Two bars, and what they leave out.

Two bars, and what they leave out.

Draw the planned programme. Underneath it, draw what was actually built. Measure the gap.

That is the oldest comparison in the subject and still one of the most used, partly because anybody can follow it — a commercial director, a judge, a site manager — without a word of explanation.

It is also the point at which this phase turns around. Everything up to now started from causes: identify the events, model them, calculate the effect. This method starts from the effect. Here is what happened; now let us work out what was responsible for it.

Two things share the name

At project level the comparison is between the planned completion date and the actual one, and the whole difference is presented as the claim. This is the total time version, and it is the one that gets served most often.

At activity level the comparison is done bar by bar: this activity was planned for eleven weeks and took fifteen, that one was planned for six and took six. The output isn't a single number but a pattern of variances, and the work is then to explain which of them mattered.

Both get called as-planned versus as-built. They are not remotely the same document, and only one of them tends to survive a careful reading.

The bar the total-time version has to clear

A contractor claiming the whole overrun as excusable and compensable, on the strength of a project-level comparison, is taking on considerably more than the arithmetic suggests.

Read the second proposition again, because it is the one that ends most of these claims. Not the contractor's delays were minor. Not they were outweighed. Nothing at all — no contribution to critical delay anywhere in the job.

On a project long enough to produce this kind of dispute, that proposition is almost never true, and both sides usually know it before the report is opened.

How the defence is actually run

Respondents faced with this analysis rarely bother with the calculation. There isn't much of one to attack.

They go to the list instead, and to the two easiest items on it. First, find culpable events the claimant's analysis left out — a subcontractor who didn't turn up, a rig that broke down, a section rebuilt after a failed inspection. Second, find the places where the programme logic was changed during the works, which undermines using the original plan as the yardstick for anything.

Neither of those is difficult on a real project. That is the honest position: the defence is cheap, and it is cheap by design.

What it asks for, and what it lets you off

Now the reason this method has not been retired, because the trade it offers is genuinely unusual.

It doesn't need a monthly update series — the whole apparatus Week 12 depends on. It doesn't need the as-planned programme to contain logical relationships or float values, which means it can be run against a baseline too crude for any modelled method. And nothing is inserted or subtracted anywhere, so none of the objections aimed at hypothetical models apply to it at all.

What it demands instead is a defensible as-built, an as-built sequence that can be mapped onto the planned one, and enough factual record to explain the variances. It also demands that the analyst deduce the as-built critical path by hand, without monthly updates to lean on — which is Week 7's inference problem arriving without the safety net.

So it forgives a weak baseline and a broken update series, and it does not forgive a weak site record. That is the opposite of every method in the first half of this phase, and it is why the ordering of these weeks is worth remembering.

Variance is not cause

Before any of that, the limitation the picture hides best.

The comparison measures a gap. It says nothing whatsoever about why the gap is there, and because the diagram is so legible people read causation straight off it.

Three activities each running twenty days over look identical on the page. One of them is the rock — an employer risk event, and a claim. One is rebar the contractor ordered late, which nobody pays for. And one is an activity that simply took longer than the estimator allowed, which is not an event at all and belongs to whoever priced the tender. Risk Week 5 is worth remembering here: the $48,450 was always conditional, and a bar chart cannot tell you whether the condition held.

Everything that decides the money sits outside the diagram, in the records from Week 6. The comparison is a very good instrument for deciding where to look. It is not evidence of what you will find when you get there, and a report that puts the two bars on one page and a number on the next has skipped the argument entirely.

This is Week 2's point arriving in its most tempting form. An agreed picture of what happened and an agreed picture of what was planned still do not, between them, contain a single sentence about causation.

The version that works

Done properly, this can be among the most convincing analyses available, and it gets there without modelling anything.

Properly means at activity level. It means each significant variance identified and then explained from contemporaneous records rather than asserted. It means the contractor's own delays appearing in the analysis, priced and set against the rest, because a report that shows them is far harder to dismiss than one that doesn't.

And it means the same two virtues Week 12 identified: concurrency visible in the period the work was actually done, and critical delay landing in the months the costs were actually incurred. This method has both, without needing a single update.

Cutting it into periods

There is a windowed version, which is where most competent practitioners end up.

The comparison is done as before, but the variances are summarised at the end of each period rather than across the whole job. The boundaries are usually set at key milestones, at points where the contractor's intentions or method changed — an attempted acceleration, a change in conditions — rather than at arbitrary monthly intervals.

The naming is a small trap worth flagging. The American guidance files this under a technical label built from its characteristics, while the British protocol calls it plainly what it is. Two names, one technique, and a certain amount of unnecessary argument between people who have each read only one of the documents.

Practical insight

Take your job and produce the crudest possible version of this in an hour: planned dates against actual dates for the twenty most significant activities, in one table, with the variance in days beside each.

Then sort the table by variance, largest first, and look at the top five. Those five are your delay claim, whatever else the narrative says.

Now write one sentence against each explaining why it ran late, and mark the sentence evidenced or believed. The honest split is usually uncomfortable and always useful. The evidenced ones are your case. The believed ones are either work to do while the job is still running, or ground you'll concede later without meaning to.

An hour of that will tell you more about the strength of your position than a month of anything else in this phase.

Key takeaways

✔ This is the first effect-based method: it starts from what happened rather than from a list of causes.

✔ Project-level and activity-level comparisons share a name and are not the same exercise.

✔ The total-time version requires proving the contractor contributed nothing at all to critical delay, which is rarely true.

✔ The standard defence ignores the arithmetic and looks for omitted culpable events and mid-project logic changes.

✔ It needs no update series, no baseline logic and no float, so it works where modelled methods cannot.

✔ It requires a defensible as-built and an as-built critical path deduced by hand, so a weak site record kills it.

✔ Done at activity level, with the contractor's own delays shown, it is one of the most convincing analyses available.

What's coming next

The last method in this phase runs the comparison in reverse. Instead of adding events to a plan or measuring the gap between two programmes, it starts from the job as built and removes the delays to see what would have been left — the most intuitive idea in the subject and the one with the most judgement hidden inside it.