Recommendation algorithms, tuned to your site.
A recommendation engine that combines crowd wisdom, predictive models, each visitor’s own history and AI that understands what every product and article is about, so each visitor sees what they are most likely to want next, in product recommendations and content recommendations. It adapts to your kind of site on its own: outfits in a fashion store, accessories that fit in electronics, the next read on a news site.
Not sure where to start? Get a free recommendations plan
4.8
4.6
ViewingSherpa Bomber, Navy$129
Merino Crew, Charcoal$79
Knit Beanie, Charcoal$45
Court Sneaker, White$74
ViewingStudio Headphones$129
Field Earbuds$59
Over-Ear Headphones$89
Canvas Backpack 25L$89Court Sneaker, White$74Over-Ear Headphones$89
Canvas Backpack$79Knit Beanie, Charcoal$45
Merino Watch Cap$39
Trail Runner Low$89
Weekender Duffel$119
Knit Beanie, Mustard$45NewCourt Sneaker, White$74
Court Sneaker, Olive$74
Court Sneaker, Tan$74Trail Runner Low$89Studio Headphones$129Court Sneaker, Tan$74Weekender Duffel$119Canvas Backpack 25L$89
Knit Beanie, Rust$45Knit Beanie, Charcoal$45Knit Beanie, Mustard$45Merino Watch Cap$39Court Sneaker, Olive$74Canvas Backpack$79Court Sneaker, White$74Court Sneaker, Tan$74
Product recommendation algorithms, tuned to what you sell.
Personyze reads your catalog, works out what kind of shop it is and adapts its algorithms to it. A fashion store gets outfits, an electronics store gets the accessories that fit, a pet store gets the refill before the bag runs out. Nothing to configure.
Complete the look, Similar style, In your style
Personyze reads each product’s picture and title once and records its look: the kind of item, colour, pattern, style, occasion, season and who it is for. It weighs that with what your shoppers bought and viewed, so every row reads as “this goes with that” or “more like this”.
- Complete the lookPieces of other kinds that go with the item on screen. A navy bomber gets a sweater, a beanie and sneakers, never twelve more jackets.
- Similar styleThe same kind of item in a matching style, for shoppers still choosing.
- In your styleAcross the site, items ranked by the looks this shopper has browsed and put in the cart.
See it on a store: a fashion site, personalized · personalization for fashion websites
Merino Crew, Charcoal$79Knit Beanie, Charcoal$45Court Sneaker, White$74
Chunky Wool Scarf$49
Hooded Down Parka, Navy$129
Insulated Parka, Navy$139
Alpine Parka, Moss$149
Alpine Parka, Slate$149
Merino Gloves, Charcoal$35
Winter Boots, Brown$139Merino Watch Cap$39
Cable Beanie, Oatmeal$42Accessories that fit
Cases, chargers and cables that work with this exact product, in the same brand ecosystem. After someone buys a TV, they see the soundbar and the wall mount, not more TVs.
- Made for this productOnly accessories that fit the model on screen or in the order.
- Alternatives at a similar priceOn a product page, other models stay in the same price band.
- After the purchaseWhat completes it, on the thank-you page and in the follow-up email.
The algorithm behind it: customers who bought this also bought · the benefits of bundle sales
Slim Soundbar$199Fits your TV
Tilting Wall Mount$49Fits your TV
HDMI Cable, 2 m$19Fits your TVComplete the room
A sofa gets a coffee table, a rug and a lamp in the same style, read from each product’s picture: the wood, the colours, the period. A shopper furnishing one room sees one room, not three styles at once.
- Other kinds, one styleThe pieces that complete the room, never another twelve sofas.
- Learned from real ordersWhat your shoppers furnished together, cleaned up by each piece’s look.
- Priced below the piece on screenAdd-ons cost less than the sofa they go with.
More on it: product recommendations, an example
ViewingEllis Sofa, Sage$1,190Its lookMid-century · walnut · sage
Round Coffee Table, Walnut$340Walnut, like its legs
Flat-weave Rug, Cream$260Sage and terracotta
Floor Lamp, Brass$180Walnut and brassTime to restock
Products come back to a shopper when they typically run out, learned from how often people rebuy each item. A serum, a bag of coffee and a sack of dog food each keep their own rhythm.
- Learned per productFrom how often your shoppers rebuy it, not one rule for the whole store.
