Small Focused Experiment
Each A/B test or design experiment must change exactly one variable (or a tightly coupled cluster of related changes) — never bundle unrelated changes into a single experiment.
$ aoe install @community/rule-small-focused-experiment Projection
Always in _index.xml · the agent never has to ask for this.
SmallFocusedExperiment [rule] v1.0.0
Each A/B test or design experiment must change exactly one variable (or a tightly coupled cluster of related changes) — never bundle unrelated changes into a single experiment.
Loaded when retrieval picks the atom as adjacent / supporting.
SmallFocusedExperiment [rule] v1.0.0
Each A/B test or design experiment must change exactly one variable (or a tightly coupled cluster of related changes) — never bundle unrelated changes into a single experiment.
Meta
- License: Apache-2.0
- License Provenance: package-authorship-claim
Severity
warning
Applies When
scoping any A/B test, design experiment, or product change evaluation
Verify By
Confirm the experiment changes exactly one variable or a tightly coupled cluster; reject experiments that bundle independent changes (e.g., changing CTA copy AND layout AND color simultaneously).
Loaded when retrieval picks the atom as a focal / direct hit.
SmallFocusedExperiment [rule] v1.0.0
Each A/B test or design experiment must change exactly one variable (or a tightly coupled cluster of related changes) — never bundle unrelated changes into a single experiment.
Meta
- License: Apache-2.0
- License Provenance: package-authorship-claim
Severity
warning
Applies When
scoping any A/B test, design experiment, or product change evaluation
Verify By
Confirm the experiment changes exactly one variable or a tightly coupled cluster; reject experiments that bundle independent changes (e.g., changing CTA copy AND layout AND color simultaneously).
Examples
- Good: test only the CTA button label — 'Start free trial' vs 'Get started'.
- Good: test a redesigned hero section (headline + subtext + CTA are one coupled cluster).
- Bad: test new nav + new pricing table + new onboarding modal in a single experiment — impossible to attribute.
Relations
related: [@community/rule-user-value-metrics]
Rationale
Multi-change experiments produce ambiguous results where it is impossible to attribute outcomes to specific decisions. When an experiment moves the needle, you must know what caused it.
Meta
- License: Apache-2.0
- License Provenance: package-authorship-claim
Severity
warning
Applies When
scoping any A/B test, design experiment, or product change evaluation
Verify By
Confirm the experiment changes exactly one variable or a tightly coupled cluster; reject experiments that bundle independent changes (e.g., changing CTA copy AND layout AND color simultaneously).
Source
aoe-engine/examples/frontend-design/primes/compiled/@community/rule-small-focused-experiment/atom.yaml