How cross-sell differs from upsell
Both tools aim at a higher average order value, but they work differently.
| Cross-sell | Upsell | |
|---|---|---|
| What is offered | A product from a different category | A more expensive version of the same product |
| Example | Smartphone → case + headphones | Smartphone 128 GB → smartphone 256 GB |
| Timing | Before payment or right after | Before payment |
| Effect | Higher AOV through a wider basket | Higher AOV through the price of the main item |
In practice the two are usually used together — upsell on the product page (similar items at a
higher price), cross-sell in the cart (complementary products).
Where cross-sell works
The product page — a “frequently bought with this” block. The shopper has not made the final
call yet, so an addition to the main choice reads as part of the same decision.
The cart — the most effective spot: purchase intent is at its peak and the decision to buy is
already made. Cross-sell here works as a “you forgot to take…” prompt.
The order confirmation page — the post-purchase moment. Conversion is lower, but there are no
objections left: the money is spent and the offer reads as helpfulness.
The order confirmation email — through a recommendation block embedded into the email your own
ESP sends. Worth doing for products with a consumption cycle: consumables and accessories.
Algorithms for cross-sell
The classic is Frequently Bought Together: transaction analysis surfaces the pairs and triples
of products that get bought together. It is simple, transparent and works well for FMCG and
electronics.
For fashion there is Complete the Look: an algorithm trained on stylistic combinations. It
offers not just “often bought together” but “goes with this outfit”.
Collaborative filtering performs worse for cross-sell — it optimises similarity of preference,
not complementarity of products.
Tip: test algorithms through an A/B test with AOV and attach rate as metrics, not CTR alone —
a click does not guarantee that the cross-sell item is bought.
Common mistakes
- Offering too many items — three to five products is optimal; a long list dilutes attention.
- Offering incompatible items — a case for a different phone model, or accessories without a
category compatibility check. - Ignoring cart context — if the cross-sell item is already in the cart, do not show it again.
- Judging by CTR alone — measure AOV and attach rate, or you will end up optimising
clickability instead of the actual economics.