Merchandising

Merchandising allows you to make advanced rules to govern the way your search and recommendations behave. It allows you to highlight or bury entities, to help drive views and conversion rates for the products you deem important.

Creating a Merchandising Rule

From the My Relewise control panel, you can easily create and manage your Merchandising rules.

Creating a rule requires two things to function: The Affected Context of where the rule should apply, and the Target entity that the rule should apply to. A rule can have multiple Contexts and multiple Targets.

Affected Context

The Context denotes where the API call is coming from, and thus where on the site the rule should be applied.

At the broadest level is Currency and Language. Relewise is capable of sorting between revenue streams on the basis of language and currency via the Revenue section of My Relewise, so setting up these types of contexts is only recommended if you have special need to target large swathes of your site at once.

Specific Location lets you target a particular page or page type on the site. This lets you set up rules for recommendations being made on the Product Details Page or the Power Step, or it can let you affect the search weighting of products via the search overlay. Important to note is that Location will affect every request made from the particular page, which means that if there are multiple recommendation bars on a PDP, they will all be affected by the Merchandising rule.

Search and Recommendation Types denote the types of requests being sent. my.relewise.com will generate a list of the request types that the system has registered from your site, which makes it easy to set up rules for the specific request types you are using. This also allows you to debug the requests by checking if a particular request type is actually being reflected in the generated list.

The generated lists of requests are not exhaustive, and it is possible to view all types of requests by clicking the "Show all types" button below each list. Staging a Merchandising rule on a request type not currently in use can be useful for pre-loading a series of rules to features not yet implemented on the site, and lets you anticipate the needs of your solution ahead of time.

Considering Filters for Merchandising Rules

Merchandising rules can be modified to take into account what filters are/are not included in the search or recommendation request. This can be used, for instance, to apply a merchandising rule to a request that filters for a particular category or brand.

To do this, choose the "with filter" tab under Affected Context, and choose the conditions under which the Merchandising rule should apply:

When request is filtered by...

... denotes that the Merchandising rule will take effect only if the selected filter properties are present in the request.

When request is not filtered by...

... denotes that the Merchandising rule will take effect only if the selected filter properties are not present in the request.

When request contains a certain number of filters...

... denotes that the Merchandising rule will only take effect when the request contains the specified number of requests; this can either be a number range, or a minimum/maximum amount.

Example: Applying Merchandising to Categories

To apply a Merchandising rule to a specific category, you should add the "With Filter" option, and specify the ID of the category in question. For our example, we are specifying that the request must include the filter for the category "Streaming Devices":

Click here to see an example of a complete Merchandising rule with filter context

Targeting

Targeting refers to the entities that you want to target with your rule. This can be a particular product, a product category, a content page, a brand ID, a particular product tag, etc. The most commonly used Targeting types are found front and center when setting up a Merchandising rule:

• Match a Product: This filter matches with one or more product IDs. You can search by product name or ID, and a list of suggested products will be presented to you. Each product needs to be added separately, so it is recommended to only use this for rules applying to a few products, or as a Negated target to filter out a handful of products from an otherwise active filter.
• Match Product Data: This filter matches with one or more data keys present on a product, eg. ProductType, InStock, or Vendor. The data key is then given a condition for the rule to apply: To either equal, contain, be distinct from, be less than or be greater than a given value.
• Match a Product Category: This filter matches with one or more product category IDs.
• Match a Brand: This filter matches with one or more Brand IDs.

Advanced Settings

The Merchandising UI allows for the setup of advanced Merchandising rules through the 'Show Advanced' button. Relewise is set up to handle multiple filters for a single rule, in which case all filter conditions must be true for an entity to be returned. This makes it possible to filter out entities from a broader category by nesting filters inside each other.

Negated targets

Negated targets allow you to sort exceptions out of the filters you are setting up. For example, you might want to create a Merchandising Rule that targets all but two products from a particular Product Category. In this instance, you would create a Rule that targets the Product Category ID, and then a second target that Negated the two products you do not want to boost.

Boost/Bury Modifiers

When creating a boost/bury rule, you must specify a value by which to affect the relative weight of the affected entities. Note that the weight by which the boost/bury rule affects the entity is relative. This means that an entity with a high personalization score can still show up after being buried, if its personalization score still outweighs other entities in the results list.

The MyRelewise interface allows you to set your boost/bury modifier in two ways: Either by entering a percentile value, or by referencing a data field on the affected entities.

Boost/Bury by Percentile Value

Setting a percentile value for your boost/bury rule ensures that all entities targeted by the rule will be affected by the same relative amount.

You can either select a value from the dropdown provided, or select the Custom field, which allows you to enter your own percentage value.

To put the percentages into perspective, a product being boosted by 100% will count for twice its normal weighting, and require half as much personalized relevance to be shown towards the top of the results compared to other products. Likewise, a product being buried by 50% will have its relative weight in the result ranking reduced to half, and will require twice as much relevance to be shown towards the top.

Boost/Bury by Datakey

A Merchandising rule can also boost of bury entities based on a datakey held within each entity; to do so, select "by datakey" and specify the name of the datakey found on the entity.

The datakey must be a double, with a value above 0.

When evaluating a datakey in this way, Relewise uses the value as a multiplier for the entity's relative weighting. In other words, a data key value above 1 will boost the product, while a value below 1 will bury it.

Weight calculation examples

A datakey of 1.2 will boost the relative weighting by 20%.

A datakey of 0.8 will bury the relative weighting by 20%

Pin Rule Specifications

A Pin rule imbued with multiple entity IDs will display the entities in order, starting from the position specified in the pin rule. As such, it is not possible to pin multiple entities to the same position and have them alternate between page refreshes; the list of products will simply start at the specified position, and occupy the subsequent recommendation slots in sequence.

To set up a pin rule, you must specify the entity ID(s) that you wish to pin. You cannot pin entities based on a datakey - which means that you cannot pin products on the basis of product category or similar.