Learning Repository
A centralized, searchable knowledge base of insights derived from experiments — organized by behavioral principle, audience segment, and business domain rather than by individual test.
What Is a Learning Repository?
A learning repository is the distilled wisdom layer on top of experiment documentation. While documentation records what happened in each test, the learning repository synthesizes patterns across tests into actionable knowledge: "Loss aversion framing consistently outperforms gain framing for our B2B audience in renewal contexts, but not in acquisition contexts."
The repository is insight-centric, not test-centric — one entry per validated learning, cross-referenced to the experiments that produced it.
Also Known As
- Marketing: Insights library, pattern library
- Sales: Win themes library, sales playbook insights
- Growth: Growth insights, learning library
- Product: Product insights repository, customer insights library
- Engineering: Engineering lessons, architectural insights
- Data: Research archive, evidence library
How It Works
A growth team's repository has three sections. "Confirmed learnings" hold insights validated by 2+ experiments: "Social proof works on acquisition pages, confirmed by tests #14, #47, #82." "Emerging hypotheses" hold single-test insights awaiting confirmation: "Urgency framing appears to hurt B2B trials, test #91." "Refuted assumptions" hold things the team used to believe: "We thought longer forms would filter quality; tests #23 and #57 showed they reduce quantity without improving quality."
When a new test is proposed, teams check the repository first. A repeated test on loss aversion becomes a stronger test with better hypothesis because prior learnings inform design.
Best Practices
- Organize by behavioral principle, not by test.
- Use three categories: confirmed, emerging, refuted.
- Promote emerging hypotheses to confirmed after 2+ supporting tests.
- Retire stale learnings when context has changed.
- Make repository review part of every test design process.
Common Mistakes
- Flat lists without structure — unsearchable, unusable.
- Set-and-forget repository — learnings go stale; the repository must be maintained.
- No promotion or retirement process — the repository becomes a graveyard of old insights.
Industry Context
SaaS/B2B: Repositories are especially valuable because test cadence is low and each learning is expensive to reproduce.
Ecommerce/DTC: At high velocity, the repository prevents repeated tests on already-answered questions.
Lead gen: A small repository with 20–30 core learnings covers most decisions for a focused lead gen site.
The Behavioral Science Connection
Learning repositories solve a classic commons problem — documentation benefits future teams but costs current teams time. Individual incentive doesn't align with collective value, which is why few organizations maintain active repositories despite their obvious utility. Structural solutions (making repository contribution a required workflow step) work better than exhortation.
Key Takeaway
The repository should be a living document, not an archive — continuously maintained, regularly reviewed, and central to how tests are designed.