I'm using release dates. How can I prioritise newer products?

This problem can be solved by using record ranking. There are two solutions, depending on the information contained in your product database.

1. Your database contains a “new” field

The new products in your shop have a particular attribute. Create a rule that slightly reduces the similarity value of all products that do not have this attribute:

Rule 0: Field [new] is not equal to ”true”: Reduction 1%

With this reduction, a search query for “Harry Potter” will always return the current book first (assuming the current book is marked as new and the older ones are not).

2. Your database contains a publication year/date

If you include a publication year in your product data, you can achieve even more precise results than in the first example. The options here depend on how you record this information.

If your database has a separate data field for this information, you can apply either a sales rank or time rank rule. If the year of publication merely forms part of another field (such as the article name) then you will have to use text comparison rules.

2.1 The publication year or date is held in a separate field

If the date is in a field of its own, you can apply either a top-seller or time-based ranking rule. In the case of a top-seller rule, the field in question should contain a purely numeric value. So if the field contains only the year, then this can be resolved easily. However, if other formatting characters are also used, you cannot apply this sort of rule.

A top-seller rule is intended for use in reducing the value of a range of figures, either linear, logarithmic or in fixed steps.

For example, you can define the following rule:

0% if publication year = 2010, to -4% if publication year = 2000

This means that all articles from 2010 and later are not reduced in value, while articles from 2000 and older are reduced by 4%, with a linear trend for all values in between. So for example, an article from 2005 would have a reduction of 2%.

If the field holds an actual date or timestamp, then you can also apply a time rank rule. This calculates the reduction in value on the basis of the actual current date. In contrast to the topseller ranking, it also allows you to define and subdivide multiple reduction ranges. For example,
you can define the following rule:

Time rank for publication year

Now - 1 year up to -0%
1 year - 3 years up to -2%
3 years - 10 years up to -4%
10 years - older -4%

This rule means that products younger than one year receive no reduction in value. If a product is between one and three years old, it receives a one to three percentage point reduction (linear distribution of the reduction over the period), etc. The advantage of this variant is mainly the dynamic adjustment, which means that you do not have to change the rules each year, and the fine granular configuration options.

2.2. The publication year forms part of another field

Create a rule for each year. The similarity value of the current year’s products is not reduced. The longer ago the publication year is, the greater the reduction is applied to the similarity value in the results.

Rule 0, Group A: Publication year contains ”2010”: Reduction 0%
Rule 1, Group A: Publication year contains ”2009”: Reduction 1%
Rule 2, Group A: Publication year contains ”2008”: Reduction 2%
Rule 3, Group A: Publication year contains ”2007”: Reduction 3%
Rule 4, Group A: Publication year contains "200": Reduction 4%
Rule 5, Group A: Publication year does not contain "200": Reduction 5%

In the “Harry Potter” example, this set of ranking rules will sort the results in chronological order. The disadvantage of this approach is that the values are static. You would need to update the rules each year. In 2011, 2010 becomes the previous year and will need to be reduced in value by 1%.