How to Use the Shopify Connector in Claude (Operator Prompts Included)

Using the Shopify connector in Claude with operator prompts

Build-in-public post — the actual prompts we run on connected Shopify stores. Part of our AI workflows pillar. If you haven't connected yet, start with how to connect Claude to Shopify.

The customer pain point

Connecting Claude to Shopify is the easy part. The hard part is that a vague prompt gets a vague answer. "How's my store doing?" gets you a generic summary. The operators who get real value give Claude specific, structured prompts that demand a ranked decision — not a description.

This is the prompt library. Copy them, swap in your collection names, run them on your connected store.

Table of Contents

Key Takeaways

Point Details
The skill Specific prompts that demand a ranked decision, not a summary. Vague in = vague out.
Analytics ShopifyQL through the connector covers AOV, conversion rate, traffic source, sell-through, top products — as charts.
Bulk ops The connector can bulk-update prices/compare-at. Always make Claude show the count + a sample and wait for confirmation before applying.
Discipline Every write prompt ends with "do NOT apply until I confirm." Non-negotiable.

What the Connector Can Do

Faithful summary of the connected capabilities:

The prompts below turn those capabilities into decisions.

Prompt: AOV Diagnosis

Don't ask "what's my AOV." Ask what's moving it and what to do:

Using the Shopify connector, run ShopifyQL to pull average order value
by month for the last 6 months. Then break the most recent full month's
AOV down three ways:
1. New vs returning customers
2. Top 5 collections by order volume
3. Discount-code orders vs full-price orders

Tell me which single segment is dragging AOV down the most, the most
likely reason, and the one highest-leverage change to lift it. Show
the 6-month trend as a chart. Be decisive — one recommendation, not five.

This forces a diagnosis ("returning-customer AOV is fine; new-customer AOV fell because the welcome discount is too deep") instead of a number.

Prompt: Conversion-Rate Change Analysis

The useful question is never "what's my CVR" — it's "why did it change":

Run a ShopifyQL query for site conversion rate, weekly, for the last
12 weeks. Identify any week where CVR dropped more than 15%
week-over-week.

For each flagged week, pull the traffic-source mix and the
device split (mobile vs desktop) for that week vs the prior week.

Tell me, for each drop: was it a traffic-quality change (the mix
shifted toward lower-intent sources) or an on-site change (CVR fell
within the same source mix)? Give me the evidence, then the fix.
Chart the 12-week CVR trend.

The traffic-mix cross-reference is what separates "you have a CVR problem" from "you don't have a CVR problem, you have a traffic-mix problem" — completely different fixes. We cover this distinction in the 7 KPIs that actually move revenue.

Prompt: Bulk Comparison-Price Update for a Sale

This is a write operation — note the confirm gate:

I'm running a 20% sale on the "Summer" collection.

Using the Shopify connector:
1. List every product (and variant) in the "Summer" collection with
   its current price.
2. For each variant, plan to set compare-at price = current price,
   then set price = 80% of current price (round to .99).
3. Show me a before/after table for the first 10 variants AND the
   total count of variants that will change.

Do NOT apply anything yet. Wait for me to reply "apply" after I
review the table.

After you review and reply apply, Claude executes the bulk update. Reversing it later is the mirror prompt: "set price back to compare-at, clear compare-at, for the Summer collection — show count, confirm before apply."

Prompt: Sell-Through → Discount

Chain analytics into an action:

Run a ShopifyQL sell-through query for the last 30 days. Flag every
SKU that has BOTH: sell-through under 10% AND more than 60 days of
inventory on hand at the current sales rate.

List those SKUs with their on-hand quantity. Then draft (do not
create yet) a percentage discount code scoped to ONLY those SKUs'
collection or product set, sized to clear ~50% of the excess in 30
days based on the current rate. Show me the code config and wait
for "create" before making it live.

This turns "we have dead stock somewhere" into "here are the 14 SKUs, here's the targeted code, approve it."

Prompt: PDP vs Pricing Diagnosis

Find why products underperform, not just that they do:

Pull the top 10 products by revenue (last 90 days) and the bottom 10
by conversion rate among products with more than 100 sessions.

For the bottom 10, compare add-to-cart rate vs checkout-completion
rate. Classify each as:
- PRICING problem (decent add-to-cart, drops at checkout)
- PDP problem (low add-to-cart despite traffic)
- TRAFFIC problem (high bounce, low engagement)

Give me the classification table and the single fix for the worst
offender. Chart revenue concentration across the top 10.

The Confirm-Before-Apply Rule

Every prompt that writes — bulk prices, status changes, discounts, collection edits — ends with an explicit "do NOT apply until I confirm." This isn't optional ceremony. The connector has write access to your live store. The value is Claude doing the analysis and staging the change; the safety is you approving it. A bulk price update applied to the wrong collection is a real-money mistake that's tedious to reverse.

The pattern, every time:

  1. Claude analyzes and stages the change
  2. Claude shows you a sample + the total count
  3. You review and explicitly confirm
  4. Claude applies
  5. Claude reports what changed (so you can verify or roll back)

We hold this discipline across every AI workflow we run — the reasoning is in what AI-run means when the AI screws up.

Talk to Branva

We run Claude-operated Shopify analytics and bulk operations — with these exact prompts and review gates — as part of our managed operations. Book a free call and we'll run a live AOV + conversion diagnosis on your connected store.

Frequently Asked Questions

Why does a vague prompt give a useless answer?

Because "how's my store" has no decision attached. Claude pattern-matches to a generic summary. Prompts that specify the cuts (by segment, by week, by source) and demand a ranked output force a real diagnosis. The structure is the value.

Can it really do bulk price updates safely?

Yes — if you keep the confirm gate. The connector stages the change and shows you the count + sample. You approve. Skipping the approval on a bulk write is the only way this goes wrong.

What's ShopifyQL?

Shopify's query language for analytics. The connector runs it for you — you don't write ShopifyQL by hand, you ask in plain language ("conversion rate weekly, last 12 weeks") and Claude composes the query and renders the chart.

Will these prompts work on any store?

The structure works on any connected store. Swap collection names and thresholds for yours. The diagnostic logic (segment the metric, cross-reference, demand a ranked fix) is universal.

Should I let it create discount codes automatically?

Stage, don't auto-create. The sell-through prompt deliberately ends with "wait for create." A mis-scoped discount code live on your store is a margin leak. Same confirm-before-apply rule as bulk prices.

Related reading