- On the site and in emailOn the home page, in the cart and in your recommendation emails.
- Not before they need itUntil then, what they bought stays out of their recommendations.
In the library: Buy it Again · recommendations in email
BeautyDaily Face SerumBought 5 weeks ago · Rebought about every 6 weeksTime to restock
GroceryWhole Bean Coffee, 1 kgBought 13 days ago · Rebought about every 2 weeksTime to restock
PetsAdult Dog Food, 6 kgBought 4 weeks ago · Rebought about every 4 weeksTime to restockThe line: when shoppers usually run out of it
Gear for the same activity
Trail shoes get running socks, a bottle and a headlamp; a tent gets a sleeping bag, a pad and a lantern. Personyze keeps each row to one activity: running, football, camping, ski.
- One activity per rowNo tent next to running shoes, no ski wax next to a football.
- From the picture and the ordersWhat the gear is for, and what your shoppers bought for the same trip.
- Add-ons for lessEvery piece costs less than the shoe or the tent it goes with.
More on it: recommendations on an ecommerce site
ViewingTrail Runner, Olive$129
Merino Running Socks$18
Sports Bottle$16
Compact Headlamp$39
Viewing
More on this topic, Go deeper
Personyze reads every article once: its topics, the people, companies and places in it, news or guide, beginner or expert, and whether it stays relevant. A story published an hour ago has related reading before anyone has opened it, and the list is always full.
- More on this topicThe same subject from another angle.
- Go deeperThe natural next step: from the news story to the explainer, from a guide to a case study, and on to the page that converts.
- For youArticles matched to the topics and kinds of content this reader reads.
More on it: content recommendations · the content recommendations playbook
Go deeperInside the lab that measures ocean heat
More on this topicOffshore wind’s next ten years
Another angleWhat a warmer coast means for marketsTwo signals for every product. The balance shifts by itself.
For each product, Personyze weighs what shoppers did with it and how it looks. A bestseller is recommended mostly from real purchases; a new arrival mostly from its look, so it gets good recommendations the day it is added.
Purchases count from day one. The balance moves on its own as they build up.
Bought togetherShoppers who buy items together are building an outfit.
Viewed togetherShoppers who view items together are usually comparing alternatives.
MLKidsAdultGIFT CARDNot only which items. If shoppers who buy black sporty hoodies also buy black track pants, a hoodie added yesterday is matched the same way.
The same item in another size, children’s items next to adults’, and things bought with everything, such as gift cards.
A combination many people bought leads one that two people happened to buy.
What readers did, and what each article is about: its topics, the people and places in it, news or guide. More on content recommendations.
Nothing to configure. Automatic includes Complete the look on product pages for shops with look data. Reading product pictures is one switch in Recommendation settings, and before you turn it on you see the estimated AI credits for your own catalog.
The algorithm picks. The rules put what fits them first.
Add Match the visitor to any algorithm: brand and size from the shopper’s own history, their gender, and a price like the ones they look at. The algorithm still chooses the products; the rules decide which come first.
Bought a Northline jacket in size M last winter, and browses men’s jackets at around $130.
The rules, in the order you set- 1Brandfrom their own historyNorthline
- 2Sizefrom their own purchasesM
- 3Genderunisex matches everyoneMen’s
- 4Price like theirswhat they usually look atAround $130
0201Insulated Parka, NavyNorthlineS–XXLMen’s$1390402Hooded Down Parka, NavyNorthlineM–XXLMen’s$1290303Alpine Parka, MossAlderXS–XLUnisex$1490104Alpine Parka, SlateAlderXS–MWomen’s$149The rules re-order its own top picks. They never fill the row with random items.
With no history yet, every rule comes up empty and the algorithm’s order stands.
On your site, in recommendation emails, in the JSON API and in feeds.
Match the visitor on the recommendations page · how it works, with recipes for each kind of site
Five things every list gets right. Built in.
Whichever algorithm a widget runs and whatever you sell, these hold for every row of recommendations.
BoughtNever what they just bought
A shopper who bought the tan sneakers yesterday is not offered them again today.
Each product once
One jacket in three colours and six sizes takes one place in the row, not nine.
$129≈$139$129>$45Alternatives at a similar price, add-ons for less
A $129 jacket gets other jackets near $129, and add-ons priced below it.
Lists always full
When one source runs thin, the next fills in, so a row never shows an empty slot.
