This is the sixth article in the series ‘The Art and Science of Taxonomy Development for Market and Competitive Intelligence’. Click here to view other articles.
S tart with the end-goal, the decisions to be made, problems to be addressed, tasks to be performed with the help of your taxonomy, and write it down. Align this end-goal with the needs and priorities of your organization. And then agree on this goal or purpose with all the stakeholders. Writing it down is important. Without it, there will be endless discussions and arguments about what terms to be included or how they should be structured. A documented purpose lends a framework (a single source of truth) to enable subjective decision-making with consistency over time.
To build your taxonomy, group the terms that were selected in the previous phases, and put them in a formal structure that is aligned with your use-case. The terms can be arranged in a simple list of terms or a complex system of hundreds of terms with interlinked relationships between them.
Define the purpose; Do not attempt a be-all and end-all taxonomy
Methods for building a new taxonomy
There are two prominent methods of building a new taxonomy — A Deductive (bottom-up) and Inductive (top-down)
1. Deductive method (Bottom-up)
Selected terms are reviewed by users or subject experts, either individually or as a group. They first identify terms that represent the broadest category and then allocate the remaining terms to these categories on the basis of their relationships according to the use-case, such that the hierarchies are established on a broader-to-narrower basis. Maintain, as much as possible, a consistent level of specificity within a category. For example, if one of the categories ‘Suppliers’ has a narrower hierarchy as ‘Component Supplier’, then other categories should go to the same level of specificity. Such consistency in selecting terms and defining hierarchies helps in setting the right user expectations and help them intuitively navigate their way through the taxonomy. This is a subjective area, therefore, it has to be a part of the taxonomy governance document, with rules and policies explicitly spelled out. One common technique for the deductive method of taxonomy building is ‘Card Sorting’.
- Card Sorting
Card sorting is a technique in user experience design in which a group of users (or domain experts) generates a category tree. All the terms selected for the taxonomy are written on index cards or post-it notes. The users then arrange the cards to represent how they see the taxonomy structure and relationships between the information. The same idea can be used in building business taxonomy also.
Tip: Allow users to suggest new terms. There could be other forms and variations of the terms (synonyms) that can be used to represent the same information.
2. Inductive method (Top-down)
This method is recommended when you need to develop a taxonomy without the privilege of having access to historical content for reference. For example, developing a taxonomy for a brand new website or internal project. In this situation, we don’t know the kind of information that will be indexed or how the taxonomy will be used. In this top-down method, taxonomy development is regarded as a continuous process from the onset. New terms are selected for potential inclusion in the taxonomy as they are encountered in the content. Each term, as it is admitted, is designated as a member of one or broader classes that are constructed on an ad-hoc basis. The taxonomy is therefore established on a narrower-to-broader term basis.
In our experience, we found that a hybrid of both, top-down and bottom-up, methods is the most useful approach to building a new taxonomy.
Types of Taxonomy
Taxonomies can be of different types — flat, hierarchical, network, etc. In our experience, a hierarchical structure is most suited for market and competitive intelligence. There are three kinds of hierarchical structures:
1. Standard Hierarchy
Standard hierarchy taxonomy consists of a single root, which is subdivided in categories as necessary to structure the information and is often represented as a tree. It is easy to understand a single hierarchy structure and develop a mental model to find information. The terms in taxonomy have three kinds of relations with other terms:
- Parent (or broader) term, for example, ‘Automobile’ is the parent of ‘Autonomous Cars’
- Child (or narrower) term, for example, ‘Sensors’ is a child of ‘Autonomous Cars’
- Equivalent (or synonyms, may or may not exist), for example, ‘Radar’ and ‘Lidar’ are equivalent to ‘Sensors’
A term in a taxonomy can be repeated in different categories. This is referred to as a ‘polyhierarchy’. For example, the term ‘Legal’ can be placed under ‘Regulatory’ as well as under ‘Business Challenges’. Polyhierarchy makes it easier for the user to tag (not retrieve) a term with the content. It is not recommended because, sometimes, it also causes confusion. For example, if the content is related to ‘Legal’ matters, it may or may not be related to the ‘Business Challenges’. It could be an update to the legal requirements by the regulators of the industry. Polyhierarchy is against the first and second principles of naming taxonomy terms: “Terms should be unambiguous and mutually exclusive”. It also disregards another important principle for selecting terms for your taxonomy. That is, users of taxonomy are more important than the kind of information it will organize. When there are trade-offs to be made, design for the end-user, not for the ease of tagging.
Standard hierarchy taxonomy is a top-down single dimension structure. But we have already established earlier that information is multi-dimensional. Therefore, a better structure is required to represent the content. A faceted classification system is multi-dimensional. It consists of multiple taxonomies or ‘facets’, whereby the top-level node of each represents a different type of taxonomy, attribute, or context. For example, one facet (taxonomy) could be ‘Competitors’, and another could be ‘Topics’ such as ‘Partnerships’. By combining the tags in the two facets, we can find partnerships by anyone (or more) competitors.
In our experience, some of the standard business taxonomies in the market and competitive intelligence domain are:
- Key Accounts
- Document Types
The standard hierarchy model asks the question, “Where do I put this?” In contrast, the faceted approach asks the question, “How can I describe this?” A business taxonomy shouldn’t be too deep, which frustrates users with too many clicks, or too broad, which overwhelms users with too many initial options. It should be easy to scan and the user should reach the desired information in as few clicks as possible. Based on our experience, top-level facets shouldn’t be more than 10 – 15; and each facet shouldn’t be more than 2 – 4 levels deep. Therefore, faceted hierarchy is our recommended structure of choice for organizing market and competitive intelligence. It should be accepted from the outset that some decisions regarding the terms and their interrelationships may have to be revised as the users start using the taxonomy. How the taxonomy will be maintained and changes will be incorporated, therefore, is a part of the overall taxonomy development.
Next, let’s look at the guidelines for taxonomy maintenance.