Collaboration & Process › Estimation & Planning

Breaking Down Tasks

Splitting work into pieces small enough to finish and estimate.

Also known as: task breakdown, splitting work, work breakdown, decomposing tasks, slicing work

Big tasks hide uncertainty. “Build the notifications feature” can’t be estimated, started or reviewed meaningfully. Breaking it down means splitting it into pieces small enough to finish, estimate and ship, ideally each taking a day or two.

How to split

Instead of splitting by technical layer (“do the database, then the API, then the UI”), prefer thin vertical slices that each deliver something that works end to end (vertical slice):

Too big:  Build notifications

Slices:
  1. Send a single email notification when an order ships (hard-coded template)
  2. User can turn email notifications off in settings
  3. Add notification history page
  4. Add push notifications on mobile
  5. Group notifications that arrive close together

Each slice is usable, testable and can be shipped. You get feedback earlier, and can stop when the value runs out.

Techniques

• Ask “what’s the smallest thing that’s useful?” and start there.
• Split by user steps, rules or data variations: one payment method first, then the others.
• Separate the risky or unknown parts and investigate them first with a time-boxed spike.
• List what you don’t know as questions to answer before committing to an estimate.
• Peel off non-functional work: tests, migrations, monitoring, documentation, as explicit items instead of an afterthought.
• Use the Definition of Done as a checklist for each piece (definition of done).

Signs a task is still too big

• You can’t describe it in a sentence, or can’t say when it’s finished.
• Estimating it feels like a guess (“a week… maybe two?”).
• It would produce a huge pull request (small PRs).
• It depends on several other things being finished first.

Good habits

Write the breakdown down in the ticket, review it with a teammate, and adjust as you learn. A breakdown is a plan, not a contract. Better to discover a missing piece on day one than on day nine. See estimation.