A calendar that knows nothing you did not tell it.

A season is a dozen productions. Each one has an opening night that will not move, and thirty pieces of work that have to happen before it: a set model signed off by the workshop, costumes cut and sewn, a lighting run, a poster shoot, the December venue booked.

It is project work, and it has the two properties that make project work hard. The productions share people — one wardrobe supervisor across three shows — so they cannot live in three separate systems. And the thing everybody actually needs to know is not how much is left but when, and what is standing in the way.

Before this, the team could answer neither from a tracker. A board says what is moving. A list of due dates says what is late. Neither says that costumes and the poster shoot are on a collision course, three weeks before anybody notices.

One board for the whole season; a production is a cut of it. Not a project per show — that would mean a dozen numbering schemes, and an issue that turns out to belong to a different production would have to change its name to move. A production here is a label the project has declared meaningful. Work joins by carrying it and leaves by dropping it.

A new view: the calendar. Beside the board, the table, the wiki, the statistics and the forecast — every piece of work on one axis, with today marked. The production switch above it is the same switch that sits above the board, and the same choice: narrow the board to Twelfth Night, open the calendar, and you are still looking at Twelfth Night.

Dependencies, drawn only where somebody stated them. An arrow in red means this work is standing before the thing it waits on is finished. That is the answer to “what is this production up against”, and it is the one thing on a chart that must never be inferred: two bars that happen to line up are not a dependency, and drawing one invents a constraint nobody agreed to.

An issue here has a due date. It has no start date, because nobody has ever reliably filled one in, anywhere.

So where does the left end of a bar come from? Four answers, and only one of them is honest:

1. 1.

   Add a start field. True, and it becomes one more field nobody fills. A chart whose left ends are blank on nine rows in ten is a chart nobody opens twice.

2. 2.

   Guess it. Created-at, or the due date minus a fortnight. Always available, always draws, looks finished. This is the one that cannot be allowed. A bar drawn from an invented start is pixel-for-pixel identical to one drawn between two recorded days, and the reader’s entire reason for opening a Gantt is to find out how long things take. A blank says “we don’t know”. A guessed bar says “three weeks”, in the same voice the true ones use.

3. 3.

   Draw only the deadlines, as points. Honest and thin. That is a deadline calendar, not a schedule, and it cannot show that something has been running for three weeks.

4. 4.

   Use what was already recorded. The tracker does not know when work was planned to start. It knows when it did: the first time somebody moved it into work.

We took the fourth, and the third where the fourth has nothing to say:

• A bar runs from the day work was first picked up to the day it closed.

• While it is still running, the bar reaches today — and past today, up to the stated deadline, it is drawn as an outline. That part is a date somebody named, not work that has happened. One colour for both would be the same over-claim as an invented start, arriving from the other end.

• One known day and no other is a point, not a bar.

• No known day at all is not on the chart. It is named underneath, by key and by title, so a chart that leaves out a third of a season cannot be mistaken for the season.

• The deadline is a mark, never an end. Work finished on the 10th that was due on the 20th ran until the 10th — and the 20th is still drawn, off the end of the bar. Finished late shows. So does finished early.

Every refusal above costs something. The chart has holes in it that a competitor’s does not. A season with no dates entered shows a month, a line marking today, and a sentence explaining that nothing is on the calendar yet.

That is the trade, made deliberately: a schedule you can act on has to be a schedule you can believe, and the cheapest way to lose that is to fill the gaps with something plausible. A plan is worth reading to the extent that you know which parts of it are facts.