Frequently Bought Together

FBT · Customers Also Bought · Bought Together

Definition

FBT is a recommendation algorithm that suggests products based on an analysis of what other shoppers bought together inside a single order.

How the FBT algorithm works

Frequently Bought Together is built on association rule analysis: the system looks for pairs and
triples of products that show up in the same order most often. Two concepts from association rule
mining carry the work:

Support: how often the products are bought together relative to all orders
Confidence: the probability of buying product B given that product A was bought

Support(A→B)    = N(orders containing A and B) / N(all orders)
Confidence(A→B) = N(orders containing A and B) / N(orders containing A)

Modern implementations use item embeddings and neural co-purchase models, which handle sparse
matrices better than classic Apriori.

FBT vs other recommendation blocks

Block	Logic	Business goal
FBT	Co-purchase inside orders	Cross-sell, higher AOV
Similar items	Similarity by attributes or behaviour	Substitution, fewer bounces
Recently viewed	The shopper’s own history	Navigation, return to an earlier interest
Personalized recommendations	The shopper’s profile	Discovery, higher CR

FBT is the better choice for cross-sell because it rests on real purchase decisions rather than on
attribute similarity. If buyers of a cordless drill consistently pick up drill bits from one
particular brand, the algorithm will find that on its own, without a manual rule.

Where to place an FBT block

On the PDP: below the add-to-cart button, or inside a “complete the set” section — this is the
standard scenario. The shopper has not made the final call yet, so the extra products read as part
of the same decision.

In the cart: the highest-converting spot for FBT. The shopper has already decided to buy the
main item, which is the ideal moment for a “take this with it” prompt. Show a “frequently bought
with this” strip above the checkout button.

Important: do not overload an FBT block with items. Two to four products work better than
eight to ten — the shopper decides faster.

Cold start and fallback strategies

New products with no purchase history cannot take part in a co-purchase algorithm. The options are:

Category-based: show popular products from compatible categories
Merchandising rules: manually configure “for this product, show these”
Content-based similarity: products with matching attributes (brand, size, compatibility)

Once enough transactions accumulate — usually 50 to 100 co-purchases — the product moves onto
algorithmic recommendations automatically.

Frequently asked questions

How is FBT different from similar items?

Similar items are interchangeable — another laptop instead of this one. FBT is complementary — a case and a mouse to go with the laptop. FBT raises AOV through cross-sell, while similar items work through substitution, for out-of-stock situations or when the shopper is hunting for an alternative.

Does FBT need a lot of data to work?

Yes — the algorithm needs a statistically meaningful number of co-purchases. Below roughly 100 transactions per product the output is unreliable. The fix for new products (cold start) is to fall back on category compatibility or merchandising rules.

Where does an FBT block work best?

The two strongest placements are the product detail page, close to the add-to-cart button, and the cart, right before checkout. FBT converts better in the cart because the shopper is already in buying mode, so the add-on feels natural.

How do you tell FBT apart from other cross-sell blocks?

FBT is strictly based on co-purchase inside real orders. That separates it from editorial or rule-based blocks such as “Recommended with this product” and from co-view blocks such as “Often viewed together”. The method behind it is association rule mining (Apriori) or a more modern embedding-based co-purchase similarity model.

Gravity Field