Which Customer Data Should Drive Magento Personalisation?
Magento personalisation should start with what the shopper is doing now, then add catalogue, stock, purchase and performance data in that order. A live basket usually tells you more than an old customer label. Use history to refine the choice, store rules to keep it sensible, and outcome data to see whether the result earned its pixels. “For you” is not a strategy. It is two small words.
What data does Magento personalisation actually use?
Useful Magento personalisation combines six types of data. They do different jobs, so tipping everything into one mysterious score is rarely the best starting point.
| Data type | Examples | Best use | Main risk |
|---|---|---|---|
| Catalogue data | SKU, category, attributes, price | Finding products that genuinely relate or fit | Broad categories create silly pairings |
| Availability data | Salable stock, low stock, end-of-line status | Removing products that cannot or should not be promoted | Cached results can outlive a stock change |
| Live session data | Current product, viewed category, basket, subtotal, coupon | Responding to the shopper's immediate job | Too many reactions make the page twitchy |
| Behaviour history | Product views, category views, previous orders | Reordering choices for a returning shopper | Old interest can be mistaken for current intent |
| Customer data | Guest or logged in, customer group, account details | Eligibility, pricing and targeted promotions | A broad segment can become a lazy substitute for relevance |
| Commercial data | Sales velocity, margin, conversion, campaign boosts | Balancing shopper relevance with store priorities | High margin is not the same thing as a good match |
Outcome data sits around all six. Impressions, clicks, adds, purchases, revenue and profit tell you which decisions helped. Without that feedback loop, personalisation is simply a collection of confident guesses wearing a dashboard.
How is personalisation data different from placement?
Placement decides where a recommendation appears. Data decides which product appears and why. Both matter, but they are separate decisions.
Our guide to product-page, cart and post-purchase offer placement covers the first question. This article covers the second: the signal hierarchy behind each choice. A perfectly placed widget can still recommend printer toner to someone buying a tent. The furniture is in the right room; the guest list needs work.
Which signal should win when the data disagrees?
Hard eligibility rules and current intent should beat historical signals. A shopper's basket today is normally more useful than something they viewed months ago.
- Start with eligibility. Remove products that are out of stock, already owned, already in the basket, excluded, incompatible or unavailable to that customer group.
- Read the current session. Use the product, category and live basket to understand the job the shopper is trying to finish.
- Check direct relationships. Attributes, model compatibility and well-maintained product links give the choice a factual reason to exist.
- Add recent behaviour. Views and purchases can reorder eligible products, but they should refine current intent rather than replace it.
- Apply store priorities. Margin, sales velocity and campaign boosts can break a tie between relevant products. They should not rescue an irrelevant one.
- Use popularity as a fallback. Bestsellers help when a new or anonymous shopper gives you little else to work with.
This order also gives you a useful debugging path. If a recommendation looks wrong, you can tell whether eligibility failed, context was missing, or a business weight shouted too loudly.
What does Magento handle by default?
Magento Open Source gives merchants manual product relationships, customer groups, catalogue price rules and cart price rules. Those are useful building blocks, but Magento does not automatically combine browsing, purchase, basket and commercial signals into a personalised product ranking.
Adobe Commerce goes further. Its Product Recommendations service combines behavioural and catalogue data, including product views, basket actions and purchases. Adobe Commerce also supports dynamic customer segments based on details such as order history and basket contents. Those services are not the same as the manual relationships included with Magento Open Source, so check which edition and services a recommendation guide assumes before copying its setup.
How does SmartListings use browsing, purchase and store data?
SmartListings gives Magento category ordering a layered data model. Availability comes first, merchandising boosts can steer a campaign, and the Pro plan can personalise category results from each visitor's browsing and purchase signals.
The behaviour settings are deliberately visible. Product views have a default weight of 1, purchases have a default weight of 5, and the behaviour contribution is capped at 25. Activity older than the default 90-day retention period is removed automatically. Merchants can use AJAX personalisation, which reorders products after the cached page loads, or server-side ranking at the SQL layer.
SmartListings can also blend in ProfitEasy sales velocity, profit velocity and conversion data on Pro. That is commercial data, not customer intent, so it belongs later in the score. A profitable product still has to be in stock and relevant. On product pages, the SmartListings Product Block can work through Magento's related items, upgrade options, cart add-ons and category siblings, then rank the useful group instead of presenting a fixed row forever.
