Throughput

The number of work items a team finishes in a given period, such as per week.

2 min read · Practitioner

In 30 seconds

Counts finished items per week or sprint.
No estimates needed; just count.
The main input for flow-based forecasting.

Try it

Items finished per week Recent weeks, oldest first Items remaining

–50% likely by

–85% likely by

–95% likely by

Monte Carlo forecast: 10,000 simulated futures, each week drawn at random from your history. Plan commitments around the 85% figure.

How to measure it

Count items reaching "done" each week. Track it over time; a run chart shows the range and stability. Throughput works best when items are split to similar, small sizes.

Forecasting

Divide remaining items by throughput to get a range, or use Monte Carlo simulation on past weekly throughput for a probability-based date. See forecasting and velocity vs throughput.

Little's Law

Average lead time = average work in progress / average throughput. Raising throughput or lowering WIP shortens lead time.

Myth vs reality

Myth · Throughput ignores size.

Reality · It relies on keeping items similar in size.

Myth · Higher throughput is always better.

Reality · Splitting work artificially just inflates the count.

Myth · You need estimates to forecast.

Reality · Throughput forecasts well without them.

The AI angle

Throughput counts finished items, which makes it more robust than story points when AI changes effort. Keep items small and similar so the count stays meaningful.

Estimation and Velocity with AI →

FAQ

What counts as an item?

Whatever the team finishes: stories, bugs, tasks. Be consistent.

Should bugs count?

Track them separately or together, but consistently.

How much history do we need?

A few weeks to start; more improves forecasts.