Spike
A short, timeboxed investigation or prototype used to answer a specific question and reduce uncertainty before a delivery decision.
Download Markdown (.md)Soroush’s Product Wiki · https://wiki.soroush.uk/topic/spike/
01What it means
A spike is a bounded exploration used when a team lacks enough knowledge to choose an approach, assess feasibility, split work, or form a useful forecast. Its output is evidence or a decision, not a production-ready feature; spikes are an optional agile technique and are not prescribed Scrum artifacts.[1][2]
02Why it matters
Unknowns can turn estimates and commitments into guesses. A focused spike buys a controlled amount of learning before larger investment and makes uncertainty visible instead of hiding research inside delivery work.
03When to use it
Use a spike for a consequential technical, functional, design, or operational question that cannot be answered cheaply from existing evidence. Avoid spikes for routine refinement, open-ended research, or questions that a quick conversation or documentation check can resolve.
04How it works
- State the decision or uncertainty and why resolving it matters.
- Define the questions, evidence expected, and explicit timebox.
- Choose the smallest research, experiment, or prototype capable of producing that evidence.
- Conduct the investigation without turning it into hidden feature delivery.
- Summarize findings, remaining uncertainty, and the recommended next decision.
- Stop at the timebox and decide whether to proceed, adapt, investigate further, or abandon the work.
05A practical example
Before promising real-time bank synchronization, a finance team runs a two-day spike against three representative institutions. It measures authentication failures and refresh latency, documents unsupported flows, and recommends a limited launch instead of building a polished integration prototype.
06Where it breaks
A spike provides bounded evidence, not certainty or production-quality implementation. Teams can misuse spikes to postpone decisions, inflate velocity, or repeatedly relearn the same facts. Throwaway prototypes may also create pressure to ship unsafe code.
Common mistakes
- Starting without a specific question or decision.
- Allowing the investigation to continue beyond its timebox by default.
- Counting spike output as delivered customer value.
- Shipping exploratory code without normal quality work.