Measured per source
Clicks and purchases are counted for each recommendation source, so you see which one sells.
Every algorithm, wherever you need it.
The products an algorithm picks for someone are not tied to one widget. The same picks, for the same person, go into your emails, notifications, landing pages, app and your own systems.
Trail Runner, Olive$129Canvas Backpack 25L$89Merino Watch Cap$39One widget for every page. It picks the algorithm itself.
Put Automatic on every page and it works down one order: the page’s own algorithm first, then what this visitor is likely to want, then what is popular where they are and what is trending — and the newest products last, so it is never empty.
Automatic: one widget, every page
In the panel, Automatic is the first card when you set up a widget, and a new widget starts on it; pick All algorithms beside it to choose one yourself.
Merino Watch Cap$39Trail Runner Low$89Weekender Duffel$119Knit Beanie, Mustard$45A first visit: nothing personal to go on yet, so it shows what is popular where they are.
Every algorithm in the panel is one of five kinds: predictive and popular, crowd wisdom, their own items, what just changed, your rules.
On a content site the order is Read Next, then Similar Content (AI) on an article page; then Content Recommended for You, Popular in the Visitor's Country and Trending Now on every page; and the newest articles last, so it is never empty.
In a store with look data Automatic also shows Complete the look on product pages, and every step is tuned to your kind of shop. By kind of store ↑
Every recommendation algorithm, by what it learns from.
Five kinds: what each visitor is likely to want and what is popular, crowd wisdom around one item, their own items, what just changed in your catalog, and your own rules. Pick one to see what it shows, the page it fits and what it needs.
What this visitor is most likely to want, and what everyone is viewing.
Personalized Recommendations
Predicts what each shopper is most likely to buy next, learned from the shoppers who behave like them.
- Best on
- Home — and any page, starting from the product, cart or order on it
- Needs
- Visits
- For articles
- Content Recommended for You
Most Popular
Your crowd-pleasers: the most viewed products in a period, across the store or in one category.
- Best on
- Home, Category
- Needs
- Product views
- For articles
- The same name
Popular in the Visitor's Country
What shoppers where they are view most — picks that fit from a visitor’s very first page.
- Best on
- Home; first visits; stores with shoppers in many places
- Needs
- Product views, by visitor location
- For articles
- The same name
Trending Now
Catches what is taking off: products whose views jump above their own weekly average, before they top the charts.
- Best on
- Home; also safe for email
- Needs
- Product views
- For articles
- The same name
Sales Falling
Spots the products selling less than the week before and puts them back in front of shoppers before they go stale.
- Best on
- Reviving slow movers; the same for every visitor
- Needs
- Orders
- For articles
- Products only
Best Sellers
Your most purchased products, across the store or in the category being browsed — the strongest signal on a category page.
- Best on
- Category
- Needs
- Orders
- For articles
- Products only
Most-Added to Wishlist
The products shoppers save most — a strong signal of intent for retargeting and cart emails.
- Best on
- Retargeting and abandonment emails
- Needs
- Wishlist tracking
- For articles
- Most Liked Content (favorite or wishlist)
Inspired by Past Purchases
Learns from shoppers who bought the same things: what they went on to view and buy next.
- Best on
- Home, for past buyers
- Needs
- Orders
- For articles
- Inspired by Reached Goals
What your shoppers view, buy and open next around the product on screen.
Others Who Viewed Also Viewed
The classic “customers also viewed”: what other shoppers viewed in the same visits as this product.
- Best on
- Product page
- Needs
- Product views
- For articles
- Visitors Who Read This Also Read
Viewed Next
Where shoppers actually go next from this product, learned from the order of views in each visit.
- Best on
- Product page
- Needs
- Product views
- For articles
- Read Next
What Others Viewed Then Bought
What shoppers who viewed the same products went on to buy — a step closer to the sale than “also viewed”.
- Best on
- Product page, Cart
- Needs
- Orders
- For articles
- What Readers Went On to Convert On
Cross-Sells
What goes with it: the products your shoppers bought together, learned from your own orders.
- Best on
- Cart, Thank-you page
- Needs
- Orders with several items
- For articles
- Products only
Inspired by Category
More from the category or interest being browsed, so the next good find is one click away.
- Best on
- Product page, Search results
- Needs
- Categories in your catalog
- For articles
- More From This Topic
Up-Sells
One step up: higher-priced alternatives to the product viewed or in the cart, to raise the order value.
