Guide

Delay analysis in Gantt CSS

Nine ways to work out what an event cost in time, run from one set of records, with every result set side by side.

Where it lives

Delay analysis and the contract registers it reads from are part of the Contract add-on, which works on top of Planning. The registers hold the notices, compensation events and claims; the programme holds the updates and the actual dates. The analysis brings the two together.

No single method suits every job or every set of records. Gantt CSS runs each method from the same entries, fragnets and updates, so you can rely on the one your contract and your records support, and show how the others compare.

The methods

For each one: the question it answers, what it needs from the programme and the registers, and when it suits. Methods marked “added up” give shares that total the delay; the others take each event alone, so their results are shown but not totalled.

  • Method 1 · Each event alone

    Time impact, each event alone

    It asks
    How far would this one event, on its own, move the finish of the programme as it stood just before the event?
    It needs
    A programme update with a data date before the event, a fragnet that models the event, and the date it started.
    It suits
    Work still under way: backing a notice or a claim as the event happens, or a quotation on the accepted programme.
  • Method 2 · Added up

    Time impact, events added in date order

    It asks
    Adding the events one after another, in the order they started, what does each one add to the finish on top of those before it?
    It needs
    A fragnet for every event and the update that was current before the first of them.
    It suits
    Several events in the same period, or a claim for the extension they cause together.
  • Method 3 · Added up

    Impacted as-planned

    It asks
    If every event is added to the as-planned programme, how far does the planned finish move, and how much of that is each event’s?
    It needs
    The as-planned programme kept as a baseline with its whole programme, and a fragnet for every event.
    It suits
    A final account, or any review where the contract programme is the agreed starting point. It leaves out how the work really went, so read it beside an as-built method.
  • Method 4 · Added up

    Windows

    It asks
    Between one programme update and the next, how far did the finish move, and which entries were live on the work that drove it?
    It needs
    An update with a data date for each period, register entries linked to activities, and who carries the risk of each entry.
    It suits
    Any job with regular updates, during the work or after it. Its totals are the control figure the other methods are checked against.
  • Method 5 · Added up

    Half-step

    It asks
    Within each window, how much of the movement came from the progress recorded, and how much from changes to the programme itself?
    It needs
    A full copy of the programme at every update, and register entries linked to activities.
    It suits
    Updates that also changed durations, links or activities, where delay from progress has to be kept apart from re-planning.
  • Method 6 · Added up

    As-planned vs as-built

    It asks
    Which activities on the planned critical path finished late, by how much, and which entries were on them?
    It needs
    The planned programme kept as a baseline, actual dates for the work, and register entries linked to activities.
    It suits
    After the work, with a good record of actual dates: the plainest picture to put in front of someone. It does not follow the logic, so parallel delays can add up to more than the completion delay.
  • Method 7 · Each event alone

    Collapsed as-built, each event taken out alone

    It asks
    Had this one event not happened, when would the work have finished?
    It needs
    Actual dates, logic, and a fragnet for every event. The rebuilt as-built finish should land on the real one before any event is taken out.
    It suits
    After the work, to show what the event cost. Taking out the employer’s events, the contractor’s events and both gives a cross-check on concurrent delay.
  • Method 8 · Added up

    Collapsed as-built, events taken out in turn

    It asks
    Taking the events out one at a time, latest first, what does each one save on top of those already taken out?
    It needs
    The same as collapsing one event at a time.
    It suits
    After the work, when several events overlap and each needs its own share.
  • Method 9 · Added up

    Weather register

    It asks
    How many days did weather cost beyond the contract’s allowance, counting only days that fell on critical work (and, once activities are marked weather-sensitive, only on those)?
    It needs
    Daily weather records, the weather rules for the contract and the allowance for each month.
    It suits
    Any contract that gives time for exceptional weather.

The records

The methods share their inputs, so a change to an entry or a fragnet shows in every result.

  • Register entries linked to the programme

    Each notice, compensation event or claim names the activities it affects and who carries its risk: the employer, the contractor or neither.

  • A fragnet for each entry

    The new activities and links that model an event are kept on its register entry, so every method that adds or removes the event uses the same model.

  • Programme updates with data dates

    Each update is kept inside the programme file with its data date, as a whole programme or as dates alone. The methods that follow the logic need the whole programme.

  • An as-built record

    The actual dates of every activity as each update recorded them at the time, with later changes kept as revisions and anything doubtful flagged for you to look at.

  • Weather records

    Daily records set against the contract’s monthly allowance, with a rain allowance activity in the programme where you want one.

  • A fingerprint for every run

    Each run carries a fingerprint of what it read and what it gave, so you can show that the same inputs give the same days.

Reading them together

  • Concurrent delay is counted once: where employer and contractor delay drive the same window, the days go in as time without money.
  • Pacing candidates are listed: contractor work off the driving path that fell behind while an employer event drove the window, for you to judge.
  • Results that take each event alone are shown but not totalled, because overlapping events would count the same delay twice.
  • Every method’s days sit side by side for each cause, with a reconciliation against the windows totals.

What comes out

Every output starts from the same register. For the notices and time bars behind it, see contract time bars.

  • Overall delay register

    One row per cause with every method’s days side by side, the reconciliation against the windows, an event by window matrix, a Scott schedule and a short guide to the methods. To Excel and to A3 PDF.

  • Combined EOT report

    One A4 PDF with the contract, the events and whether their notices went in on time, the weather, the analyses, your introduction and conclusion, and the programme printed as an appendix.

  • Claim pack

    A folder with the as-built programme and the programmes behind the analysis a claim relies on, filed with the programme.

  • Save as programme

    Any analysis view can be saved as a programme of its own, so the impacted or collapsed programme can be opened, printed and checked.

  • Draft delay narrative

    A first draft written with AI from the analysis’s facts, each sentence citing the facts it rests on, using your own API key and only when you ask. You check and edit it.

Run it on your own programme

Delay analysis comes with the Contract add-on, on top of Planning.