Decisions.

The framework

1. 01

   The last five deliveries are your forecast.

   This is Day 54's base rate, pointed at time rather than at odds. You do not need an industry study to run it, because the reference class that will actually convince you is the one you delivered yourself. Take the last five pieces of work you scoped and write each as a pair: what you said, what it took. The ratio between those two columns is the most useful number in your business, and almost nobody has it anywhere in writing, which is why it keeps arriving as a surprise instead of as an input. Do it before you produce this morning's figure, not after — a number you have already quoted stops being an estimate and becomes a position you defend. And when you hear yourself explaining that this job is not comparable, notice that every job on that list was also not comparable.

2. 02

   Count the waiting, not just the working.

   Ask how long something takes and you will price the hours of effort, because effort is the part you can picture. What actually consumes the calendar is everything between the effort: the wait for a decision, the handover to a team with its own queue, the review that comes back marked up, the supplier who has gone quiet, the fortnight when the one person who understands it is away. Give each of those its own line in the estimate, in the open, or they arrive later as surprises with a story attached. Working time is optimistic by a margin. Waiting time is where the schedule is genuinely spent.

3. 03

   Hold one visible buffer at the end — never padding inside every task.

   Most teams answer the fallacy by quietly padding each task, which is the worst available fix. Cyril Northcote Parkinson made the mechanism famous in The Economist in 1955: work expands to fill the time available. Hidden slack gets consumed anyway, nobody starts early on a task that looks comfortable, and the project is late regardless. So strip the individual estimates back to honest working figures, then hold a single buffer at the end of the chain, owned by one named person, spent deliberately and reported. A public buffer you can watch draining. Scattered padding you discover has gone in the final week.

4. 04

   Separate the estimate from the commitment — then log the actual.

   Two very different sentences get compressed into one number, and the compression is where the damage happens. An estimate is your honest read, uplifted, still moving as you learn. A commitment is a date you are staking something on — a customer, a launch, a hire — and it must carry the uplift already inside it. Name which of the two you are handing over, in the message that carries the number, every time. Then do the part almost nobody does: on delivery day, write what it really took beside what you said. Ten of those rows and the argument about whether you are an optimist is over. You have the ratio instead, and next quarter's estimate starts from it.