- Best on
- Product page, Cart
- Needs
- Prices in your catalog
- For articles
- Products only
Similar Items (AI)
AI reads your catalog’s titles, brands, descriptions and categories, so a brand-new product has matches from day one.
- Best on
- Product page
- Needs
- Switched on in Catalog settings (uses AI credits)
- For articles
- Similar Content (AI)
Managed Cross-Sells
Your own pairs, uploaded to the catalog, for the times you know exactly what goes together.
- Best on
- Product page, Cart
- Needs
- An uploaded cross-sell list
- For articles
- Managed Related Articles
Cross-Sells for Last Item Added to Cart
What goes with the one product they just added, while they are still at the cart.
- Best on
- Cart, right after an add-to-cart
- Needs
- Add-to-cart tracking and orders
- For articles
- Products only
Last Bought Item: Those Who Bought This Also Bought
What goes with the product they just bought — on the thank-you page and in the follow-up email.
- Best on
- Thank-you page; post-purchase email
- Needs
- Orders
- For articles
- Products only
Back to what this visitor viewed, saved or bought before.
Buy it Again
What they bought before, back in front of them — made for refills and everyday things.
- Best on
- Refills and consumables; after a purchase
- Needs
- Orders
- For articles
- Reached Goal
Matches Their Search
Products matching their last search, on your site or at the search engine that sent them.
- Best on
- Search results; Home after a search
- Needs
- A search
- For articles
- The same name
Recently Viewed
An easy way back to the products they were considering.
- Best on
- Home, Cart; returning visitors
- Needs
- Product views
- For articles
- The same name
Items in Cart
What is in their cart right now — a reminder on any page, or in an abandoned-cart email.
- Best on
- Any page as a reminder; abandoned-cart email
- Needs
- Cart tracking
- For articles
- You Commented
Your Wishlist
The products this shopper saved, back in front of them when they return.
- Best on
- Returning visitors
- Needs
- Wishlist tracking
- For articles
- Content You Liked
Cheaper, back in stock or new since their last visit.
Price Dropped
Products cheaper now than on their last visit — a reason to come back.
- Best on
- Returning visitors; email
- Needs
- Prices in your feed
- For articles
- Products only
Back in Stock
Products that came back after selling out — recently, or since their last visit.
- Best on
- Returning visitors; restock reminders
- Needs
- Stock in your feed
- For articles
- Products only
New in Stock
What is new in your catalog, or new since their last visit.
- Best on
- Home, for returning visitors
- Needs
- A catalog feed
- For articles
- Recently Published
Any products, filtered and ordered your way.
Any products
Any products you choose, narrowed by your own rules and ordered your way.
- Best on
- Anywhere a fixed selection is wanted
- Needs
- Nothing
- For articles
- Any Articles (custom filter)
What this reader is most likely to read, and what everyone is reading.
Content Recommended for You
Predicts what each reader is most likely to read next, learned from the readers who behave like them.
- Best on
- Home — and any page
- Needs
- Visits
- For products
- Personalized Recommendations
Content Recommended for You — Published Since Last Visit
For a returning reader: what they are likely to read among the articles published since their last visit.
- Best on
- Home, for returning readers
- Needs
- Visits, and publish dates in your catalog
- For products
- Articles only
Most Popular
Your most read articles in a period, across the site or in one section.
- Best on
- Home, Category
- Needs
- Article views
- For products
- The same name
Popular in the Visitor's Country
What readers where they are read most — picks that fit from a reader’s very first page.
- Best on
- Home; first visits
- Needs
- Article views, by reader location
- For products
- The same name
Trending Now
Catches the story taking off: articles read far above their own weekly average, before raw counts catch up.
- Best on
- Home
- Needs
- Article views
- For products
- The same name
Most Liked Content (favorite or wishlist)
The articles readers like and save most — the ones they want to come back to.
- Best on
- Home, Category
- Needs
- Likes or saves
- For products
- Most-Added to Wishlist
Most Commented
The articles with the most comments in a period — where the conversation is.
- Best on
- Home, Category
- Needs
- Comments on your articles
- For products
- Articles only
Inspired by Reached Goals
Learns from readers who reached the same goals: what they went on to read.
