AI Analytics for Shopify: Ask Your Store Data Questions
You can now ask your store which products lose money after ads, in plain English, and get an answer in seconds. Whether that answer is true depends on one boring column.
Your store can talk now. That's the easy part.
Type "which products lost money after ads last month?" into your Shopify admin and you'll get an answer in about the time it takes to sip your coffee. The hard part is whether the answer is true.
Here's the short version for anyone searching ai analytics shopify: yes, you can ask your store data questions in plain English today. Shopify's analytics editor works with Sidekick, its built-in assistant, and it turns a sentence into a report. A general chat assistant can read a CSV export. Both are fast. Both will confidently give you the wrong product to cut if your cost data is blank, your shipping is missing or your ad spend lives somewhere they can't see.
So this guide does two things. It shows you the questions worth asking. And it shows you the twelve fixes that make the answers worth trusting.
What can AI analytics on Shopify actually do today?
More than it could a year ago, and less than the demos suggest.
Shopify's own changelog says the analytics query editor now works with Sidekick. You ask something like "what was my best converting traffic source in September?" and it builds the report. You ask a follow-up and it remembers the conversation. It explains what each report measures, and you can still write ShopifyQL yourself if you want to check its work.
The Sidekick help page goes further: Sidekick can write and run its own code to forecast, compare or summarize your store data. Shopify also notes that in some stores it can't run code. And Shopify's enterprise guide to Sidekick questions says it's included on all plans and sees only what the logged-in user can see.
Then there's the do-it-yourself route. Export a report, upload it to a chat assistant, ask away. That's what most founders mean when they search "chatgpt shopify data". It works for a one-off. A Coupler.io guide (they sell a connector, so weigh it) lists the traps: Shopify exports one report type at a time, files carry summary rows the model may count as orders, and the export habit usually dies by month three.
| Route | Good for | Blind spot |
|---|---|---|
| Sidekick in the Shopify admin | Sales, orders, customers, traffic, live, no export | Ad spend from Meta and Google isn't in your Shopify data |
| CSV into a chat assistant | One-off questions, joining files you choose | Stale snapshot, summary rows, you do the SKU matching |
| A profit app feeding the AI | Ad spend and fees pulled in daily | You inherit the app's attribution choices |
AI didn't make your data smarter. It made your data faster. If the data was lying before, now it lies in ten seconds.
Why does AI give confident wrong answers about profit?
Because it reads what's there and says nothing about what's missing.
Here's the line from Shopify's profit reports documentation that matters most: profit is reported only for products and variants that had a cost recorded at the time they were sold. Each variant needs its own cost per item, and that field excludes shipping.
Read that twice. If your eye balm has no cost entered, it doesn't show up as a loser. It doesn't show up at all. Ask the AI for your least profitable product and it'll rank the products it can see, and hand you an answer with a straight face.
And ad spend? It lives in Meta and Google. Unless something pipes it in, Shopify's numbers stop at gross profit. Which means the one question every founder actually wants answered, "which products lose money after ads?", is the one question your store can't answer on its own.
I'll admit something. The first time I saw a product ranked by "profit" in a clean AI-built report, I believed it. It looked so finished. Then I checked the variant costs and a third of them were blank.
Which products lose money after ads? Here's the math
Run the math on a store like this, a hypothetical. A skincare brand, four products, one month.
| Product | Orders | Price | Revenue | Product cost | Shipping and fees | Ad spend | After ads |
|---|---|---|---|---|---|---|---|
| Vitamin C serum | 400 | $60 | $24,000 | $6,000 | $3,600 | $9,000 | $5,400 |
| Night cream | 300 | $48 | $14,400 | $4,320 | $2,700 | $3,000 | $4,380 |
| Eye balm | 250 | $36 | $9,000 | $3,150 | $2,250 | $5,200 | -$1,600 |
| Cleanser | 200 | $28 | $5,600 | $1,680 | $1,800 | $600 | $1,520 |
Shipping and fees are $9 an order across the board. Ask the AI for revenue by product and the eye balm looks like a solid third-place seller. Ask for gross profit and it still looks fine: $9,000 minus $3,150 is $5,850.
Now add the two columns Shopify can't see on its own.
Contribution before ads is revenue minus product cost, shipping and fees. Ad spend is bigger, so the product loses $1,600.
The eye balm earns $3,600 after product cost, shipping and fees. You spend $5,200 to sell it. Every month it runs, you pay $1,600 for the privilege.
Here's the math on the page itself. The eye balm page gets 12,500 sessions a month. 250 orders is a 2.0% conversion rate. Average order value is $36. That means revenue per visitor is 2.0% x $36 = $0.72. On 12,500 visitors, that's $9,000. Meanwhile ads cost $5,200 divided by 12,500, about 42 cents a visitor, on a page earning 29 cents a visitor before ads ($3,600 divided by 12,500).
That's a leak with an ad budget attached. Pour more spend in and you amplify the leak.
