Categorizing essentially consists of grouping like items together (classification) and assigning good labels to those groups (labeling). If it is going to be effective, categorizing needs to be done according to sound principles. Alternatively, if something is not categorized well, it is probably not understood well and will likely not communicate well.
This document outlines the key principles for how to categorize. It provides the framework to follow when grouping our content by topic (see the Topic Listing document for a list of our current topics and The Nature of Our Topic Index for a summary of the purposes and policies for our topic index) or into various other category groupings (such as those defined in the document Desiring God Classification Schemes).
Why Categorizing Matters
Categorization matters for several reasons, especially on websites. First, good categorization reveals site content. For example, thousands of site pages become more manageable when “summarized” by 8 primary categories. Second, good categorization reveals the structure of the content. Categories provide a framework for understanding and aid exploration because similar things are grouped together. Hence, presenting people with a logical set of categories that reveals the contours of the content is actually a form of teaching—a way of informing people how to think about and interpret what they find and associate it with other content.
Third, good categorization enables ease of use. This is especially important given the nature of our website. Our website is, in essence, “The Works of John Piper”—comprehensive and ever-growing. Therefore good categorization matters, as do consistent titling conventions, grammar, and so forth. Through good categorization, our site can be the complete Works of John Piper, easily navigable. Fourth, good categorization expands people’s horizons by exposing them to terms and groupings they would not have thought to search for—and probably couldn’t search for well, if they tried.
The Principles of Good Classification
Classification needs to follow solid principles if it is going to make sense and be useful. The characteristics of good classification, or grouping, are as follows:
- Follow a classification scheme. Identify the common trait that you are using as the basis of your grouping. This is the classification scheme, or organizing principle—that which “defines the shared characteristics of content items and influences the logical grouping of those items.”[1] Classifications without a clear, definite, and consistent organizing scheme are confusing and reveal lack of clarity and understanding. See the document Desiring God Classification Schemes for a list of our major classification schemes.
- Do not mix classification schemes. Multiple schemes can be applied, but they cannot be mixed. For example, you would not want to use this category structure: CDs, MP3’s, and Christian living. The reason is that the first two items are product categories, while “Christian living” is a topic. You could subdivide CDs, however, by topic, or provide “Browse by Topic” or “Browse by Product Category” as two alternate means of browsing.
- Make the categories mutually exclusive. Categories should be mutually exclusive as much as possible. Otherwise, classification looses much of its value.
- Make the categories unambiguous. It needs to be clear what the categories pertain to, especially since ambiguity in effect works against mutual exclusivity.
- Eliminate redundancy. There should be no redundancy in labels, sections, and categories. For example, we should not have both “effectual calling” and “irresistible grace” as subject categories. If both terms are common, then by the least common one put a “see [name of the other term].”
- Keep cross-listing to a minimum. This means that each item should be in only one category. This follows from the principle of eliminating redundancy.
- Use “related items” sections to show people the full network of connections among resources. When an item seems to “fit” in more than one category, list it according to its primary category and designate it as a related item in other categories. This enables the item to be found, without making the user have to mentally sort each new category they go to in order to “screen out” the redundant elements.
- Make the categories comprehensive. There should be no gaps. The categories should be exhaustive, without being redundant.
- Seek to keep lists between 5 and 9 items. Groups of similar choices should generally be limited to 5-9 items; people can’t keep more in their mind for evaluation. If tons of links are needed, cluster them in groups of 5-9 links if you can. Of course, this is not always possible.
- Make the groupings intuitive. The customer should not have to work hard to figure them out. One way to accomplish this is through user testing. In user testing, the key is to observe how something works with them, not necessarily to incorporate all of their suggestions.
- Make the groupings logical. For the groupings to be logical means that they reflect the actual structure of the content. Since one of the purposes of categorization is to reveal the structure of the content, this principle is very important. At times, it may seem in conflict with making things intuitive. But in general, that which is logical is what is intuitive, or what will be found to be most intuitive upon use.
