The right product. The right content.
Every time.

Software that decides, for each shopper, which products to show — on a product page, in the cart, in an email — from what shoppers like them viewed and bought. Personyze’s ecommerce recommendation engine runs from one tag on your site, with an algorithm for each page, your own rules, and the same picks in your emails and your app — and it learns from every visit.

Leave a widget on Automatic — the first card when you set one up, and where a new widget starts — and it chooses by itself on every page: where shoppers go next on a product page (Viewed Next), the best sellers on a category page, cross-sells for what is in the cart, then personalized picks, what is popular in the visitor’s country and what is trending, with the newest products last, so it is never empty. Or pick All algorithms beside it, or a page, to choose one yourself: predictive and popular ones, crowd wisdom around the product on screen (Viewed Next, Cross-Sells, Similar Items (AI)), the visitor’s own items, what just changed in your catalog, and your own rules.

Yes. Personyze reads your catalog, works out what kind of shop it is, and tunes the algorithms to it. In fashion it completes the look: a black hoodie gets matching track pants and sneakers, never twelve more hoodies. In electronics it shows the accessories that fit this exact product, so after a TV come the soundbar and the wall mount. In home and furniture it completes the room in the same style; in beauty, grocery and pet stores it brings a product back when a shopper usually runs out; in sports it keeps to the same activity; and on a blog it leads to the next step on the same topic. It also reads what each product looks like, so a new arrival gets good recommendations the day it is added. Nothing to configure.

Yes — any website, whatever it is built on: one tag on any platform or custom build, with or without a store connection. Your catalog and your orders then come in four ways. Your store: Shopify, Magento, WooCommerce, BigCommerce or Shopware, connected once by API with your own keys. Products, prices and stock, past orders and customers stay up to date on their own, and it only reads. Your feed or files: The feed you already give Google Shopping or Facebook at its address, a file you upload, files your system drops by SFTP, or the REST API — in CSV, Excel, XML or JSON. We read your site: No feed at all? We visit your product pages and read each product’s name, picture and price. Your orders, online and in-store: Upload past orders — one row per order line — and the engine starts from what already sold together instead of waiting weeks. A recurring feed or SFTP keeps it learning from every sale, in-store included.

As a slider, a grid, a list, a bought-together strip or tabs — under the product, in the cart, on the home page, before, after or inside any element, or as a popup in the centre, a corner or on exit intent. Shoppers add to cart from the widget without leaving the page, and the slider and grid set their own count of cards for desktop, tablet and phone.

Your way: ask the AI assistant and it builds the widget, matches your site’s look, places it and checks it on your page; or point and click the spot on your page in the visual editor; or start from the template gallery and describe any change for AI to make. Cards can carry badges, an add-to-cart button and a wishlist heart, and developers can edit the HTML.

Yes. Skip what the visitor is viewing now, has viewed, has in the cart, has saved or has bought; never recommend chosen categories; keep only products on sale, with an image or from chosen categories; write a filter on any catalog field; show one per brand; and nudge the order with gentle boosts — more in stock first, new items, profit margin. Out-of-stock products are never recommended.

Similar Items (AI) reads your titles, brands, descriptions and categories, so a new product has similar items before anyone has viewed it. It is switched on once for the whole account in your catalog settings and uses AI credits. New in Stock shows what arrived since the visitor’s last visit, a widget can boost new items or keep a slot for them, and fallbacks fill a short list.

Yes — it is self-learning. It weighs many signals at once: who the shopper is, what they view, add to cart, save, search for and buy, and where they are. Every round goes into the next build of the lists, clicks lift what works within limits, and an A/B test can hand the winner all the traffic.

Per widget and per product: times shown (counted only when at least two thirds of the item was on screen), clicks and the click rate, and purchases — a purchase counts when it follows a click on that item in that widget within 7 days. Revenue is credited per widget, priced from your catalog, and items shown a lot but rarely clicked are flagged for you.

Yes. Put one widget against another — two algorithms, two designs, or a widget against none. New tests ask for 98% confidence and 1,000 sessions before a winner is called, and with auto-pick on, the winner can get all the tested traffic. More on A/B testing.

Yes. The widget translates its own words — titles, the add to cart button, labels — into each shopper’s language automatically, and your catalog feed can carry several languages: set a language and map which feed fields go with it, so names and descriptions show in the shopper’s language too.

Yes. Native product rails in your iOS and Android apps come through our SDK — Product Recs for Apps, the same engine and the same picks as on the web. For anything else — your back end, a kiosk, an email template — the JSON API returns a visitor’s recommended products from one address, by user id, email or the item being viewed.

One HTML snippet goes into the emails you already send, from Mailchimp, Klaviyo or any other platform, and each recipient gets products picked for them at the moment they open it, from what is in stock then. More on open-time emails.

Yes. Articles have algorithms of their own — what readers open next, what readers of this article also read, the most popular in its category, trending reads, content recommended for each reader and more — for publishers, knowledge bases and media sites. See content recommendations.

Yes — it’s one engine shared across the platform. The same recommendations can render inside open-time emails sent through your existing email platform, the visitor signals behind them come from behavioral targeting, and you can A/B test algorithms and placements against each other. One tag, one visitor profile, every channel.