Compare the serum. 20,000 sessions, 400 orders, 2.0% conversion rate, $60 average order value. Revenue per visitor: 2.0% x $60 = $1.20. On 20,000 visitors, $24,000. Same conversion rate, very different machine. For the full method of ranking products this way, the profit per product walkthrough does it line by line.
What questions should you ask your store data?
Start with questions that end in a decision. "Show me sales trends" ends in a chart. "Which product should I stop advertising?" ends in a Monday action.
Here are the ones I'd type first, in this order:
- "Show contribution after ads by product for the last 30 days, sorted lowest first." (Only after ad spend is in the data.)
- "Which products have no cost per item on one or more variants?"
- "Show conversion rate and average order value for my top five product pages, side by side."
- "Compare new versus returning customer revenue for the last 90 days."
- "Which traffic source converted best last month, and which converted worst?"
- "Which products had refunds above 8% of orders this quarter?"
Notice question 2. It's the question about the question. Ask it before you trust anything else. If you want the six numbers these answers should feed into, the Shopify profit dashboard lays them out on one screen.
The best AI question ends with a decision. The worst ends with a chart you admire and close.
How do you get your data clean enough for AI to be right?
Boring work. One afternoon. It pays every time you ask a question after.
Fill in cost per item on every variant, not just the parent product. Rebuild those costs when freight, duty or a supplier price moves. Record your real shipping cost per order and your payment fee rate somewhere the AI can see them. Get ad spend split by product, either from campaign naming you control or from a profit app, because a single "Meta spend" line can't tell the serum from the eye balm.
Then test the AI against one answer you already know. Pick last month's top product. Work its contribution by hand. Ask the AI the same question. If the two match, keep going. If they don't, you just found the hole before it cost you anything.
Two more traps worth a sentence each. Refunds: make sure the AI is reading net sales, not gross. And attribution: Shopify and your ad platforms count conversions differently, which is why Shopify Analytics and Google Analytics disagree on the same day. Pick one source of truth per number and tell the AI which one.
The deeper method, with cohorts and payback windows, lives in the ecommerce profit analytics guide. And if you want to see the cost stack the AI should be subtracting, the contribution margin breakdown walks it from CM1 to CM3.
What should you do when the AI finds a loser?
Don't kill it on reflex. Ask one more question first: is the product weak, or is the page weak?
Cutting the eye balm's ads stops the $1,600 bleed. Fixing the page can flip it. If that same page moved from 2.0% to 3.0% conversion rate and average order value went from $36 to $44 through a two-pack, revenue per visitor would be 3.0% x $44 = $1.32. On the same 12,500 visitors, that's $16,500 instead of $9,000, from the same $5,200 in ads. Still hypothetical. But it's the right question to ask before you delete anything.
Here's a real one. On a copper bamboo sheets product, conversion rate went from 1.0% to 4.3% and average order value from $125 to $221. Revenue per visitor moved from $1.25 to $9.50, which on 10,000 visitors is $95,000 instead of $12,500. See the full case study numbers. Real client numbers, not typical results, and not a promise of what your store will do.
No new ad budget. The page did the work.
What to do next
This week, ask your store one question: "Which products have no cost per item on one or more variants?" Fill every blank it finds. Then ask for contribution after ads by product, sorted lowest first, and check the bottom row by hand.
Book Your Profit Audit
If the AI just told you a product loses money after ads, the fix usually lives on the page, not in the prompt. On a free profit audit we run the numbers on your hero products, find the page dragging revenue per visitor down, and show you how to rebuild a high-converting product sales page in less than 15 minutes.
Or go here to check it out → revenueflows.ai
P.S. AI answers whatever you ask in ten seconds. Clean data decides whether the answer prints money or bleeds it.
Frequently asked questions
Can Shopify's AI answer questions about my store data?
Yes. Shopify's analytics editor works with Sidekick, so you can type a business question in plain language and get a report back, with follow-up questions and an explanation of what each report measures. Shopify says Sidekick is included on all plans and only sees what the logged-in user can see.
Can I use ChatGPT to analyze my Shopify data?
You can export a report as a CSV and upload it to a chat assistant, and for a one-off question that's often the fastest path. The catches: each Shopify export covers one report type, exports carry summary rows the model may count as orders, and the file is stale the day after you pull it.
Can AI tell me which products lose money after ads?
Only if three things are in the data: cost per item on every variant, your real shipping and payment costs per order, and ad spend split by product. Shopify's profit reports skip any product without a recorded cost, and ad spend lives in Meta and Google, so an AI reading Shopify alone will call some losers winners.
Is AI data analysis for ecommerce accurate?
The arithmetic is usually fine. The inputs are where it breaks. Blank costs, refunds counted twice and ad spend left out produce confident, wrong answers, so check every AI answer once by hand before you act on it.
What's the best first question to ask AI about my store?
Ask for contribution after ads by product for the last 30 days, sorted lowest first. That one question finds the product your ads are quietly paying people to buy.