How does SmartCart use live basket and order data?
SmartCart starts with the live basket. Rules can trigger from a SKU, category or attribute in the cart, while candidate products can come from a chosen list, category, attribute or co-purchase pattern found in completed orders.
Before anything appears, SmartCart filters out unavailable products, items already in the basket, excluded categories and products the shopper already bought when purchase-history suppression is enabled. Merchants can also restrict rules by customer group, minimum subtotal and coupon status. That is personalisation with a paper trail: each decision has a visible rule instead of disappearing into a black box.
Order-history mining runs offline and proposes draft rules for review. The storefront does not wait for an external AI call, and nothing mined has to go live automatically. That matters because a pair can be statistically common and still be a daft thing to suggest. Your catalogue knowledge gets the final vote.
What about wishlists, account data and dynamic content?
Wishlist and account signals can help, but they need a narrower job than “make the site personal”. A wishlist shows interest, not necessarily intent to buy today. A customer group is useful for eligibility and promotions, but it says little about whether two products belong together.
Use these signals to adjust an already sensible choice. A wishlist can nudge a saved item upward, a customer group can control access to an offer, and account or address data can select appropriate delivery or regional content where your consent and privacy setup allow it. None should overrule stock, compatibility or the shopper's current basket.
How should you measure Magento personalisation?
Measure the decision chain, not just the final order value. A recommendation can raise average order value among the people who click it while distracting everyone else.
- Impression to click rate tests whether the suggestion and its explanation look relevant.
- Click to add rate tests the product, price and landing experience.
- Add to purchase rate shows whether the extra item survives checkout.
- Overall conversion rate catches widgets that distract more shoppers than they help.
- Revenue and profit per rule separate busy recommendations from useful ones.
- Repeat purchase rate matters when purchase history is meant to support replenishment or retention.
SmartCart records 30-day impressions, add rate, purchase rate and revenue per rule. ProfitEasy can show attributed SmartCart and SmartListings revenue, while AnalyticsEasy can carry SmartListings attribution into GA4. Start with a baseline, change one signal or weight, and compare the full chain. Changing the ranking, widget, copy and discount together produces a result, but not an explanation.
A practical first personalisation test
Start with one busy category or one cart rule where the correct choices are easy to explain. You want a test that can fail clearly.
- Write down the shopper job in one sentence.
- Choose the strongest current-intent signal: viewed product, category or basket item.
- Add the stock, compatibility and prior-purchase exclusions.
- Choose one secondary signal, such as recent views, order history or a merchandising boost.
- Record impressions, adds, purchases, conversion, revenue and profit before changing the rule.
- Run the test until normal traffic produces a useful comparison, then keep, adjust or remove it.
For implementation steps, use our Magento personalisation and product-recommendations guide. For deeper cart logic, see the cart recommendation audit and the worked Magento recommendation-rule examples.
FAQ
What is Magento personalisation?
Magento personalisation changes products, offers or content according to useful shopper and store signals. Good personalisation combines current intent with catalogue relationships, availability, recent behaviour and clear business rules, then measures the outcome.
What customer data is most useful for product recommendations?
The current product and live basket are usually the strongest signals because they show what the shopper is trying to do now. Recent views and purchases can refine those choices. Stock, compatibility and prior-purchase checks should remove bad candidates before any ranking begins.
Does Magento Open Source personalise products automatically?
No. Magento Open Source includes manual product relationships, customer groups and pricing rules, but it does not automatically combine shopper behaviour and catalogue data into per-visitor product rankings. That needs custom work or an extension.
Do Magento recommendations need AI?
No. Rules based on product attributes, the live basket, stock and order history cover many useful cases. AI or machine learning can help find patterns when enough data exists, but it should propose choices that still pass your eligibility and relevance rules.
How do you keep Magento personalisation from feeling intrusive?
Use only the data needed for a clear customer benefit, keep retention sensible, respect consent choices and avoid revealing how much history you hold. A recommendation should feel like the next useful shelf, not like the shelf has been following someone home.
The useful question is not “How personal can we make this?” It is “Which signal makes this choice more helpful right now?” Start there, and let the data earn its place.
