What is iteration?
Iteration is the disciplined practice of cycling through small, measurable changes — building, releasing, measuring impact, learning, and changing course — rather than executing a long predetermined plan. Eric Ries’ Lean Startup formalised the build-measure-learn loop as the core mechanism by which early-stage companies find product-market fit. The shorter the cycle, the faster the learning.
The iteration loop
- Hypothesis: a specific, falsifiable claim — “adding feature X will lift activation rate by 5 points”.
- Build: the smallest implementation that can test the hypothesis.
- Measure: instrument and run the test for enough volume to draw a conclusion.
- Learn: compare result to hypothesis. Validate, invalidate or refine.
- Decide: persevere, pivot or kill.
What good iteration looks like
- Short cycle time: weekly or biweekly cycles for product changes; even faster for marketing experiments.
- Small bets: each iteration tests one hypothesis at a time; avoid bundling.
- Written hypotheses: documented predictions before the test, so post-hoc rationalisation does not corrupt learning.
- Honest measurement: picking metrics in advance, including the kill criteria.
Iteration vs. related concepts
- Iteration vs. agile: agile is the engineering process for short release cycles; iteration is the product-discovery mindset that uses those cycles to learn.
- Iteration vs. pivot: iteration is incremental refinement within a hypothesis; pivot is a fundamental change of hypothesis or business model.
- Iteration vs. perfection: iteration trades polish at any single point for cumulative learning velocity.
Common iteration failures
- Confirmation bias — designing tests that cannot disprove the founder’s belief.
- Insufficient volume — concluding from too few users or too short a window.
- Pivoting without learning — abandoning hypotheses based on noise rather than signal.
- Iterating in the wrong dimension — improving the product when the binding constraint is positioning or pricing.
Do: write the hypothesis and success criteria before you build, and run for long enough to draw a clean conclusion.
Don’t: iterate on product micro-features when the data says retention or positioning is the binding constraint — solve the right problem.