Using AI to spot slow and fast sellers

Ask the tool to rank your products from fastest to slowest, then look hard at both ends of the list. A ranking is only as sound as the sales period behind it.

What "fast" and "slow" mean

Retailers often measure speed with inventory turnover, the number of times inventory is sold or used in a time period such as a year.¹ A product that turns over often is a fast seller. One that sits is slow.

Context changes the meaning. Wikipedia points out that only intra-industry comparison will be appropriate, because a supermarket sells fast-moving goods and a car dealer sells slow-moving ones.¹ Compare your products with each other, not with a figure from another trade.

Sorting into groups

An older method, ABC analysis, groups stock by importance. It starts from the idea that inventories are not of equal value.² The page adds that there are no fixed thresholds for each class, and that proportions vary between companies.² So where you draw the line is your choice. It can also be done in Excel without special software.²

A simple three-way split works for a small shop: fast, middle, slow. Ask the tool to put each product in a group and to say what rule it used. If it can't state the rule, don't accept the groups.

A way to ask

Paste a table with product, code and weekly units sold. Then write something like:

Rank these products by average weekly units sold. Show the average for each. Mark the top fifth as fast and the bottom fifth as slow. Tell me which products have fewer than eight weeks of history.

Microsoft says Copilot in Excel can apply sorting, filtering and conditional formatting to focus on the data that matters most.³ Either way, ask for the averages, so you can check them.

Where the ranking misleads

Work through these before acting on it:

• Stockouts. A product that was empty for weeks sold little because there was none. It may be a fast seller in disguise.
• One-off sales. A single large order can make a slow product look fast. An outlier is a data point that differs significantly from other observations.⁴ Look at the biggest week for each fast seller.
• New products. Few weeks of data means a weak ranking. Leave them out of the first pass.
• Price and promotions. A discount can lift a product for a short stretch.
• Season. A winter item looks slow in summer.
• Value. A slow seller with a high price can still matter. Fast and valuable aren't the same thing.

Checks before you act

Take the five fastest and five slowest products. For each, look at the weekly sales yourself. Do the numbers fit what you see in the shop? If a "slow" product has a loyal handful of buyers, you may still want to keep it.

Then decide what to do with the ranking. Common uses are reordering the fast sellers more often, holding less of the slow ones, and checking the slow ones for damage or age. Any cut to a range is your decision. Don't let the tool delist anything. A hand-count routine, set out here, shows why the shelf check comes first.

Rerun the ranking once a month or once a season, with the same prompt, and compare. If the same products jump around, the data is too thin or too noisy to rank. The forecast failures page lists more causes.