2018] EWHC 1 (TCC) the court accepted expert evidence that a prospective analysis “does not necessarily produce the same answer as an analysis carried out retrospectively” (Fluor ¶275). The worked example above is one mechanism by which that happens.
The question selects the approach. Of the two, “The former is the correct approach when considering matters such as the award of an extension of time” (Fluor ¶275). But that was not the exercise before that court, so “I agree that some form of retrospective analysis is required” (Fluor ¶275). Read it with ¶280, where the court records that it did not need to decide which expert’s approach was correct.
Both experts thought the debate was beside the point. In Walter Lilly & Company Ltd v Mackay [2012] EWHC 1773 (TCC) the judge called the debate sterile because both delay experts accepted that “if each approach was done correctly, they should produce the same result” (Walter Lilly ¶380). The convergence is the experts’ common ground, not a finding that the methods always agree, and note the conditional it rests on. What the same paragraph identifies as the real problem is elsewhere: “An underlying problem, accepted as such by both experts, was the absence of a contemporaneous critical path programme” (Walter Lilly ¶380).
Contemporaneity is the substance. “One needs to consider what critically delayed the Works as they went along” (Walter Lilly ¶365). And in Saga Cruises BDF Ltd v Fincantieri SPA [2016] EWHC 1875 (Comm): “Causation in fact must be proved based on the situation at the time as regards delay” (Saga ¶251).
And a caution about leaning on the Protocol. Adyard Abu Dhabi v SD Marine Services [2011] EWHC 848 (Comm) records that the SCL Protocol was not in general use in contracts and had not been approved in any reported case, and so was of little assistance on the legal causation issues before that court (Adyard ¶289–¶290). Standards tell you what a method is. They do not tell you what a tribunal has done with it.
Incomplete by construction. Ten English judgments have been read for this work and four are cited here. This is not a survey of English authority on delay analysis. Where the judgments pull in different directions, and Walter Lilly ¶380 and Fluor ¶275 do, both are cited and the conflict is left standing rather than resolved into a trend.
What this leaves a delay analyst with
Name the exercise, not the standard. Say which question you are answering and why that dictates the approach. A methodology section that only cites standards is exposed, on the authority of Adyard ¶289–¶290.
Check the label against the mechanics. Additive or subtractive; one base or one per event; sequential or isolated. Those three answers determine what your method is called, regardless of what the column heading says.
Expect the methods to disagree, and say so first. The worked example inverts which event mattered. If your analysis would give a different answer under a different method, an opposing expert will find it. Better to have addressed it. This is the same fault line that the choice between prospective and retrospective analysis turns on in the English judgments.
The record decides more cases than the method. The determinative finding in Walter Lilly was the absence of a contemporaneous critical path programme (¶380), not a ruling on methodology.
About this article
AI-assisted, human-verified. Drafted with AI assistance. Every quotation was checked verbatim against the retained judgment text at the paragraph cited by an automated check, and every cited paragraph was then read in context by a person before publication.
Read the judgments. Quotations here are short and illustrative. The full judgments are public on Find Case Law:
- Walter Lilly & Company Ltd v Mackay & Anor [2012] EWHC 1773 (TCC): caselaw.nationalarchives.gov.uk/ewhc/tcc/2012/1773
- Fluor v Shanghai Zhenhua Heavy Industry Co Ltd [2018] EWHC 1 (TCC): caselaw.nationalarchives.gov.uk/ewhc/tcc/2018/1
- Adyard Abu Dhabi v SD Marine Services [2011] EWHC 848 (Comm): caselaw.nationalarchives.gov.uk/ewhc/comm/2011/848
- Saga Cruises BDF Ltd & Anor v Fincantieri SPA [2016] EWHC 1875 (Comm): caselaw.nationalarchives.gov.uk/ewhc/comm/2016/1875
Not legal advice. This is commentary on published judgments and industry standards, written for delay practitioners. It is not legal advice and must not be relied on as such. The standards are summarised and cited, not reproduced. Consult your own copies of AACE RP 29R-03, the SCL Delay and Disruption Protocol and the CIOB Guide to Good Practice for their full text.