Part 2: Your Product Taxonomy Sucks (And Here’s HOW You Can Fix It)

Part 2: Your Product Taxonomy Sucks (And Here’s HOW You Can Fix It)

By Jason Hein, Principal Strategist at B2X Partners

With any new initiative comes a learning curve and occasional “learn by failure” moments, and eCommerce is no different. Selling products online involves a different set of skills for every step in the customer journey. Companies use different techniques to help customers discover, research, compare, and purchase products. So if you feel bad about your product taxonomy after reading the last article, you really shouldn’t – it’s perfectly natural! Today we’ll cover techniques to address the five problems we discussed earlier and start putting your taxonomy on the path to success.

So, your product taxonomy sucks.
Here are the top challenges, and how you can start fixing them.


Duplicate Categories

For many companies, one of the challenging aspects of this problem depends on the tools available to manage their taxonomy. The best way to find redundant categories in a taxonomy is to view the entire taxonomy at once. Some firms attempt this simply by clicking through the categories on their site, but that approach is the proverbial “needle in a haystack”. Using your site for find duplicates only works well for reviewing “sibling” categories – and it falls to the tester to remember where they may have seen similar categories in other nodes.  When managing large or broad taxonomies, this method quickly becomes ineffective.

This is where tools like a Product Information Management (or PIM) system can come in very useful.  Almost all of them provide the ability to view or export the entire taxonomy and make it easier to view more of the taxonomy simultaneously or let you run simple queries against shared keywords in a simple tool like Excel. Some PIMs have governance functionality that can check for potentially redundant categories at the time they are created.   However you approach it, your ability to resolve duplicate categories depends on your ability to view your taxonomy as broadly and completely as possible.


Too Many Choices

A similar problem to duplicates, the issue of having customers choose from more than 15 subcategories at a time is slightly easier to solve. Certainly, if you have the ability to review the full taxonomy, then this issue is one that can be watched for in parallel. However, in this case, you can also spot your worst offenders by simply “walking the store” on your website. As you click through your taxonomy, simply count how many choices of subcategories you are asked to select from at each point. If you have more than 15 (or even more than 13 in most cases) consider changing the structure to simplify the approach.

Once you have identified cases where too many choices are offered, there are three approaches you can try to resolve them.  First (and perhaps the most obvious) is to make sure none of the options are redundant – are all the options you are offering mutually exclusive? If not, then merging categories is appropriate. Second, you can employ “grouping categories” – inserting a level of your taxonomy between the parent category and the 15+ subcategories you offer.

See an example here.

Without grouping categories:

With grouping categories:

As you can see, it is easier (and usually faster) to choose one of three options and then one of 5 options than to search through 15 options all at once.

Another option is to use faceted search (also called “attribute filters”) to find ways to let customers refine their results using features or aspects of products that are not managed in taxonomy. If your taxonomy offers different categories for each material variety (e.g. “Brass Cap Screws”, “Stainless Steel Cap Screws”, Galvanized Steel Cap Screws”) consider just one node for “Cap Screws” and letting customers filter selection using a “Screw Material” attribute. Faceted search may not for all cases, but it can easily be used in combination with other solutions, and customers often engage with it regularly when offered correctly. And discovery tools that customers use are often ones that boost conversion and sales.

See an example of facted search on the left column from McMaster-Carr:


Missing Categories

Generally, at the time a taxonomy is built, Missing Categories are often not an issue – the people who build them have typically done considerable research and therefore most significant selection is accounted for.  But over time, businesses evolve – manufacturers invent and launch new kinds of products, and then those products become popular and eventually dominate the market, supplanting other categories that slowly scale down and eventually are obsoleted. Because markets evolve, you have to make sure your taxonomy keeps pace with the latest developments.

Start by asking – where do we hear about ideas for new products first? Is it your suppliers? Your sales people? Perhaps your customers? How can you leverage your access to those sources to help you avoid missing categories as new products come online? Can you schedule a call with your manufacturer’s rep to hear about new products in their industry? Can you have your best sales people review the taxonomy for missing categories? Do you have a way for customers to tell you what is missing? Are you reviewing reports for popular keyword searches that have zero results?  All of these options can provide valuable insights into what’s missing quickly, and usually with little incremental cost.

