A PRACTICAL WORKFLOW FOR EXPERIMENTATION TEAMS: HYPOTHESIS GENERATION, ICE SCORING, AND THE COMPOUNDING LOOP WITH CLAUDE + AIRTABLE
The Hypothesis Bank Playbook: How to Generate and Score Experiment Ideas with Claude + Airtable
This playbook is designed for product, growth, and experimentation teams who want a structured, repeatable system for generating high-quality A/B test hypotheses. It covers the two arms of hypothesis generation — proactive and reactive — how to build your Airtable hypothesis bank, and how to use Claude for unbiased ICE scoring, so your team always has 30+ well-evidenced ideas ready to test.
Why Download the Hypothesis Bank Playbook?
This playbook gives you a concrete, step-by-step workflow to stop running out of experiment ideas — and start building a system that compounds with every test you run.
✅ Build Your Hypothesis Bank in Airtable: Set up the four core databases — Hypothesis Bank, Scoring DB, Results DB, and Meeting Notes — with every field your team needs to run the system.
✅ Generate Hypotheses with Claude (Two Ways): Use the proactive arm to pull from Amplitude, session replays, UX audits, and customer feedback. Use the reactive arm to automatically generate new ideas from completed experiment results.
✅ Write Hypotheses That Actually Hold Up: Use the standard "We believe that X for Y will cause Z because [evidence]" formula — with a real client example from a checkout optimisation test.
✅ Score with ICE Without the Bias: Learn the exact sequence — human scores first, Claude scores second, then compare the gap — so prioritisation is grounded in data, not whoever speaks loudest in the room.
✅ Close the Compounding Loop: Understand how the three phases (Generate & Score → Run & Evaluate → Learn & Compound) connect so every result automatically feeds the next sprint.
Who is This For?
✅ Product & Growth Teams: Get a structured system to fill your experiment backlog with well-evidenced, properly scored hypotheses — and stop starting every sprint from scratch.
✅ Analysts & Data Teams: Learn how to connect Amplitude data, session replays, and customer feedback directly to Claude so hypothesis generation becomes a repeatable, data-driven workflow rather than a brainstorm.
✅ CRO & Marketing Teams: Understand how to use ICE scoring correctly, why the human-first sequence matters, and how to document learnings so your experimentation programme compounds over time.