Getting Started with Ranking Rules

Best practices

To ensure you are using ranking rules efficiently, you might want to explore and apply these recommendations.

Maintain at least two default ranking rules

Ranking rules determine the sorting order of items on lister pages. If no ranking rule triggers for a query, Fredhopper lists items using natural sort. Natural sort is a sorting order determined during the indexing of your data. Because it relies on the indexing process, it might vary between each reindex. You cannot control natural sort.

To avoid sorting items by natural sort, you need to maintain default ranking rules. You need one for lister pages shown after navigation and one for lister pages shown after search.

When your Fredhopper team configures Fredhopper for the first time, they will create default ranking rules for you. However, you might need to re-create or change them or add more defaults. In this case, follow these guidelines:

• Name your rankings 'Default navigation ranking' and 'Default search ranking'.
• Label them as default.
• Use the Result type trigger for each, and set it to Navigation or Search, as needed.
• Use a ranking that works for your product types and shoppers across all categories. For example, sort by relevance, margin, featured items, or new arrivals.
• Avoid sorting by brand for your default, because this strategy of ranking/sorting may change over time.
• Avoid sorting by top-rated for your default, because this strategy of ranking/sorting may change over time.

Maintain category ranking rules

Different categories contain items with different characteristics, some of which might be category-specific. Consider maintaining additional ranking rules for your categories that reflect the unique purpose of the category.

For example: in a Clearance category, sort by 'percent off' in an Apparel category, sort by new arrivals; in a thematic category, sort by relevance or featured.

To create a rule that triggers for a category, you can use a User location trigger.

Use labels

In a lengthy list of rules, labels help you find the rule you are looking for more quickly. Make sure that you have a concise list of labels and a clear naming convention for new labels. Establish internal guidelines for labelling and apply them.

In compound ranking rules, start with the attributes with the fewer possible values

In ranking rules with several sorting actions, begin with the attributes with the fewest possible values, i.e. an attribute with only two possible values. This increases the odds that the next sorting action will impact the sorting order.

Example

This example shows how attribute values influence the sorting of items.

Table 1: List of unsorted items and their respective attributes

In this case, all attributes affect the sorting.

In this case, only the first attribute affects the sorting.

Use numbers instead of text to flag items

Flag attributes let you mark an item as having or not having a characteristic. You can later use the flag to filter items when you present them to shoppers.

You can create a flag with text. For example, yes and no, or true and false values. As an alternative, you can use numbers. For example, 1 signifies yes or true and 0 signifies no or false. If you use numbers, you can then use the flag attribute in ranking rules and ranking cocktails.

Use caution when ranking by historical data

In some cases, past data might lead to continuously promoting or demoting the same items. In others, it might erroneously promote some items over others.

When you use historical data, choose relative attributes and limited time frames. For example, sort by average gross sales per day instead of an absolute number of sales for all time.

Example 1

You have a ranking rule that sorts items by page views. Items with the most page views appear first.

In this example, newer items always appear on the last pages because they don't have page views. Furthermore, items at the beginning of the list usually receive more attention. As a result, items with many page views continue to gather page views and appear at the top.

Example 2

You have a ranking rule that sorts items by average user rating. The average user rating is calculated as the sum of all ratings divided by the number of users who submitted ratings. Items with the highest scores appear first.

In this example, an item with just one or two very high ratings (5/5) might appear before an item with a lot of reviews but a lower average score (4.9/5).

If available, use predictive data

If you have access to any predictive data, use it in your rankings. This lets you create dynamic lists of items tailored to shoppers and their history.