Shopify AI: Ask Your Store Data Questions

An AI answer about your store is only useful if its net sales matches the number Shopify shows. Here is what has to be true for that, and the questions worth asking first.

You can ask a Shopify store questions in plain English, and the answers are useful as long as the underlying numbers match what Shopify reports. The test is simple: ask for net sales last month and compare it with Shopify's own sales report. If they agree, the rest of the questions are worth asking. This post covers the definitions that have to line up, the questions that pay off first, and where store data runs out.

Make the first answer match Shopify

Before trusting any tool, pick a closed month and write down three figures from Shopify's reports: net sales, orders and average order value. Then ask the AI layer for the same three. Most disagreements trace to a short list:

• Returns on the wrong date. Shopify counts a return in the month it was refunded, not the month of the original order. A tool that subtracts returns from the original order month will not match, and the difference is largest in months after a holiday peak.
• Shipping and tax. Net sales excludes them. Total sales adds them.
• Gift cards. Excluded from sales in Shopify's reports.
• Time zone. Days and months follow the store's time zone, so orders near midnight land on the right day only if the tool uses the same zone.
• Test orders and cancellations. Test orders are excluded. Orders counts include cancelled orders.

The full walkthrough is in Net Sales vs Gross Sales in Shopify. The short version is that net sales equals gross sales plus discounts plus returns, where discounts and returns are negative numbers.

The questions worth asking first

1. What were net sales by month this year? The reference check and the baseline.
2. Which products drove the most net sales last quarter? Usually reveals that a few products carry the store.
3. What is our average order value by sales channel this year? Online store, point of sale and other channels often behave differently.
4. How many new customers did we get each month? Separates acquisition from repeat purchase.
5. Which products get returned most? Returns concentrate in a few items more often than they spread evenly.
6. Which products are running low on inventory? Reads the current inventory quantity per variant.

How Quiriz's Shopify connector answers them

The connector is read-only and refreshes daily; it is in beta. It imports six datasets: [Shopify] Orders, [Shopify] Order line items, [Shopify] Refunds, [Shopify] Sales, [Shopify] Products and [Shopify] Customers. The Sales dataset is a ledger with one row per dated sale or return line, built the way Shopify's sales report is: orders sit on the order date and returns, which are negative, sit on the refund date. Net sales there is gross sales plus discounts plus returns, so it is designed to match Shopify.

Suggested metrics include Gross sales, Discounts, Returns, Net sales, Total sales, Orders, Order value, Average order value, New customers and Units sold. Average order value is order value divided by orders, which is Shopify's definition: before returns, shipping and tax. An admin stores the definitions once in Company Context so the answer to "what is AOV" does not depend on who asks.

Customer data comes without names, emails, phone numbers or addresses. A customer row holds an id, order count, amount spent, first and last order dates and marketing state, which is enough for new-versus-returning analysis without personal details.

An example of a useful follow-up

Say net sales for a month look flat but orders are up. A good second question is "average order value by sales channel, this month versus last". Illustrative result: if point-of-sale AOV held and online AOV fell, the discount code rather than demand is the likely story. From there, "discounts by month" confirms it. The point is that each follow-up is a short question against the same defined metrics, not a new pivot table.

What to do in the first hour

Connect the store, wait for the first import to finish, then spend ten minutes on the reference check rather than exploring. Ask for net sales, orders and average order value for one closed month and put them beside Shopify's report. Then try one question from each family: a sales question, a customer question and a product question. That tells you quickly whether the data you have is deep enough, particularly whether full order history was approved, before you build habits around it.

If you run more than one store, each connection covers a single store, so check which one you are asking about, and say the store name in the question when you have access to several. Currency is the store currency, and days follow the store time zone, so multi-store comparisons need the currencies noted.

Honest limits

• Order history. Without approval for full order history, only the last 60 days of orders are available. Reconnecting and approving access fixes it.
• Store size cap. Very large stores import the newest whole months up to an order cap, and the connector says so when it applies.
• Product cost is current cost. Margin on old orders uses today's unit cost.
• Guest checkouts have no customer record, so they do not count as new customers.
• It can still be wrong. Governed definitions and visible work reduce errors; they do not eliminate them. Check important figures against Shopify.

If you advertise too, joining Shopify and ad exports covers the next step, and Quiriz for ecommerce collects the store workflows.

Frequently asked questions