Working with product sorting¶

The sort order of products in Voyado Elevate 3 can be based on a variety of things. By default, the sort order of a product list is by relevance according to Elevate, relevance. The sort_by argument can be included in a panel query to specify another sort order.

The type of sorting used should be aligned with the purpose of the panel used. An example of this is a panel that showcases new products. A relevant sorting for this panel can be based on a product attribute with the release date in descending order. Sort attributes for specific products attributes can be configured in the Voyado Elevate manager.

Sorting modifiers¶

Sorting modifiers asc and desc can be included in a sort_by argument to sort the data ascending or descending. If a sorting modifier is not included in an argument, the default modifier of the sorting type will be used.

It is not recommended to use a sorting modifier when sorting by relevance, personal, sales, or clicks. These sort orders default to desc. Using asc as a sorting modifier with these sort orders may produce unexpected results.

Sort by relevance¶

The sort order relevance is the default sorting type if the sort_by argument is not included in a panel query. Relevance of a specific product is based on context and will differ based on what panel or search phrase is used.

The relevance is calculated by Elevate using parameters such as product information, strategy score, and customer history.

The default sorting modifier for relevance is desc. It is not recommended to include a sorting modifier when sorting by relevance. Products in some panels sorted by relevance are affected by exposure strategies, boost, and bury.

Example arguments¶

sort_by: relevance

Sort by product attribute¶

Any product attribute can be used for sorting of products if a sort attribute has been configured for it. This attribute is then used in the argument sort_by. When sorting by product attribute, the result set can contain a maximum of 100 000 variants.

The default sorting modifier for a product attribute is asc. It is recommended to include a sorting modifier when sorting by a product attribute.

Example arguments¶

sort_by: release_date desc
sort_by: price asc

Sort by personal¶

The personal sort order tries to find products that can be associated to a visitor based on the following historical visitor actions: clicks on products, adding a products to cart, and purchasing products. It is not recommended to use the personal sort order with search, such as with a search panel, as the match quality is not considered for the sorting.

The intended use of personal sort is for category or other landing pages. On these pages broad queries span large sets of products and match quality is either evident or meaningless. The personal sort brings up to five personalized products to the top and backfills with the current best-selling products.

The default sorting modifier for personal is desc. It is not recommended to include a sorting modifier when sorting by personal. Products in panels sorted by personal are affected by exposure strategies, boost, and bury.

Product selection¶

A score is given to the most recent of historical visitor actions: clicks on products, adding products to cart, and purchasing products. Historical actions does not include actions from the current session. The score for clicks and add-to-cart actions are lowered with age. For each of these product actions a probabilistic distribution is created which connects the product to other products based on criteria such as how often the products are bought together, and how often the products are clicked in the same session. The scored distributions are then merged and the top products are returned.

The personal sort order uses the top five products returned. The Recommend based on customer panel use the same method for product selection but with the arguments window_first and window_last to determine what number of products to return.

Example arguments¶

sort_by: personal

Sort by sales¶

The sales sort order sorts products according to their strategy score.

The optimal period of time for the strategy score is automatically and individually chosen per product. This enables statistically significant trends to be recognized, and adequately compare products with a steady purchase rate to those that have quick changes in popularity.

The default sorting modifier for sales is desc. It is not recommended to include a sorting modifier when sorting by sales. Products in some panels sorted by sales are affected by exposure strategies, boost, and bury.

Example arguments¶

sort_by: sales

Sort by clicks¶

The clicks sort order sorts products according to popularity based on recent clicks. This type of sorting is useful when products cannot be purchased, such as for sites where accessories are sold for devices, but not the actual device itself. Sorting with clicks can be used as a good alternative for sites with low traffic as clicks are more frequent than purchases.

The default sorting modifier for clicks is desc. It is not recommended to include a sorting modifier when sorting by clicks.

Example arguments¶

sort_by: clicks

Sort by arithmetics¶

Custom sort orders can be created with arithmetics, using operators with sorting types. Normal execution precedence is applied but can be overridden with parentheses. It is also possible to use floating point literals.

It is not recommended to use the relevance sorting type in arithmetic expressions.

The default sorting modifier for a custom sort order is asc. It is recommended to include a sorting modifier when sorting by custom order.

Operations¶

The following operators are allowed when creating custom sort orders.

Example arguments¶

sort_by: (avg_rating * rating_count + 300) / (rating_count + 100) desc

Multi-level sorting¶

Additional sorting levels can be used when sorting products. This is useful when there is a likelihood of multiple products having the same primary rank. The levels of sort orders are separated by a comma, ,, in the sort_by argument and are listed in priority order.

Multi-level sorting affect the behavior of exposure strategies, boost, and bury in the panels they are used. For example, a site selling Blu-Ray and DVD movies has an active boost for a certain kind of DVDs. A panel is used to list movies with format asc as a primary sort and relevance as a secondary sort. Movies with the Blu-Ray format will thus be returned before the DVD format.

The active boosting will not place the DVDs selected in the boost above the returned Blu-Ray movies as format is the primary sort, but it will affect in what order products with the DVD format are returned.

Example arguments¶

sort_by: format asc, relevance
sort_by: in_stock desc, relevance
sort_by: price desc, title asc
sort_by: price asc, sales desc

Sorting recommendations¶

Following is a recommendation of when different sort orders are appropriate.

Refinements¶

Refinements can be set for sort attributes in the Voyado Elevate manager to enable normalisation of product attributes before sorting. The most common usage of refining is to normalise text to lower case or to remove all non-digit characters before sorting by price. The following refinement settings are available:

• (no refinement) - No normalisation is performed
• Case insensitive - Normalisation to lower case
• Keep digits and letters (e.g. Text123) - Removes all characters that are neither letters or digits, including spaces
• Keep digits and letters, case insensitive - Removes all characters that are neither letters or digits, including spaces, and normalises letters to lower case
• Keep digits only (e.g. 1234) - Removes all characters that are not digits