In the Frosmo Platform, a recommendation is a piece of dynamically generated content predicted to appeal to a visitor and delivered through a modification. The platform generates recommendations using the Frosmo Recommendations feature. Product recommendations, that is, recommendations generated from product data and transaction data are the most common category of recommendations.
Figure: Most purchased products on a product category page
For more information, see:
Frosmo Recommendations is an end-to-end solution for generating recommendations in the Frosmo Platform. You first define what recommendation to generate in the FCP, and then retrieve and display the generated recommendation data in a modification on your site.
Frosmo Recommendations generates recommendations based on usage data collected from the site and using an algorithm or filter that produces relevant results from that data. For producing the results, Frosmo Recommendations supports collaborative filtering methods out of the box, while content-based filtering methods, hybrid solutions, and other machine-learning solutions are available upon request.
Figure: Recommendations in the Frosmo Platform
You can generate the following types of recommendations based on collaborative filtering:
While you can, in principle, use Frosmo Recommendations to generate recommendations from any data and using any recommendation algorithm, the feature currently provides built-in support only for product recommendations based on collaborative filtering. If you want to create other types of recommendations, contact your Frosmo representative.
In the Frosmo Platform, products are not limited to purchasable items in the traditional retail sense. You can track any website content, such as articles and downloadables, as products.
Figure: Recommendations for products purchased together on a product page
Recommendations rely on usage data collected from the site. One way of collecting this data is through the data layer. The following table shows you what data each supported recommendation type uses for generating recommendations and which data layer object you can use to collect this data from the site.
Table: Mapping recommendation types to data layer objects
|Recommendation type||Source usage data||Data layer object|
|Most viewed products||Product data||Product|
|Most purchased products||Transaction or conversion data||Transaction or conversion|
|Products viewed together||Product data||Product|
|Products purchased together||Transaction or conversion data||Transaction or conversion|
|Products viewed and purchased together|
Transaction or conversion data
Frosmo Recommendations and legacy recommendation modifications are two different solutions for creating recommendations in the Frosmo Platform. Recommendation modifications represent a legacy solution that will be phased out in the future.
You use a basic modification, not a recommendation modification, to display a recommendation generated with Frosmo Recommendations. (You can also use a cached modification, but since the recommendation data comes from the Frosmo back end, rather than from the site's custom script, the cached modification is effectively turned into a basic modification.)
The process for creating a recommendation flows as follows:
The following figure illustrates this process.
Figure: Process for creating a recommendation
In addition to showing generic recommendations for most viewed and purchased products, you can come up with more refined recommendation strategies, for example:
For an end-to-end example of using recommendations, see Example: Recommending products purchased together.
Recommending content, such as articles and stories, instead of purchasable products is relevant to both media sites and ecommerce sites. Many retail sites nowadays include a blog or user-generated content to provide more value to visitors. Content recommendations increase the time visitors spend on the site and improve visitor engagement. The revenue for the content comes from ads and, for many media sites, the increased number of subscriptions.
To implement content recommendations, you can rely on most viewed articles or articles viewed by other visitors with a similar profile. You can also employ natural language processing (NLP) algorithms that actually crawl your site and recommend content based on similar words or word combinations.
Figure: Recommending most viewed articles
The most effective recommendations are often based on combining different data sources and recommendation types. Many businesses already build their entire webshops with recommendations, that is, ranking and grouping products based on different filters.
This approach lets you both personalize content for individual visitors and highlight products you want to promote. You can combine, for example, the visitor's purchase history, default products, most sold products, and products related to the ones the visitor has recently viewed.
For the best effect, in addition to recommendations, let your visitors take control by providing them with filters they can use to customize the product catalog even further.
Figure: Building the store front based with recommendations
Shopping cart recommendations are arguably the most-used product recommendation type. Shopping cart recommendations effectively increase the average order value by cross-selling and upselling complimentary products.
The simplest way to benefit from shopping cart recommendations is to show generic, inexpensive products that most people know and use, such as socks, screen wipes, or pillowcases. You can also recommend complimentary products and highly specialized accessories to the original product. However, it's important to understand when to display recommendations. First, let the visitor proceed far enough in the funnel and only show complimentary products when the purchase decision has been made. Secondly, don't disturb the purchase flow. Show products that the visitor can add to the shopping cart without having to view details about colors, sizes, or compatibility with other products.
You can use triggers to launch recommendations based on visitor actions.
Figure: Recommendations in the shopping funnel
To start developing recommendations, see Developing recommendations.
If you're impatient and just want to get working on an example, see Example: Recommending products purchased together.