E-commerce

Cross-sell

A cross-sell suggests a related item alongside what somebody is already buying. Done well it is helpful, and done badly it is a distraction at the worst moment.

Also called cross-selling, related products, frequently bought together

SiiteWritten by SiiteUpdated September 7, 2026

A cross-sell is a suggestion for something that goes with what somebody is already buying. The case beside the phone, the filter beside the coffee machine, the delivery service beside the furniture. It is one of the oldest ideas in retail and one of the easiest to implement badly on a website.

In short

  • It suggests a companion item, not a better version of the same one.
  • It belongs after the main choice is made, not during it.
  • Relevance is the whole thing. Irrelevant suggestions read as noise.
  • It has to be measured on completed orders, not on order value alone.

Cross-sell, upsell, and why the difference matters

The two words are used interchangeably and describe opposite moments.

  • Upsell. A better or larger version of the thing being considered. More storage, the next model, the bigger size. It arrives while the customer is still deciding what to take.
  • Cross-sell. A different item that goes with the one chosen. It arrives once the decision is settled and the customer knows what they are holding.

Getting the order wrong is the most common fault. A cross-sell shown too early competes with the item somebody was choosing, so instead of adding a second purchase it splits the attention on the first. An upsell shown too late arrives after the customer has committed and reads as an attempt to change a decision they had finished making.

Where the suggestions belong

Three places carry them and each does a different job.

  • The product page. Suggestions that help somebody choose, including the things they would have to buy anyway for the item to be useful.
  • The basket. Suggestions that complete an order. This is the strongest position, because the choice is made and the customer is looking at what they have.
  • After the purchase. In the confirmation or a follow up message. Nothing is at risk here, and a customer who has just bought is a customer paying attention.

One place is usually a mistake, which is inside the payment step itself. A person entering delivery details has stopped shopping. Anything new at that point is an interruption to a transaction you have already earned, and the sale at risk is larger than the one being offered.

What makes a suggestion useful

The distinction between helpful and irritating is not subtlety. It is whether the suggestion answers a question the customer already had.

Good cross-sells tend to fall into three groups. Things the item needs in order to work, which the customer would otherwise discover at home. Things that are used together in practice, which the shop knows from its own orders. And things that solve the obvious next problem, such as storage for something bulky or care for something delicate.

Bad ones share one property. They were generated by a rule nobody checked. A category match, a price band, a tag applied at import. Customers read those instantly, and the cost is not only the ignored suggestion. A shop that recommends something obviously unrelated has told the customer that its recommendations mean nothing, which spends the credibility of every future one.

Measuring it honestly

Cross-selling is easy to declare a success because the number it moves most visibly is the flattering one.

Average order value rises whenever anybody adds anything, so a suggestion block will almost always show an improvement there. The number that decides whether it paid is the share of started checkouts that finish. Suggestions that distract, slow the page or add a decision push that share down, and a shop can raise the value of each order while ending the month with less money.

Read three figures together. What an average order comes to, how many orders were completed, and what the two multiply to. Then look at whether the suggested items are the ones being added, or whether the rise came from somewhere else entirely and the block is taking credit for it.

Choosing suggestions before you have data

Recommendation software learns from order history, which means a new shop has nothing for it to learn from. Turning it on early produces suggestions that look arbitrary because they are, and customers read that immediately.

Until there is enough history, the suggestions should be written by hand by somebody who knows the products. Three questions cover most of it.

  • What does this item need in order to be used? Batteries, a filter, a mounting kit, a cable. These are the strongest suggestions available, because the customer will need them anyway and would rather not order twice.
  • What do people who buy this already own or lack? A shop owner usually knows this from conversations long before the data would show it.
  • What is the obvious next problem? Storage for something bulky, care for something delicate, a replacement for something that wears out.

Write those as fixed pairings, review them when the season or the range changes, and switch to the automated version once there is a year of orders behind it. Rules written by somebody who understands the products beat a system with nothing to go on, and they stop being the better option the moment there is real history to learn from.

Words you will hear

  • Upsell. A better version of the same item.
  • Bundle. Two or more items sold together as one purchase, often at a set price for the group.
  • Frequently bought together. A suggestion built from what previous customers actually ordered in the same basket.
  • Attachment rate. The share of orders that include an accompanying item.
  • Basket size. How many items are in a typical order, as distinct from what it is worth.

Cross-selling is the cheapest way to raise what an order is worth, because the customer is already there and already convinced. Keep the suggestions few and genuinely related, keep them out of the payment step, and judge them on completed orders and on customer lifetime value rather than on the size of a single basket.

Questions we get

More about cross-sell

What is the difference between a cross-sell and an upsell?

A cross-sell offers something that goes with the item, such as a case for a phone. An upsell offers a better version of the same item, such as the model with more storage. They sit in different places for a reason. An upsell belongs before the choice is settled, and a cross-sell belongs after it, once the customer knows what they are taking.

Where should the suggestions appear?

The product page and the basket carry most of them well. The product page suits suggestions that help somebody choose, and the basket suits the things that complete an order. Suggestions inside the payment step usually cost more than they earn, because a customer who has started paying has finished deciding and anything new is an interruption.

How many should we show?

Few enough that each one is a considered suggestion. Three or four related items read as a recommendation, and a grid of twenty reads as a catalogue somebody gave up organising. The number matters less than the relevance, but a long list makes irrelevance more likely and gives the customer another decision when they had already made one.

Does automatic recommendation software do this well?

It depends entirely on how much sales history it has to learn from. A shop with years of orders can find genuine pairings nobody would have guessed. A newer shop produces suggestions that are close to random, which customers notice. Until there is enough history, rules written by somebody who knows the products beat the automated version.

Will it slow down or complicate the checkout?

It can, and that is the risk worth respecting. Every extra element on the way to payment is another thing to load, read and dismiss on a phone. If a suggestion block sits between the basket and the payment button, it is competing with the sale you already have, and the sale you already have is worth more than the one you might add.

How do we know whether it worked?

Look at what the shop earns per order, at how many orders contain more than one item, and at whether the checkout completion rate moved. A rise in average order value alongside a fall in completed orders is a loss dressed as a win. All three numbers have to be read together or the change cannot be judged.

Does this work with cash on delivery?

It does, with one caution. A larger order paid on delivery is a larger amount to be refused at the door, so the suggestions that work best are the small, obviously related ones rather than an expensive second item. Anything that turns a modest order into a big commitment is worth watching for a rise in refused deliveries.

Do it yourself

Guides on the same topic

Free step by step guides for owners doing the work themselves. No sign up and no email required.

Want this handled for you?

Let us take cross-sell off your desk.

The guides and these pages are yours to use for nothing. When you would rather have the work done properly than done by you, tell us what is already in place and we will put a proposal together.