personalisation-taxonomy.jpg

Personalisation - Taxonomy

Personalisation is only as good at the taxonomy. Without accuracy when tagging content, outcomes are distorted and become counter-intuitive.

6 minute read

Introduction

Personalisation is the practice of serving users with content relevant to them, based on traits including their visit behaviour, location, device and interaction history. Also known as “Context Marketing.”

What is Taxonomy?

In Information Architecture(IA), taxonomy refers to descriptive identifiers (attributes), or categories/sub categories, applied to content stored in a system. Taxonomy is used for establishing relationships with other content, and for content differentiation; In other words, categorisation. Taxonomy is traditionally applied to data using a tagging mechanism, automatically in some cases, with attributable traits of the data selected/applied as “tags.”

Amongst other opportunities, tagging content facilitates Faceted Search, which is entirely reliant on a correctly administered classification scheme. For example - when users are browsing sites, when performing content searches, interacting with filtering mechanisms, or when receiving personalised content delivery, tags are “looked up” and evaluated by the application. Typically, a system will process inputs, such as search filter choices (facets) or search(ed) terms by comparing them to content which has matching taxonomy (tags).

E.g. If the user has selected the following filters: [“Oranges,” “Small,”] then show articles which have been tagged with these attributes.

Truly extensible systems provide the facility for content authors to add/remove their own, unlimited, taxonomy categories; even further, ranking systems can be introduced for creating multidimensional taxonomic relationships.

In multi-faceted search, using this website as an example:

1. Taxonomy classifications are abstractly created

2. The classifications are associated with pieces of content relevant to the subject matter

Tagging 

3. There is now a relationship between content which share attributes, and a differentiation between content not sharing attributes

4. The article filtering mechanism is constructed from available taxonomies (categories), with the tags becoming the options. In this example, the search facets are completely extensible.

filters

5. When selected, singly or combined, the application is able to associate content based on its taxonomy, providing search results

 

This is a typical implementation of “taxonomy driven multi-faceted filtering.” Authors are, in fact, able to create primary and secondary filters for greater granularity, but that’s a nice to have.

How personalisation and taxonomy go hand in hand

Tagging content with attributes has been around for longer than personalisation, e.g. The example above.

By and large, personalisation is unable to effectively function without a taxonomic content scheme.

Taking the example of taxonomic search, or multi-faceted search; the user makes filter choices as to which types of content they would like to see (eg. “Sitecore, Umbraco, JavaScript”). They are actually selecting different taxonomies and all the system is doing is filtering results to those that match the tags. If there was no taxonomy applied to content, or if the taxonomy was applied incorrectly, the results would be null or skewed.

When selecting personalised content for users, the application does this same filtering exercise, exactly as above and behind the scenes, but automatically based on taxonomies which the system has determined to be appropriate for the user. The same applies, if taxonomy is inaccurately assigned or non-existent, the results will be of no use.

Sites with effective taxonomic / multifaceted search:

http://blog.mattoneill.london/

http://www.autotrader.co.uk/

https://www.ubereats.com/san_francisco/ 

In summary

Without properly tagged content taxonomy, the system cannot accurately respond to the information it has about the user. For example, given that the user has been looking at a certain type of restaurant on a review website (eg. Greek, Mediterranean), content authors may have configured the system to show the user related restaurants. In order for the system to be able to do this, it will have the criteria as tags (Greek, Mediterranean) and it will try to search for other content which has been tagged with this criteria. If content is not tagged responsibly and accurately, personalised/suggested results will similarly be inaccurate and ineffective.

 


  • Personalisation
  • Digital Delivery
  • Umbraco
  • Digital Marketing

Leave a comment

Submit