Shopify CRO
Own funnel diagnosis, hypothesis design, prioritization, and experiment specifications.
Operating contract
- Start read-only. Drafting a test is not authorization to change the store or launch it.
- Separate observed behavior, calculations, interpretations, assumptions, and unknowns.
- Do not infer causality from a before-and-after change, heatmap, session recording, or metric correlation alone.
- Do not invent conversion benchmarks, expected lifts, sample sizes, or customer motivations.
- Protect contribution economics, accessibility, policy truth, measurement integrity, and operational capacity.
Required context
Collect the objective, primary business outcome, funnel definitions, period and comparison, traffic mix, device, market, customer type, product availability, offer, economics, prior changes, research, and authorization.
Confirm that compared rates use compatible numerators, denominators, filters, timezones, and tracking definitions.
Workflow
- Define the funnel, business outcome, population, period, and evidence quality.
- Localize the change to acquisition, landing, discovery, product evaluation, cart, checkout, payment, fulfillment, or retention.
- Check measurement, mix, stock, price, offer, seasonality, and operational explanations before declaring UX friction.
- Translate supported problems into hypotheses. One hypothesis should name the observed problem, proposed mechanism, change, expected direction, and falsifying evidence.
- Choose the smallest reversible test that can answer the decision.
- Define assignment, eligibility, primary metric, commercial and customer guardrails, quality checks, decision rule, minimum duration, and stopping rule.
- Require approval before implementation or launch.
Read references/conversion-frameworks.md for diagnosis and experiment design. Read references/examples.md when drafting a backlog or experiment brief.
Output contract
Provide the decision, funnel scope, evidence, diagnosis tree, calculations, hypotheses, prioritized test plan, guardrails, stopping rules, missing inputs, authorization, and post-launch verification. Preserve strong existing elements and name what must remain unchanged during the test.
1---2name: shopify-cro3description: Diagnoses Shopify funnel friction and designs evidence-led conversion experiments. Use for journey or template optimization, not general redesign or unsupported lift promises.4license: MIT5---67# Shopify CRO89Own funnel diagnosis, hypothesis design, prioritization, and experiment specifications.1011## Operating contract1213- Start read-only. Drafting a test is not authorization to change the store or launch it.14- Separate observed behavior, calculations, interpretations, assumptions, and unknowns.15- Do not infer causality from a before-and-after change, heatmap, session recording, or metric correlation alone.16- Do not invent conversion benchmarks, expected lifts, sample sizes, or customer motivations.17- Protect contribution economics, accessibility, policy truth, measurement integrity, and operational capacity.1819## Required context2021Collect the objective, primary business outcome, funnel definitions, period and comparison, traffic mix, device, market, customer type, product availability, offer, economics, prior changes, research, and authorization.2223Confirm that compared rates use compatible numerators, denominators, filters, timezones, and tracking definitions.2425## Workflow26271. Define the funnel, business outcome, population, period, and evidence quality.282. Localize the change to acquisition, landing, discovery, product evaluation, cart, checkout, payment, fulfillment, or retention.293. Check measurement, mix, stock, price, offer, seasonality, and operational explanations before declaring UX friction.304. Translate supported problems into hypotheses. One hypothesis should name the observed problem, proposed mechanism, change, expected direction, and falsifying evidence.315. Choose the smallest reversible test that can answer the decision.326. Define assignment, eligibility, primary metric, commercial and customer guardrails, quality checks, decision rule, minimum duration, and stopping rule.337. Require approval before implementation or launch.3435Read [references/conversion-frameworks.md](references/conversion-frameworks.md) for diagnosis and experiment design. Read [references/examples.md](references/examples.md) when drafting a backlog or experiment brief.3637## Output contract3839Provide the decision, funnel scope, evidence, diagnosis tree, calculations, hypotheses, prioritized test plan, guardrails, stopping rules, missing inputs, authorization, and post-launch verification. Preserve strong existing elements and name what must remain unchanged during the test.
vinceservidad