- Best on
- Home
- Needs
- Goals set in Conversion Tracking
- For products
- Inspired by Past Purchases
What your readers read and open next around the article on screen.
Visitors Who Read This Also Read
What other readers read in the same visits as this article.
- Best on
- Article page
- Needs
- Article views
- For products
- Others Who Viewed Also Viewed
Read Next
Where readers actually go next from this article, learned from the order of reads in each visit.
- Best on
- Article page
- Needs
- Article views
- For products
- Viewed Next
Visitors Who Read This Also Read — From Current Category
The same crowd wisdom, kept to the category of the article being read.
- Best on
- Article page
- Needs
- Article views and categories
- For products
- Articles only
What Readers Went On to Convert On
What readers of the same articles read on their way to your goal — a sign-up, a subscription or a purchase.
- Best on
- Article page
- Needs
- Goals set in Conversion Tracking
- For products
- What Others Viewed Then Bought
More From This Topic
More on the topics of the article being read.
- Best on
- Article page, Search results
- Needs
- Topics or categories on your articles
- For products
- Inspired by Category
Similar Content (AI)
AI reads your titles, topics, authors and descriptions, so a story published an hour ago has related reads before anyone has read it.
- Best on
- Article page
- Needs
- Switched on in Catalog settings (uses AI credits)
- For products
- Similar Items (AI)
Managed Related Articles
Your own pairs of related articles, for the coverage you want read together — our support team loads them for you.
- Best on
- Article page
- Needs
- Your pairs, loaded by support
- For products
- Managed Cross-Sells
Back to what this reader read, liked or commented on before.
Reached Goal
The articles this reader reached a goal on — a finished tutorial, a recipe they completed.
- Best on
- Returning readers
- Needs
- Goals set in Conversion Tracking
- For products
- Buy it Again
Matches Their Search
Articles matching their last search, on your site or at the search engine that sent them.
- Best on
- Search results
- Needs
- A search
- For products
- The same name
Recently Viewed
Back to the articles they were reading.
- Best on
- Home; returning readers
- Needs
- Article views
- For products
- The same name
You Commented
The articles this reader commented on — the conversations they are part of.
- Best on
- Returning readers
- Needs
- Comments on your articles
- For products
- Items in Cart
Content You Liked
The articles this reader liked or saved, back in front of them.
- Best on
- Returning readers
- Needs
- Likes or saves
- For products
- Your Wishlist
Published since their last visit.
Recently Published
Your newest articles, or those published since their last visit.
- Best on
- Home, Category
- Needs
- Publish dates in your catalog
- For products
- New in Stock
Any articles, filtered your way.
Any Articles (custom filter)
Any articles you choose — by category, topic, publish date or your own fields — ranked within your selection.
- Best on
- Anywhere a fixed selection is wanted
- Needs
- Nothing
- For products
- Any products
Every widget runs one algorithm, and can keep others behind it to fill any empty slot. On top of these, Automatic tunes itself to your kind of store: Complete the look, Similar style, In your style, Accessories that fit, Complete the room, Time to restock and Gear for the same activity (by kind of store). How each family works, in plain words: recommendation algorithms explained.
Which algorithm on which page, and what fills in.
Each page in the widget editor opens with one algorithm marked Recommended. If your store has no data for it yet, the mark moves to the next in line — and every widget keeps fallbacks behind its algorithm, so a row is never empty.
Choose the algorithm, then keep the row full.
Two walkthroughs in the real panel: one widget built properly, with a filter, a fallback and a title for each algorithm — and Matches Their Search putting what a visitor searched for on the home page.
Each algorithm, up close.
The algorithms shoppers recognise have pages of their own, and the guides show them at work on real kinds of stores.
++AlgorithmCustomers who bought this also boughtWhat goes with the product, learned from your own orders.AlgorithmCustomers who viewed this also viewedAlternatives from the same visits, for shoppers still choosing.TrendingAlgorithmTrending and popular productsWhat is taking off this week, and what everyone buys.AlgorithmYou may also likePicks for each shopper, learned from shoppers like them.ViewedAlgorithmRecently viewedAn easy way back to what they were considering.
GuideRecommendation algorithms explainedCollaborative filtering, bought together, trending and AI similarity, in plain words.
GuideProduct recommendation examplesFourteen personalized ways to sell more, page by page.
