Field Notes · Elle Anderson · andersonco.uk

Flow in a nutshell

Third of three short pieces on ways of working, after Scrum and Kanban. Flow is the physics underneath both, and the reason busy organisations are so often slow ones.

One law explains most of it

Little's Law: the time a piece of work takes to get through your system equals the amount of work in progress divided by your throughput. It is arithmetic, not opinion. If your team completes ten items a month and has sixty in flight, the average item takes six months, whatever the plan says, however hard anyone works, and regardless of which framework the calendar is wearing. The fastest way to speed up is almost never to work faster; it is to have less in progress.

Why full utilisation makes you slow

Queueing theory has an inconvenient shape: as utilisation approaches one hundred percent, waiting time does not rise gently, it curves towards the vertical. A motorway at 95 percent capacity is a car park. Organisations that plan every person to full allocation are choosing the car park, and then holding meetings about why everything is late. Slack is not waste; slack is the price of responsiveness, and the organisations that refuse to pay it pay far more in queues.

Batch size is a choice, and most choose badly

Big batches feel efficient and behave terribly: they wait longer, fail bigger, hide defects, and delay feedback in proportion to their size. Halving batch size roughly halves cycle time before you improve anything else. This is why the health check asks whether the typical item reaches customers in weeks, not quarters: the answer is mostly a batch-size confession.

Measure the work, not the workers

Four numbers tell you nearly everything: cycle time (door to door, in calendar days), throughput (finished items per week), work in progress, and the age of every unfinished item. Note what is missing: velocity, story points, and utilisation, which measure effort and estimation rather than delivery. Work item age is the most underused of the four, because it is the only one that warns you while there is still time to act, and the oldest item on the board is almost always the most important conversation nobody is having.

What to do on Monday

Cap WIP somewhere, anywhere, and lower the cap until it hurts slightly. Split the next big thing into the smallest slice that reaches a customer. Put item age on the board and discuss the oldest three at every replan. Measure where work waits, because in most systems work spends over eighty percent of its life waiting for someone, and nobody is measuring the waiting. None of this requires permission, a transformation programme, or a single new meeting.