The biggest barrier to working with AI isn’t technology – it’s the blank page. Below is a bank of questions you can copy word for word. They work because no analyst made them up – this is how real store owners actually ask.
You don’t reach for data every day at a fixed hour. You reach for it in four specific moments: when something feels off, when you’re weighing options, when you want to verify something – and when you need to know right now. These 10 questions are arranged exactly that way. Find the moment you’re in today and start there.
One note before you start: none of these questions requires knowing any metrics or the language of reports. You ask them the way you think. It’s the data that should adapt to the question – not the other way around.
Moment 1 · When something feels off
You’re hunting for where the problem is
-
“Last week feels weak. What could it be?”
The classic first question. A good answer first verifies whether the week really was weak (hunches can mislead), and only then starts looking for causes – in traffic, the funnel and campaigns.
at work: sales · purchase funnel · campaigns -
“Sales in [category] have been falling for several weeks. What changed?”
Narrowing it to a category is half the battle – the answer can compare periods, check prices, variant availability and whether the drop affects every channel or just one.
at work: products · categories · period comparisons -
“Why does GA4 show different numbers than my store?”
A question every owner asks eventually – usually suspecting that something is broken. An answer backed by data shows the scale of the discrepancy in your store and explains how much of it is normal (ad blockers, attribution) and how much needs attention.
at work: GA4 · orders · data health
Moment 2 · When you’re weighing options
You’re facing a decision and want to see the options
-
“I want to grow sales without raising the ad budget. Where is my potential?”
The best question in this category – because it reverses the direction of the search. Instead of “what to buy more of”, it asks “what do I already have and underuse”. The example below shows what such a conversation really looks like.
at work: product funnel · margins · exposure
Question 4 in practice · anonymized data, pet store
get_product_funnel · last 30 days · sort: views
I reviewed the purchase paths of the products with the most traffic. One case clearly stands out:
| Product | Views | To cart |
|---|---|---|
| Dry lamb food 2 kg | 854 | 85% |
| Training treats mix 400 g | 840 | 89% |
| XXL treats mix bundle | 810 | 2.5% |
The XXL bundle is viewed by as many people as the bestsellers – 810 views a month – but only 2.5% of them add it to the cart, versus 85–89% for the others. You already have the traffic to the page, and it’s paid for. Something on the page itself is blocking the decision: the price relative to the contents, the description, or the lack of a smaller variant. Those who do add it to the cart buy normally – the problem is purely on the product page. Want me to compare its price per kilogram with the individual products in the bundle?
-
“Which products are sitting unsold – and how much frozen cash is tied up in them?”
A sales report shows what’s selling. This question flips the mirror: it shows what’s just sitting there. Often that’s more money than a good month brings in.
at work: inventory · purchase costs · turnover -
“What do customers buy together? What’s missing from my bundles?”
Baskets reveal how customers think – and that’s knowledge you can use directly: bundles, recommendations, on-site exposure.
at work: baskets · product co-occurrence
Moment 3 · When you want to verify
You have a hypothesis – yours or someone else’s
-
“The agency says the campaign is working. Show me the numbers.”
It’s not about distrust – it’s that the campaign report comes from whoever runs the campaign. Looking at your own numbers (ROAS over time, cost per click, share of budget) takes five minutes and ends the guessing.
at work: campaigns · spend · ad revenue -
“I tell everyone I have loyal customers. How many actually come back within 90 days?”
One of the healthiest questions in e-commerce – because it tests a belief on which an entire strategy often rests. The answer can be surprising in either direction.
at work: customer history · repeat purchases · segments
Moment 4 · When you need it right now
Meeting in an hour, deadline, ASAP
-
“I have a meeting in an hour. Prepare the key numbers from the last month – and one thing they might ask about.”
The second part of this question makes the difference. You don’t just want numbers – you want to know where the weak spot is before someone else points it out.
at work: sales · trends · anomalies -
“I ran a promotion last week. What did it really bring in?”
“Really” is the key word – a good answer compares not just revenue, but how much of it the discounts ate up, and whether the promotion drew in new customers or simply gave a discount to people who would have bought anyway.
at work: coupons · discounts · new vs returning
None of these questions ends the conversation. “Expand on this”, “show it per channel”, “and how did it look a year ago?” – the model remembers the conversation context, so every answer is a springboard to the next question. The best analyses are dialogues, not single shots.
Takeaway
A good question for AI is the question you already have in your head.
There’s no “query language” you need to learn. The barrier that for years separated a store owner from their data was the need to translate thoughts into the language of reports – and that barrier has just disappeared.
These 10 questions work everywhere, by the way: you can put them to an analyst, an agency, or to yourself over a spreadsheet. With your data connected to the chat there’s just one difference – the answer arrives in a minute, so you’ll also have time to ask the eleventh, twelfth and thirteenth.
DataOrganizer · MCP
DataOrganizer connects your store’s data to your favorite chat. 20 minutes of setup – and you ask the way you think.