Skip to content

Can AI Predict What My Shop Will Sell?

October 28, 2026

You overstock some items and run out of others. Both cost you money. Extra stock ties up cash and shelf space. Missing stock sends customers elsewhere.

You want to know if AI can fix that. It can help, with limits. This post covers what it does, where it fails, and how to try it without buying anything.

What “forecasting” means

A forecast is a guess about how much of something you will sell next week or next month. You already make these guesses. You look at what sold last time, think about the season, and place an order.

AI does the same thing with more history and less guessing. It reads your past sales, finds the pattern, and projects it forward. It needs no special insight. It needs your numbers.

What it can do well

Steady sellers. If a product sells most weeks, AI can estimate how much you will need. This is the easiest case.

Seasonal swings. If you sell more candles in November every year, a tool that sees two or more years of sales will expect it. A single year of data is thin, because the tool can’t tell a season from a one-off.

Many items at once. You can eyeball ten products. You can’t eyeball 300. A tool can check all of them every week and flag the ones heading for trouble.

Reorder timing. A forecast is most useful when you pair it with your supplier’s lead time. More on that below.

What it can’t do

Predict things it can’t see. A local festival, a competitor’s sale, or a post that goes viral will not show up in your sales history until after it happens.

Forecast new products. With no history, there is nothing to learn from. You are guessing, and so is the tool.

Fix bad records. If your counts are wrong, the forecast will be wrong. A tool that reads “0 in stock” for an item sitting on your shelf will assume you sold nothing.

Handle rare items well. An item you sell twice a year is hard for anyone to predict. For those, a simple rule such as “always keep two” often works better.

Check your own numbers first

You don’t need software to see whether you have a forecasting problem. Pull the last four weeks of sales and your current stock for a few items. Divide the four-week total by four to get a weekly average. Then divide stock on hand by that average.

Here is an example with three made-up items:

Item Sold in 4 weeks Weekly average On hand Weeks of stock left
Candle A 48 12 60 5.0
Mug B 8 2 70 35.0
Tote C 36 9 9 1.0
Total 92 23 139 6.0

Mug B is overstocked. You have 35 weeks of it. Tote C is about to run out. Say your supplier takes three weeks to deliver. You will be without totes for roughly two weeks, even if you order today.

That is the whole idea. A forecasting tool does this math for every item, every week, and adjusts for seasons. You can do a rough version in a spreadsheet today.

How to start

  1. Export your sales. Most point-of-sale and online store systems can export sales as a spreadsheet. Get as much history as you have.
  2. Check the data. Look for gaps, returns entered as negative sales, and items that were out of stock for weeks. Those weeks understate real demand.
  3. Try what you already have. Some store and inventory tools now include forecasting or reorder suggestions. Look in your settings before you buy anything new. If a tool added AI features on its own, our post on software that added AI without asking covers what to check.
  4. Start with your top items. Pick the 10 products that matter most. Compare the tool’s suggestion to what you would have ordered.
  5. Keep your judgment. Treat the forecast as a second opinion. You know about the festival. The tool doesn’t.

Questions to ask before you pay for a tool

  • How much sales history does it need? Ask for the minimum.
  • What happens with new products and one-off spikes?
  • Can you see why it suggests an order, or does it only show a number?
  • Where does your sales data go, and can you delete it? Our post on where pasted customer information goes has more on that.
  • Can you cancel without losing your own records?

Some vendors publish large accuracy and savings numbers. Treat those as claims, not promises. Ask how they were measured and on what kind of shop.

Is your shop too small?

Probably not. The main requirement is sales history, not size. A shop with a year of clean records and a few dozen products has enough to try a basic forecast. If you are unsure, read is my business too small for AI.

The short answer

Yes, AI can predict demand for a small shop. It works best on steady sellers with clean records. It works poorly on new products and surprises. Start with the spreadsheet math above. If it shows real gaps, a tool can save you the weekly work.

If you want help checking your own setup, talk to us. We will tell you honestly whether a tool is worth it for your shop.

Want to talk it through?

Tell me about the task that takes up your team's time.

Newsletter

Plain-language notes on software and AI for small business

One short email when there is something worth sharing. No spam. Unsubscribe any time.