← Community

Timebox a spike

Answer one unknown in a clock, write the answer down, and throw most of the code away — instead of starting the feature in order to learn. Use when a slice cannot be estimated, before a rewrite, or when two approaches are both plausible.

---
name: timebox-a-spike
category: coding
description: >-
  Answer one unknown in a clock, write the answer down, and throw most of
  the code away — instead of starting the feature in order to learn. Use
  when a slice cannot be estimated, before a rewrite, or when two
  approaches are both plausible.
---

# Timebox a Spike

A spike is a **question with a deadline**. It is not a small version of
the feature, not a prototype you will "just clean up," and not a branch
that grows until it is the rewrite.

If the deliverable is code you intend to ship, you are implementing. Use
a behaviour slice. A spike is for when you cannot write the Then because
you do not yet know what is true.

## Write the question and the clock first

One question, answerable in hours or a day or two, not a week:

- Weak: "Look into the search stack."
- Strong: "Can we get p95 < 200ms on last week's production queries with
  the existing index, or do we need a new store? Answer: yes with X, or
  no because Y."

The output is **a written answer**, not a PR. If you cannot phrase the
question so that two people would recognise the answer, you are not
ready to spike — you are fishing.

Set the clock before you open the editor. When it rings, you stop
coding and write. Extending "because it is almost working" is how
spikes become the project. If you need more time, that is a **new**
question, budgeted, not a silent overtime of the old one.

Shared publicly by the author and open for any use. See the Terms.