One other option to quickly find new categories is to review and critically assess the content of your “junk drawer” categories and look for certain product types which have enough representation in those nodes to warrant being broken out into a more specific node just for that kind of product. Of course, this brings us to…


Junk Drawer Categories

One of the first things to be done when a junk drawer category is authorized is to establish a metric for the maximum number of SKUs that will be allowed in it. Once the SKU count hits that target number, an audit must be done and the designated owner of that node must either move some products out (potentially to new categories) or set a new, higher-SKU count number for the next trigger. If the decisions are marginal, then the owner can establish guidelines for what is the maximum number of SKUs for a given type of product can be stored in a junk drawer category before they warrant a node of their own.

The critical aspect is that these categories must not be allowed to grow on their own without any supervision.  Each junk drawer category must have an assigned (human) owner who is responsible for knowing what is in that node and whether it is appropriate. Establishing this approach early (and therefore breaking down the maintenance associated with junk drawer nodes into monthly (or even weekly) tasks helps keep the workload manageable as well as keeps the selection in those nodes relevant.


Misclassified SKUs

Of all the issues with product data, this one can be the most difficult to address. Particularly in cases where automation is a factor and large numbers of SKUs are being loaded regularly, if a SKU is loaded into the wrong category, but nobody sees it (neither an employee or customer), then did that SKU really get loaded? Products in the wrong categories are effectively marooned, meaning relevant product can be hidden from customers. But the difficult part of this issue is – how do you know which SKUs are misclassified without looking at all your SKUs? A complete review of every product in your catalog can be an expensive and time-consuming project, therefore most firms are reluctant to do it. But there are some tactics you can try that are simpler and that (to some extent) scale up.

Use your keyword search tool

First, you may be able to use your keyword search tool as a rudimentary classification checker. Often, the results of keywords are displayed in the context of the categories where the SKUs in results are listed. For example, a search for “threading dies” on the Amazon Business Site returns over 1000 results across 16 different categories of their taxonomy. And while some of them are appropriate, there are also clear errors of Threading Die SKUs being in categories such as “Abrasive & Finishing Products”, Fasteners”, “Hydraulics, Pneumatics, & Plumbing”, etc. – all of which are places where that selection should be reviewed as a potential misclassified selection.

Use the data governance functionality within your PIM

Another option for finding and fixing misclassification issues is to leverage data governance functionality within a PIM tool.  When implementing a PIM, it is important that you design a Data Model, which defines which attributes are relevant (or even required) for all SKUs loaded into a given category. PIMs not only store that data, but they usually also provide the ability to automatically review whether the SKUs loaded into a category follow that data model. Does each SKU have values populated for the “required” attributes? Are they populated with values that meet the model? If not, then those SKUs with serious defects can be generated as a report of potentially “Defective” SKUs where one of the potential solutions would be to relocate them to a perhaps more suitable category.



Regardless of how well you think you designed your taxonomy on day one, eventually it’s going to fail. And that’s a *good* thing! The failures you find are the ones you can fix, and as your site attracts more users, you’ll get more data on how well your taxonomy is performing which you can use to find new problems and the process repeats itself. With every new thing you learn about how customers use and interact with your taxonomy you’ll improve your site’s ability to create a unique experience for your customers, providing the easy path for customers to find, research, compare, and purchase new products that they would not otherwise have known about.  And that ability to be a trusted, easy-to-do-business-with source of *information* as well as goods, is what will differentiate you from the disruptors and help grow your business online.


About Jason Hein

Jason Hein is an eCommerce expert and product strategy leader. He brings over 20 years industry-leading experience in industrial distribution, including at Amazon Business and McMaster-Carr. At B2X Partners, Jason works to expand content strategy and develop world-class processes for rich product data and optimized product merchandising.