Best sellers and trending products or top lists in short is a recommendation type based on overall trend data and popularity displaying what is currently “hot and popular” on the online store.

Required page type context: None

Minimum data required: product views

Available settings

Displays shopper's product browsing history in a linear order. Similar to personalized recommendations, but it doesn’t add any weight for products meaning that products are displayed in the order they have been browsed. Since the recommendation type is based on personal shopping history, it’s only displayed to shoppers who have browsed more than one product during current or previous visits to the site.

Required page type context: none

Minimum data required: Products viewed

Available settings

Displays products related to a visitor’s recent browsing history, excluding products a visitor has already browsed or bought, by comparing visitor’s browsing history to similar profiles. In practice it displays new and unseen products to an individual visitor, based on their personal preferences. Since the recommendation type is based on personal shopping history, it’s only displayed to shoppers who have browsed at least one product during current or previous visits to the site.

Required page type context: none

Minimum data required: Products viewed

Available settings

Dynamic Bundles are a way to create sub-collections of items within Nosto that go well together. Popular use-cases might be “Frequently bought together” or “Get the Look” kind of inspirational content that are then exposed through the use of Onsite Product Recommendations. Technically, Dynamic Bundles can only work on PDPs. Bundles also work as static bundles similar to cherry-picked recommendations.

Required page type context: none

Minimum data required: Products viewed

Available settings

Manual merchandising. Displays always the same products based on productIDs and order of productIDs. Remember to verify that products are in stock, since even manual merchandising with Cherry-Pick adheres to availability status.

Required page type context: none

Minimum data required: Products

Available settings

Displays product recommendations that are related to the product that the visitor is viewing, such as “Customers Who Viewed This Also Viewed.” Cross-selling and up-selling has include and exclude rules allowing filtering which type of products are displayed in a recommendation, for example controlling whether the recommendation displays alternatives or supplementary products.

Required page type context: Product page

Minimum data required: Products, product views

Available settings

Free shipping recommendation is a similar setting to shopping cart recommendation, but including customisable visibility settings based on total cart value. Otherwise the recommendation type works similarly to the shopping cart recommendation.

Required page type context: Cart

Minimum data required: Products

Available settings

Geo-targeted is geo-based version of best sellers, featuring similar options, but recommending products that are trending in user's location: country, state, city or even city district.

Required page type context: None

Minimum data required: product views

Available settings

Recommendation type displays products depending on the external source of traffic, read from the utm parameters and Google gclid parameters. Recommendation looks into behaviour of the shoppers and what the same group browsed, carted and bought.

Typically used on landing pages, blog posts talking about a specific category, brand or seasonal products to automatically determine common denominators for shoppers after engaging with content. Grouping shoppers by advertisement channel and determining interest after engaging with content happens automatically. Landing Page Recommendations work automatically when deployed on the target page and don't have any additional options, but due to its nature we recommend using Fallback Recommendations with it. The same experience with more granular controls for content can be achieved with segmenting shoppers by source of traffic.

Required page type context: None

Minimum data required: Traffic with utm or gclid parameter, product views

Available settings

Designed as ticker version and instead of showing products that are trending using longer time periods, the recommendation shows in real-time what products current customers interact with: view, add to cart and buy, optionally displaying user's location at a city level. Live feed is designed to create similar buzz effect online that is common during sales seasons in an offline environment.

Required page type context: None

Minimum data required: product views

Available settings

Displays product recommendations that are related to shopper’s previous order.

Required page type context: None

Minimum data required: purchase conversions, identified shopper

Available settings

The recommendation type is basically weighted browsing history with alternatives. Recommendation displays shoppers the items they viewed or added to their cart, but did not buy, during the current visit or previous visits to the store giving more weight for items visited more frequently, multiple times and especially those added to the cart, optionally excluding bought items.

If a user only interacted with small amount of products, recommendation automatically falls back to relevant alternative items instead of browsing history. Since the recommendation type is based on personal shopping history, it’s only displayed to shoppers who have browsed more than one product during current or previous visits to the site.

Required page type context: None

Minimum data required: Products, product views

Available settings

Shows completely random set of products indexed by Nosto. Intended for debugging and testing purposes for example on test-environments.

Required page type context: none

Minimum data required: products

Available settings

Displays products previously bought by the same shopper as a personalized version, and those which are commonly bought by shoppers and which a shopper bought before as the general version. Effective especially for consumables where rebuying the same product is common.

Required page type context: None

Minimum data required: purchase conversions, identified shopper

Available settings

Displays products that other shoppers have gone on to view and buy following the same search query using site’s search. Alongside with all searches made, recommendation type also catches trending search terms, common typos and in practice is often able to display a recommendation whereas site’s internal search returns an empty search result.

