Schedule update analysis: the update moved six days, the calculation forty two

Scale

The four fixed panels share one scale, forty days either side of the previous forecast, and it does not move. The live chart above rescales with the dates you type.

Basis

Every figure is a whole number of calendar days between two dates, measured from the previous forecast completion.

Assumptions

Three dates and nothing else; the progress only date recalculated to the later data date; the two issued programmes taken as given; no calendar, holiday or float behind any figure.

Model version

1.0

The formula, in words

• Three dates, two subtractions, and one addition that has to close Take the previous forecast completion from the progress only date and you have the progress movement. Take the progress only date from the submitted date and you have the revision movement. Add them and you must get the distance between the two issued programmes, exactly, every time. That closure is the only check the page can run on the dates you gave it, and it is why the readout shows all three.
• The middle date is not on either programme, and that is the point of it It is the previous accepted programme with the later update's recorded progress imported into it and nothing else imported, recalculated to the later data date. No logic changed, no duration changed, no constraint changed. It is the answer to one question: where would the completion date have gone if the only thing that happened was the work that was recorded. Without it an update carries one number, the distance between two programmes, and one number cannot be split into two.
• The share divides by sizes, and it is exactly half when the update reports nothing Add the two movements without their direction and divide the revision movement into that total. The result is always between nought and one hundred per cent, which a share of the net is not: point two movements against each other and the net can be smaller than either, or zero, and a percentage of it stops meaning anything. Divide by sizes and one landmark falls out of the arithmetic. The share is exactly half when the two movements are the same size, and one of the two submitted dates at which that happens is the previous forecast completion itself. An update that reports no movement at all is always a fifty per cent revision share, and an update that reports no movement is the one most likely to be read as an update in which nothing happened.
• Days here are calendar days, and that is not a shortcut The distance a forecast completion date moves between two updates is an absolute measure on the calendar. It does not depend on which activities were working, on a five day week, or on a holiday list. Activity float is the opposite: it is relative, and calendars and constraints move it. This page measures the first and computes nothing that behaves like the second. If the covering report quotes the same movement in working days it is quoting a different measure of the same gap, and the two are reconciled by the programme's own calendar rather than by anything on this page.

A worked example

Shipped defaults: previous 16 March 2026, progress only 3 April 2026, submitted 10 March 2026.

The two subtractions. 16 March to 3 April is $15 + 3 = 18$ days, so the progress movement is $+18$ days, later. 3 April to 10 March is $-3 - 21 = -24$ days, so the revision movement is $-24$ days, earlier.

The check that the two are one update. $+18 + (-24) = -6$, and 16 March to 10 March is $-6$ counted directly. The two movements add to the net, always, because both are subtractions along the same line. If they do not, one of the three dates was read from the wrong programme.

The share. Added as sizes, $18 + 24 = 42$ days. The revision step holds $\frac{24}{42} = 0.5714 \approx 57\%$, so 57%, and the progress step holds $\frac{18}{42} = 0.4286 \approx 43\%$, so 43%. The two total 100%, and neither of them is a share of the six days a reader can see.

What the two issued programmes show. Six days, earlier. That is the whole of it. Forty two days of movement happened in this update and thirty six of them are invisible on the two programmes, because eighteen days of movement in one direction and twenty four in the other leave six behind.

Where the identity sits, and it is worth checking by hand. Hold the first two dates and walk the submitted date. At 3 April the revision movement is zero and the share is $\frac{0}{18} = 0\%$. At 21 March the revision movement is $-13$ and the share is $\frac{13}{31} \approx 42\%$. At 16 March, the previous forecast completion itself, the revision movement is $-18$, the share is $\frac{18}{36} = 50\%$, and the update reports no movement at all. At 21 April the revision movement is $+18$, the share is $\frac{18}{36} = 50\%$ again, and the update reports thirty six days later. The two dates at which the share is exactly half are the date that reports nothing and the date that reports double.

And the same forty two days can report almost anything. At 3 April the update reports eighteen days later and the share is nought. At 16 March it reports nothing and the share is half. At 10 March it reports six days earlier and the share is fifty seven per cent. Three updates, one recorded progress figure, and three different sentences on the covering report.

Switch to The file did most of it and the shape inverts. Previous 16 March, progress only 20 March, submitted 22 April. The progress movement is $+4$ days and the revision movement is $+33$ days, both the same way, so the net is $+37$ and equals the combined size. The share is $\frac{33}{37} \approx 89\%$. Here the net does carry the whole movement, because nothing pointed the other way, and the share is a share of what the reader can see. That equality holds whenever the two movements do not point against each other, which includes the case where the revisions moved nothing at all, and it fails the moment one of them turns back.

Check it yourself

Load The same date, twice over and read the top row of the readout against the bottom one. The two issued programmes carry the same completion date, the net is zero, and thirty six days of movement are still there, eighteen in each step, with the share sitting exactly on the reference rule in the second chart.

What this does not model

Who caused either movement and whether either was agreed, notified or accepted; the cause of anything; cost, resources and any sum of money; which activities moved and whether the critical path switched between the two programmes; float of any kind, total or free; the programme's own calendars, holidays and working week; whether the progress only date was calculated with retained logic or with progress override, which is a software setting that changes the middle date without changing a single recorded fact; inserting an event into a programme to measure it, which is a third operation and a different question from either of the two here; more than one update at a time, and therefore any accumulation across a run of months; whether the recorded progress behind the middle date is complete or evidenced; and whether either issued programme is a reasonable representation of the works.