GuideProduct recommendations, start to finishCatalog, algorithm, design, and a preview as a real shopper.ExampleFashion personalizationA fashion store that shows each shopper the right season, fit and style.PlaybookProduct recommendations playbookWhich widget on which page, and how to check it works.PlaybookContent recommendations playbookRead next, recommended for you and recently read, set up on a content site.Help centerEvery algorithm, in the help centerEach algorithm’s settings, filters and fallbacks, with starter recipes.Product recommendation algorithms, answered.
What teams ask most when they choose an algorithm. Got something else? Bring it to the call.
Every algorithm in the help center →What is a product recommendation algorithm?
The rule a recommendation widget uses to choose which products to show each shopper: what shoppers like them bought or viewed, what goes with the product on screen, what looks like it, their own history, or what just changed in your catalog. Personyze has 27 algorithms for products and 22 for articles; every widget runs one and keeps others behind it to fill any empty slot.
What are the main types of recommendation engine algorithms?
Collaborative filtering learns from what shoppers did together: Cross-Sells, Others Who Viewed Also Viewed and Viewed Next. Content-based algorithms match products by what they are and how they look: Similar Items (AI), Complete the look and Similar style. Popularity shows what everyone wants: Best Sellers, Trending Now and Popular in the Visitor's Country. History brings back the shopper’s own items: Recently Viewed and Buy it Again. And your own rules choose any products you like. Personyze weighs the collaborative and the content signals together for every product.
Which recommendation algorithm should I use on each page?
The one the widget editor marks Recommended: Personalized Recommendations on the home page, Best Sellers on a category page, Viewed Next on a product page, Cross-Sells in the cart and on the thank-you page, and Matches Their Search on search results; on an article, Read Next. If your store has no data for it yet, the mark moves to the next in line.
How does Complete the look work?
On a product page it shows pieces of other kinds that go with the item on screen: a navy bomber gets a sweater, a beanie and sneakers, never more jackets. It comes from what shoppers bought together, cleaned up by each product’s look, so the same item in another size, children’s items next to adults’ and things bought with everything are dropped. Similar style comes from what shoppers viewed together: the same kind of item in a matching style.
How do you recommend a brand-new product with no sales yet?
From its look. Personyze reads each product’s picture and title once, so a new arrival is recommended mostly from how it looks on the day it is added, and more from real purchases as they build up; the balance shifts by itself. Similar Items (AI) reads titles, brands, descriptions and categories too, and New in Stock puts what was just added in front of returning visitors.
Do the algorithms adapt to my kind of store?
Yes. Personyze reads your catalog, works out what kind of shop it is and tunes the algorithms to it: Complete the look in fashion, accessories that fit in electronics, Complete the room in home and furniture, Time to restock in beauty, grocery and pet stores, gear for the same activity in sports and outdoor, and the next step on the same topic on a content site. Nothing to configure.
Can recommendations match each shopper’s size, brand and price?
Yes. Match the visitor puts what fits each shopper first: brand and size from their own history, their gender, and a price like the ones they look at. The algorithm still picks the products and the rules only change the order, so the row is never filled with random items.
What does Automatic do?
One widget for every page that chooses what to show on each page by itself, from your own data, in one order: on a product page Viewed Next, then Similar Items (AI); on a category page the best sellers in that category; with something in the cart, cross-sells for what is in it; then on every page Personalized Recommendations, Popular in the Visitor's Country and Trending Now; and the newest products last, so it is never empty. In a store with look data it also shows Complete the look on product pages. In the widget editor it is the first card, marked Recommended, and a new widget on your site starts on it.
What is the difference between “customers who bought this also bought” and “customers who viewed this also viewed”?
Bought together shows what goes with the product, learned from orders: shoppers who buy together are building an outfit or a set, so it feeds Cross-Sells and Complete the look. Viewed together shows alternatives, learned from visits: shoppers who view together are comparing, so it feeds Others Who Viewed Also Viewed and Similar style. More on customers who bought this also bought and customers who viewed this also viewed.
Can I see which algorithm sells more?
Yes. Every widget counts times shown, clicks, click rate and purchases within 7 days of a click, for each recommendation source and each product, and an A/B test can put two widgets with two algorithms against each other.
+
+
+
+
+
+
+
+
The right algorithm on every page — learning from every sale.
Self-learning recommendations, powered by machine learning on your shoppers’ behaviour: every view, search, cart and order goes into the next build of the picks, so what you show keeps up with what sells.