- Create the categories to anticipate future content. It should be possible to add additional content without breaking the category structure.
When categorizing by topic, there are two specific principles that apply on top of the above principles:
- Our topics should be universal to our organization. For example, we should not have one set of categories for products (online store), and a different set for resources (online library).
- Our topics should be historical. We should seek for the categories to be informed by traditional categories of theology and the Christian retailing industry.
The Principles of Good Labeling
Categories that are designed well need to be labeled well. Many of the principles of classification imply as their corollary certain principles of labeling. Good labels are:
- Short.
- Clear.
- Intuitive.
- Jargon-free.
- Not cute or clever.
- Consistent in style, punctuation, presentation (fonts, sizes, colors, white space), syntax (they do not mix of verb-based and noun-based terms, they agree in parts of speech, verb tense, and they have roughly the same number of words), and granularity (they are all roughly equal in their specificity).
Classification Methods
The classification type depends upon whether you “group” your categories or just list them all without groupings. To group your categories underneath larger categories is called the encyclopedic method. To simply list them all without major groupings is the A-Z method.
Both have benefits and drawbacks, and the choice of a method depends on the purpose. The best example of an A-Z method is the phone book. This method works best for known-item-searching. If you know you need a plumber, for example, you just want to be able to flip to a section called “Plumbers.” You don’t want to have to go to a major category like “Household Services” to find plumbers grouped with electricians and carpet layers.
But for non-known item searching, the encyclopedic method can be useful, because it groups like categories together. For example, if I am interested in the field of theology, I may want to see all theological topics grouped together—rather than having to sort through a lengthy list that includes many other types of topics, in the hopes of being able to identify and remember all the ones that pertain specifically to theology.
Classification Structures
When following an encyclopedic format, there are two main types of classification structures: the generic relationship and the whole-part relationship.
In the generic relationship structure, the relationship between the super and subclasses is always “IS-A.” An example of such a structure is:
- Automobile Companies
- Ford
- Toyota
- General Motors
This is an “IS-A” relationship—Ford, Toyota, and General Motors are all types of automobile companies. According to “The Truth About Taxonomies,” an article in the Information Managmeent Journal, other characteristics of this type of structure are:
- Inclusiveness. The top class includes the subclasses. Everything beneath it is a type of it.
- Inheritance. Everything true of the given class is also true of all the items in its subclass.
- Transivity. All subclasses are members of every class above them.
- Systematic rules for association and distinction. All entities in a class are like each other in a predictable way.
- Mutual exclusivity. Each item can belong to only one class.
In a whole-part relationship, each item of the subclass is a component of the super class. An example would be:
- Airplane
- Fuselage
- Engine
- Wings
In this instance, the fuselage, engine, and wings are all parts of an airplane. You have a whole-part relationship.
If you mix classification structures, you often have confusion. For example, you would not want to create this structure:
- Airplane
- Fuselage
- Engine
- Wings
- Northwest Airlines
- 747
- DC-10
As can be seen, the problem with such a structure is that it does not adhere to a consistent classification. It mixes “Is-A” and “Whole-Part.” And among the “IS-A” relationships, you actually have two different kinds—Northwest Airlines is a type of airline, whereas the 747 and DC-10 are both types of airplanes.
Other Resources
This document provides the foundation for the two other documents in this series, Desiring God Topic Categories and Desiring God Classification Schemes. Some helpful external resources on classification are the following:
- “The Truth About Taxonomies,” The Information Management Journal (March/April 2003), 44-53).
- ECPA Christian Product Categories, 2003.
- Establishing Alphabetic, Numeric, and Subject Filing Systems, ARMA International, 2005.
- Information Architecture for the World Wide Web, Louis Rosenfeld & Peter Morville.
See also the other DG documents pertaining to classification:
- Desiring God Classification Schemes
- Desiring God Topic Listing
- The Nature of Our Topic Index
[1] Information Architecture for the World Wide Web, 55.