Personalization vs. “People Who Bought This Also Bought”

Two different jobs in the recommendation layer

“People Who Bought This Also Bought” is a recommendation pattern. Personalization is the broader practice of changing product discovery, merchandising, content, or offers according to a shopper’s context and behavior.

An also-bought module usually answers: What products are commonly purchased with this product? Personalization can answer more specific questions, such as:

• Which products should this returning customer see first?
• Which accessory fits the item currently in the cart?
• Which category should a shopper who repeatedly views running shoes explore?
• Which promotion is appropriate for a wholesale customer rather than a retail customer?

For WooCommerce agencies, the distinction matters because the two approaches require different data, rules, placement decisions, and success metrics.

How “People Who Bought This Also Bought” works

An also-bought recommendation is generally based on order relationships. If customers frequently purchase Product B in orders that contain Product A, Product B becomes a candidate when Product A is viewed or added to the cart.

A simple association score might be calculated as:

association(A, B) = orders_containing_A_and_B / orders_containing_A

A production system may add minimum order thresholds, recency weighting, category restrictions, product availability, and margin rules. It should also exclude the current product, out-of-stock products, hidden products, and unsuitable variations.

For example, if 180 of 1,000 orders containing a stainless-steel travel mug also contain a replacement lid, the replacement lid has an association rate of 18 percent for that product. That relationship can support a product-page module or a cart cross-sell.

This pattern does not necessarily know who the shopper is. It can show the same recommendation to every visitor viewing the mug, unless additional rules modify the result.

What personalization adds

Personalization uses shopper, session, account, and catalog context to select or rank recommendations. Useful signals can include:

• Recently viewed products and categories
• Search terms and filters used in the current session
• Products already purchased
• Customer role, membership level, or business account
• Geography, currency, device, or fulfillment region
• Inventory, product margin, seasonality, and campaign eligibility
• Cart contents and the intended use of the product

Consider a WooCommerce store selling commercial coffee equipment. A generic also-bought widget on an espresso machine might show a tamper, cleaning tablets, and a milk pitcher. A personalized module could prioritize a replacement gasket for a returning customer who already bought the machine, a bulk pack of filters for a café account, or a beginner bundle for a first-time visitor browsing entry-level equipment.

Recommendation logic compared

Use both instead of choosing one

Most stores benefit from a layered strategy. Use association-based recommendations where purchase relationships are strong, then personalize the ranking or fallback set.

1. Generate candidates: retrieve products frequently bought with the viewed product or current cart contents.
2. Apply eligibility rules: remove unavailable, restricted, discontinued, or already purchased products where appropriate.
3. Add contextual candidates: include products from recently viewed categories, replenishment lists, or customer-specific assortments.
4. Rank the final set: balance relevance, stock status, margin, merchandising priorities, and diversity.
5. Cache safely: cache product-level association results, but keep customer-specific ranking and eligibility separate when it can change per shopper.

For example, a cart containing a camera body could first retrieve lenses and batteries commonly bought with that model. The system could then prioritize a lens compatible with the shopper’s selected mount, suppress a battery already purchased by that customer, and promote an in-stock bundle that can ship with the cart.

Placement changes the meaning

The same recommendation can perform differently depending on its location:

• Product page: show complementary products without distracting from the primary purchase. “Often bought with” is appropriate here.
• Cart page: focus on low-friction add-ons that are compatible with the current cart.
• Post-purchase page or email: recommend replenishment items, accessories, or products that extend the original purchase.
• Account area: personalize by purchase history, subscription status, or customer-specific pricing.
• Category and search pages: use behavioral signals and merchandising rules to reorder products, while preserving filters and explicit sort choices.

Avoid placing a generic also-bought module on every page. A recommendation that is useful on a product page may be irrelevant in the cart, where compatibility and delivery constraints are more important.

WooCommerce implementation considerations

WooCommerce agencies should define the data model before selecting an extension or building custom logic. Decide whether recommendations are based on completed orders, paid orders, or all created orders. Exclude refunded, cancelled, and failed orders unless there is a deliberate reason to retain their signals.

For variable products, determine whether relationships are stored at the parent-product level or variation level. A store selling clothing may want a shirt color to share recommendations with the parent product, while a store selling technical components may need variation-specific compatibility.

Use stable product and variation identifiers, and account for products that are merged, replaced, or moved between catalogs. Recommendation jobs should be asynchronous for larger stores rather than recalculating all relationships during a customer request. Store precomputed results in a cache or dedicated table, then invalidate or refresh them when order data, inventory, or product visibility changes.

Personalized output must also respect WooCommerce rules for stock, catalog visibility, price display, tax settings, shipping zones, memberships, and customer-specific access. A product should not appear merely because its association score is high if the current customer cannot buy it.

Measurement and testing

Track impressions separately from clicks and purchases. At minimum, measure:

• Recommendation impression rate
• Click-through rate by placement
• Add-to-cart rate from recommendation clicks
• Revenue per session or per exposed order
• Average order value and item count
• Attach rate for the recommended product
• Conversion rate after exposure, including purchases without a click

Compare a product-level also-bought baseline with personalized ranking rather than comparing two unrelated placements. Use a controlled experiment where possible, keep the placement and visual treatment consistent, and segment results by new versus returning customers, device, category, and order value.

For stores with limited order volume, avoid overfitting. Set minimum support thresholds, use category or attribute fallbacks, and prefer a stable popular-products list over a recommendation based on one or two coincidental orders.

Privacy and data governance

Purchase-based recommendations can often be generated from aggregated order relationships without exposing one customer’s identity or order history to another. Do not display messages that reveal sensitive purchases, and do not use account-specific signals in a way that is visible through shared devices or cached pages.

Be careful with full-page caching. A customer-specific recommendation rendered into a cacheable product page can be shown to another visitor. Keep personalized fragments uncached, vary the response by an appropriate customer or session identifier, or render the recommendation after the page loads. Document the data used, retention period, and opt-out behavior for behavioral personalization.

A practical decision rule

Use “People Who Bought This Also Bought” when the store has enough reliable order data and the goal is to surface complementary products for a specific item. Add personalization when customer context, compatibility, lifecycle stage, or business rules materially change what should be shown.

For a first implementation, launch an aggregated also-bought baseline, establish eligibility and fallback rules, then test personalized ranking against that baseline. This gives the agency a measurable control while allowing customer-specific logic to improve relevance without replacing a proven source of product associations.