Required page type context: Technically none, but practically search results page

Minimum data required: Search query, product views

Available settings

A sibling of Search and Viewed Related Recommendation. Displays items that customers bought after using the same internal search term. Since the recommendation type only tracks purchases made after an internal search it requires massive amount of data to work effectively as only a share of search terms are taken into account. Search and viewed tracks also product views not explicitly conversions making it in almost always more practical recommendation type. Use only if your shoppers make a massive amount of searches daily.

Required page type context: Technically none, but practically search results page

Minimum data required: Search query, conversions

Available settings

Similar recommendation type to cross-selling and up-selling, but intended for shopping cart page. Shopping cart recommendation displays products based on all products in shopping cart and is often displayed as “Customers Who Bought These Bought Also.” Recommendations typically become more relevant when more items there are held in the cart at the same time, as Nosto can base recommended products on more data.

Required page type context: Cart

Minimum data required: Products, carts

Available settings

The same recommendation as "Order Related Products", but reads the context only from the current order.

Required context: Order

Available settings

The same recommendation to Product based cross selling, but only makes use of visual similarity by Nosto’s AI of products as the scoring mode e.g. visually similar to the viewed product

Required page type context: Product page

Minimum data required: products

Available settings

The same recommendation to Browsing history related, but only makes use of visual similarity by Nosto’s AI of products as the scoring mode e.g. visually similar to your browsing history.

Required page type context: None

Minimum data required: product views

Available settings

The same recommendation to Cart based recommendations, but only makes use of visual similarity by Nosto’s AI of products as the scoring mode e.g. visually similar products to those in your cart

Required page type context: Cart

Minimum data required: products

The same recommendation to Order related recommendations, but only makes use of visual similarity by Nosto’s AI of products as the scoring mode e.g. visually similar products to those ordered before.

Required page type context: order

Minimum data required: product views, purchase conversion

The same recommendation to Thank you page recommendations, but only makes use of visual similarity by Nosto’s AI of products as the scoring mode e.g. visually similar products to those ordered before.

Required page type context: order

Minimum data required: product views, purchase conversion

Scoring modes are applicable for best seller type of recommendations making use of trend data

Most viewers ranks products by the number of product views by individual shoppers

Most buyers ranks products by individual shoppers who bought them

Most buys ranks products by number of items sold

Most total revenue ranks products by the total sales they have generated

Using algorithmic models, Nosto’s relation types analyze patterns and similarities among shoppers and products by comparing all shopper sessions, journeys and purchases. By understanding shoppers' preferences and identifying similarities with other individuals and products, relation types generate the recommendation output. Different relation types are controls to recommend products based on different types of product relations e.g. control the algorithmic output.

Shopper interactions with recommendations and products update in real-time. This is easiest to comprehend through the Browsing History recommendation type, which following its name shows shopper’s linear browsing history on the site. In order to achieve this, Nosto has to take into account the last viewed product or page in order to reflect the changes in Recommendations. The same logic applies to other recommendation types: Every shopper interaction on the website is interpret as a signal, which is reflected in recommendations. Relations between products are processed at a slightly slower rate, but still within the same 24 hours.

In order to be able to recommend entirely new products when these are published, Nosto makes use of not only behavioural signals referred above, but also visual similarity and diversity and contextual matching based on product’s name, category, brand and description. These factors combined ensure that also products that are recently introduced are not buried by older and more popular products based on engagement.

As a broader topic, but related to the same concept of new products and automated data import, new accounts can in practice make use of recommendations immediately after finishing integration, yet the relevancy naturally increases after a learning period, which is commonly just some days long. Features that require product performance data like conversions, clicks add-to-carts and similar e.g. Merchandising and Popularity based algorithms, naturally become more effective after processing more interactions and conversions.

On all platform based integrations, Nosto automatically imports past 30d of product sales data in order to trim the data model and recommendations. In practice this means that there isn’t a need to import product sales data. Importing data most commonly is useful as customer lists imported as segments, in case there are existing customer groups that should be leveraged further.

Nosto’s recommendation data models are based on over a decade of serving the e-commerce and retail industry and the team actively improves and optimizes data models and algorithms used in the product. Therefore exact definition is not practical.

In a simplified manner, shopper signals are weighted differently and the models generally presume that products that are bought together repeatedly by multiple shoppers have a closer relation, but which is more applicable for supplementary options. On the contrary, products that are viewed together are likely closely related, but more in use cases when a shopper is still exploring options. Each mode is also manually controllable for users requiring closer control.

Shopper signals indicate different levels of interest, which also influence how products are recommended together and to who. As a practical example, a search-action indicates generally higher interest than viewing a top level category.