Shopify Analysis Playbook

Overview

Use this skill for fast Shopify analysis, not deep analyst mode.

Optimize for:

• speed
• signal
• clear business relevance

Answer in the user's language when it is clear. If the user's language is unclear, default to English.

Response Shape

Keep the response short and structured:

• Summary: 2-4 bullets with the main performance read
• Findings: 3-6 bullets ordered by impact
• Next checks: 1-3 follow-ups only when they materially reduce uncertainty

Do not dump raw tool payloads unless the user explicitly asks for them.

Core Rules

1. Always state the date range you analyzed.
2. If the user gave no date range, use a sensible default and say so.
3. Prefer a fast top-down read before drilling into edge cases.
4. Do not interpret attribution before checking attribution coverage.
5. Do not overclaim retention quality if order history may be incomplete.
6. Treat unattributed and <none> as real reporting outcomes, not as errors to hide.
7. Keep recommendations tied to evidence from the tools you actually ran.

Default Time Windows

Use these defaults unless the user asks for something else:

• general shop health: last 30 days vs previous 30 days
• product performance: last 30 days
• retention snapshot: cohort starting in the last full 30 days, evaluated as of today
• attribution snapshot: last 30 days
• sales drop diagnosis: current 30-day window vs previous 30-day window

Fast Workflows

1. Revenue Snapshot

Use:

• shopify_sales_overview
• shopify_sales_timeseries
• shopify_top_products
• shopify_discount_analysis when discount pressure may matter

Answer:

• is revenue up, down, or flat
• what changed in orders, AOV, units, refunds, and discounts
• which products drove the move

2. Product Performance Snapshot

Use:

• shopify_top_products
• shopify_sales_overview
• shopify_sales_timeseries if trend is relevant

Check:

• net sales vs units sold
• whether a few products dominate results
• whether product mix changed more than total demand

3. Customer and Retention Snapshot

Use:

• shopify_retention_overview
• shopify_repeat_purchase_windows
• shopify_time_to_second_order
• shopify_inactive_customer_summary if health looks weak

Check:

• repeat customer rate
• time to second order
• whether inactivity is widening

Guardrail:

If order history may be incomplete, say retention is directional rather than definitive.

4. Attribution Sanity Check

Always start with:

• shopify_attribution_quality_summary

Then use:

• shopify_sales_by_source
• shopify_sales_by_utm
• shopify_new_customers_by_attribution
• shopify_landing_page_analysis

Check:

• how much of the order base has usable source, landing page, or UTM data
• whether one source or campaign dominates
• whether the data is good enough for a confident attribution read

Guardrail:

If a large share is unattributed, say that clearly before discussing winners.

5. Sales Drop Diagnosis

Use:

• shopify_sales_overview with previous-period comparison
• shopify_sales_timeseries
• shopify_top_products
• shopify_top_customers
• shopify_discount_analysis

Add attribution checks only if the drop may be acquisition-driven:

• shopify_attribution_quality_summary
• shopify_sales_by_source
• shopify_sales_by_utm

Diagnose in this order:

1. demand change
2. mix change
3. discount or refund pressure
4. retention weakness
5. acquisition weakness

Common Traps

• Do not treat higher revenue as healthy if refunds or discounts also spiked.
• Do not treat sourceName alone as full attribution truth.
• Do not assume numberOfOrders is enough for true new-customer analysis; prefer the dedicated attribution and retention tools.
• Do not produce a long narrative when 4-6 bullets are enough.
• Do not recommend actions that the data did not support.

Output Heuristic

If confidence is high:

• give one clear read
• state the strongest driver
• give one or two next checks

If confidence is low:

• say what is knowable
• name the data-quality limit
• recommend the next highest-value check