How Auto Suggest Chooses Which Products to Recommend

Auto Suggest fills your Frequently Bought Together section, upsells and cross-sells with products that fit the one you’re editing. This guide explains where those suggestions come from, why they’re ordered the way they are, and what to expect on a store that hasn’t sold much yet.

How It Works

Auto Suggest looks at two things and combines them into a single ranked list.

• What your shoppers actually buy together, taken from your own past orders.
• What each product is, taken from its name and description, so related items can be spotted even without sales behind them.
• Whether the product can actually be offered, so nothing unavailable reaches the list.

The result is a shortlist, best match first, that you can accept as-is or edit before saving.

What Your Order History Contributes

The first signal is your own sales. Auto Suggest reviews completed and processing orders from the last twelve months and notes which products turned up in the same order.

It doesn’t simply count pairs. A product that sells constantly would otherwise appear alongside everything, so Auto Suggest measures how much more often two products are bought together than they would be by chance. A quiet product that almost always sells with one particular item can outrank a bestseller that happens to be in a lot of baskets.

Two rules keep the results steady:

• A pair has to appear in at least two orders before it counts, so a single unusual basket doesn’t create a suggestion.
• Only completed and processing orders are counted. Cancelled, failed and refunded orders are left out.

This analysis runs quietly in the background once a day, in small batches, so it doesn’t slow your store down.

What the Product Itself Contributes

Order history can only describe products that have already sold. CartFlows Pro adds a second signal by reading each product’s name and description to understand what it actually is, then finding products that make sense next to it.

That means a newly added product can be suggested on its first day, before anyone has bought it. When both signals are available they’re blended, with the product’s meaning weighted a little more heavily than the sales pattern, and the strongest match leads the list.

Reading the Suggestions

Each suggestion carries a short label that tells you why it’s there.

• A match percentage shows how strong a suggestion is relative to the best one in that list. The top match is always 100%, so treat it as a ranking, not a probability.
• Popular in your store means the product came from your bestsellers rather than from a specific buying pattern.
• New means it’s one of your most recently added products.

A suggestion with no badge and no percentage is simply one where the numbers wouldn’t tell you anything useful.

Accept adds the product to your list and Reject drops it from the suggestions. Anything you accept becomes an ordinary entry that you can remove or reorder before saving, exactly as if you’d searched for it yourself.

Suggestions are kept for up to 24 hours so the panel opens instantly the next time. If you’ve changed your products since, use Regenerate all to build the list again from scratch.

When There Isn’t Enough History

Auto Suggest never leaves you with an empty list. If your orders don’t produce enough matches, it fills the remaining slots in this order:

1. Bestsellers in a sensible price range. For a Frequently Bought Together section these usually come from a different category, since add-ons tend to sit alongside a product rather than compete with it.

2. Your remaining bestsellers, if more slots are still open.

3. Your newest products, so even a store with no sales at all gets a usable list.

On a brand-new store you’ll mostly see the second and third groups, with badges rather than percentages. As orders come in, real buying patterns take their place automatically.

Price Rules

Auto Suggest keeps the money side sensible, and the rule depends on what you’re filling in.

• For Frequently Bought Together and cross-sells, catalogue-based suggestions are kept between a quarter and one and a half times the product’s price. An add-on that costs ten times the product it sits next to rarely converts.
• For upsells, every suggestion must be priced above the product and no more than three times it. An upsell has to be a genuine step up without being a leap.

Suggestions that come straight from your buying patterns aren’t price-filtered for Frequently Bought Together and cross-sells, because a real pattern in your own orders is the stronger signal.

What Never Gets Suggested

Some products are held back so shoppers never meet an option they can’t buy.

• Products that are unpublished, hidden from the catalogue, out of stock or not purchasable.
• Product types the section can’t handle, such as grouped or external products.
• Variations that still need a choice the section has no way to collect.
• The product you’re editing, and anything already in your list.

Frequently Asked Questions