Related Products and Recommendations That Actually Relate
Recommendations help when the relationship is real: completes the job, fits the item, replaces it, or serves the same need. Random or purely promotional suggestions teach visitors to ignore the whole system.
The short answer
A recommendation earns its place when you can finish the sentence: this is shown because it completes, fits, replaces or genuinely matches what the visitor is looking at. Accessories that work with the item, refills for it, parts for it, the same need in another form, and products bought for the same job can all relate. A carousel of unrelated bestsellers, or items chosen only because the margin is higher, teaches visitors that recommendations are advertising in disguise, and they will start ignoring the useful ones too. Relevance is the whole asset. Protect it.
Name the kind of relationship
Different relationships belong in different places. Completes the job belongs beside the main product: the case for the instrument, the charge cable, the correct fitting. Fits or works with belongs where compatibility is decided, with the compatibility stated plainly. Alternatives belong where a visitor is comparing, labelled as alternatives rather than upgrades by default. More of the same need belongs on category and search pages. Mixing all kinds into one undifferentiated row forces the visitor to work out why each item is there. Label the row by its relationship, in plain words, so the logic is visible at a glance.
Rules beat vague automation for small catalogues
Large stores train recommendation systems on volume a small store does not have. For a small catalogue, hand-set rules usually win: product A needs part B, range X fits model Y, these three are the same job at different sizes. Set them at category or product-family level where possible so maintenance stays realistic, and review them when products change. Automated also-bought suggestions can supplement rules where your platform provides them, but check the output by hand. Automation that once suggested a sensible pairing can drift into nonsense as stock changes, and no one notices until a customer does.
Place recommendations after the main decision
On a product page, the first job is the product itself. Recommendations that interrupt before price, fit and availability are clear compete with the decision the page exists to support. Place completes-the-job items where they help the decision, such as compatible accessories near the buying options, and place broader suggestions lower, after the main details, for the visitor who is still exploring. In the cart, keep recommendations extremely restrained; the visitor has decided, and the checkout cluster explains why new decisions at that point are costly. Timing is what separates help from distraction.
Be honest about why something is shown
If a placement is paid, sponsored or part of a supplier arrangement, label it as such where the law and your own honesty standard require it, and never let the arrangement borrow the language of personal recommendation. If an item is shown because it is the higher-priced version, call it what it is and state the real difference. Visitors accept a clear upgrade path. They dislike discovering that a recommendation was a margin decision dressed as advice. This cluster connects to the wider proof rules in trust and social proof: advice is a trust surface.
Internal recommendations include editorial links
Not every recommendation is a product row. A link from a product to its sizing guide, care guide, compatibility list or the category that helps choose is often more valuable than another product. Likewise, a category page linking to the guide that explains the choice serves the visitor better than a fifth carousel. Build these links from the questions support receives. When you retire a product, check its incoming recommendations and links so nothing points to a dead end or, worse, quietly substitutes a different product as if nothing had changed.
Review the output like a customer
Once a month, open a sample of product pages as a customer would and apply the one-sentence test to every suggested item. Remove or fix any suggestion you cannot defend. Watch also for practical failures: recommended items out of stock, incompatible variants shown as fitting, and the same item recommended back to itself through a bundle or collection loop. Measure gently: clicks on a row mean little if they replace the main purchase with confusion. The healthy sign is visitors using recommendations to complete a purchase or to find a better fit, then proceeding without extra searching.
Sources and evidence on the record
- Baymard Institute, baymard.com - large-sample usability research on ecommerce product pages, carts and checkout.
- CXL, CXL blog - practitioner writing on testing method, research process and optimisation programme discipline.
- US Federal Trade Commission, advertising and marketing guidance - the honesty rules for claims, reviews, endorsements and disclosures.
Evidence basis: Method guide. Any figure quoted is attributed to a named published source listed below; worked examples are labelled illustrations with round numbers, not measurements. Applies when: Your product pages or categories show related, recommended or also-bought items