The do’s and don’ts of Product Taxonomy

Launching a website, changing your back-office solution or simply reviewing your current organisation, there are many moments in a company life stage where giving a bit of thought to your product taxonomy will help you enhance your future performance and eventually save a lot of time in the future.

For those of you who haven’t read it yet, you can find out about the perks of a robust and well-maintained product taxonomy right here. After that previous publication, we’ve received a lot of emails asking what would be the best practices for product taxonomy, how to build an efficient category tree or how to differentiate between what’s category material and what should be treated as an attribute.

First of all, a bit of vocabulary about product taxonomy. There are three main tools :

1. The category tree (also known as product hierarchy): it helps group products by nature in a set of hierarchized categories.

Example: Home > Home Appliances > Kitchen > Cooking > Rice Cookers

2. The attribute list (also known as facets, dimensions or refinements): additional information attributes, used to qualify the products in a same category. When applied to a product, an attribute will require a value.

Example: Bowl capacity, color, type of sole …

3. The value list: for some attributes, the finite, predefined set of possible values, for Booleans (yes/no) or rich attributes

Example: Blue, 1.5 litres, 220V, strong, …

Each attribute will apply to several categories, and have various values. Values are linked to one and only attribute.

Understanding the relationships between those three tools is essential to the understanding of taxonomy design. What will follow is a set of advice based on our experience and research and intended to help the would-be and the seasoned taxonomists in their designing or revamping of product classifications

How many is too many ?

In our opinion a robust and understandable category tree should have a hierarchy depth of 3 to 5 category levels, and between 5 and 15 level one categories. This is of course dependent on the width of the assortment you are selling and your product normalization.

For the attributes, and the values, there’s no limit! The more you have, the more accurate product description will be.

A simple category hierarchy

Where should I start ?

Check your inventory and the products you sell. That should give you a good view on how deep your category tree should be. If you are running a physical store and wish to go online, do not necessarily replicate the organization of your store on your website categories. That might help the loyal customers, but the whole point of an e-commerce website is to go beyond your customer base and reach new customers. The use of physical space and the use of screen space are very different, as well as the customer behaviour online and offline. Remember that the categories will help your customer find your product, through navigation or search. Attributes will help the customer access in-deep information about the product.

How should I name my categories and attributes ?

Make it:

• Short
• Self-explanatory (at least within your organization and for your target customer)
• Unique (as much as possible)

On the other hand, Attributes are to be re-used across the whole category tree and are stand alone. They are never too accurate, so don’t hesitate to have “Diameter of watch dial” and “Diameter of bicycle wheel”, rather than a global “Diameter”. That will strengthen your data consistency (maybe not all diameters are commonly expressed in the same unit), enhance the ease of use of people within your organization, as well as improve clarity for customers.

One thing we see a lot is a “Other” subcategory in almost every category. We strongly advise against this. Customers are never searching for “others”. While it might seem convenient for products not fitting in an existing category, it is always better to create the adequate category rather than having a messy and vague category. A category is never too small or too specific.

So what’s attribute material?

There is often a question about whether a difference in product nature should be treated as an attribute or a subcategory :

Example: should I have a category for denim shirts and one for jersey shirts, or a shirt category, with the attribute fabric = “denim” or “jersey”

The answer often comes with a bit of common sense and experience of the e-commerce industry (category = product class that can stand by itself, attribute = descriptive element that can be used on several product types), but there are a few rules:

• If you can re-use the qualifier on other categories (“denim jackets”), it makes sense to define it as an attribute (DRY = Don’t repeat yourself !)
• It all boils down to having one additional attribute or one additional level of categories: try to balance between those two lists
• Try to understand what your customers are looking for: would they consider denim a subcategory of shirts (i.e. they would look for denim shirts exclusively) or an additional information on the product ?

Should products appear in more than one category ?

Offering the same product in several aisles of a physical store or various categories of a website can enhance the visibility of this product and thus its sales.

For instance a Ski jacket could be in Sports equipment and also in Apparel>Jackets

Sometimes a product can be displayed also in a special category (“offers of the week”, “Best deals”…)

However that has to remain a front-end or store display option only. In your product management system or WMS, a product cannot be in more than one category. Otherwise, there are many risks: double-counting at reporting, mix-up in the stock management, inconsistencies in your purchase orders…

Most e-commerce software allow to display a product on several categories online, but this should be seen as a marketing tool only, and not infer with the product back-end management

What level of standardization should I apply to my values ?

Once again, choosing to standardize values for an attribute (= all the values for this attribute can only be picked in a pre-defined, finite set of values) is often a tough choice, mixing flexibility, IT constraints, ease of management of the product list…

We’re not going to deny that there are perks to a strictly standardized value system, among others:

• IT-wise, it helps ensuring the quality of the data appearing online
• It can help to do cross-category and cross-attribute reporting or filtering
• It keeps the consumer in known grounds and will be less confusing for him (I know what Pink is, but what colour is Amaranth ?)

That might sound appealing, but one shall not forget the constraints:

• Having to reformat/”translate” all the information received from your suppliers can be extremely time-consuming if you don’t have the right tools (Scalia can help !)
• You might be missing slight nuances in meaning
• Forcing your system to accept only some values can, and will, lead to issues at product creation: products that will not update, system rejections…

We believe this has to be done on an attribute-by-attribute basis. It boils down to the number of different values for an attribute. Managing a set of 8 or 10 shirt sizes is achievable, but some attributes just don’t do very well with fixed values even though it might seem possible to standardize, simply because there are too many possible values (Brand, Color…)

There are no strict rules to product taxonomy. However, the various elements listed in this article need to be thought of beforehand to make sure your product taxonomy will be able to help your business grow and never represent a hinder to your daily management. Having the right tools to help you manage and interconnect your taxonomy with your business partners is also a wise choice.