Quick answer
A taxonomy is a system for classifying things into named categories, usually arranged in a hierarchy and drawn from a closed, planned list of terms. In analytics it governs event names, traffic sources and content categories.
What a taxonomy is
A taxonomy is a system for classifying things into named categories, usually arranged in a hierarchy from broad to specific. Nielsen Norman Group defines it for content work as a closed list of acceptable terms, arranged hierarchically, used to describe and classify content (NN/g). The idea started in biology, where the ranks run from domain and kingdom down to genus and species, and it now covers anything you need to sort: documents, products, support tickets, analytics events.
Two properties make a taxonomy different from a pile of tags:
- Hierarchy. Concepts have parents and children. Going deeper makes them more specific.
- Controlled vocabulary. The terms are decided in advance. People tag with them and cannot invent new ones on the spot.
The taxonomy usually works backstage. Visitors do not see it directly; they see navigation and filters built from it.
Taxonomy, ontology, folksonomy and IA
These words are often used as if they meant the same thing. They do not.
| Term | What it is | Relationships | Who decides the terms |
|---|---|---|---|
| Taxonomy | Hierarchy of categories | Parent and child ("is a kind of") | A planned, closed list |
| Ontology | Model of a domain and how concepts connect | Many kinds (part of, causes, located in) | Specialists, per domain |
| Folksonomy | Free tags added by users | None formal | Whoever tags |
| Information architecture (IA) | How content is structured and navigated | Site structure, labels, navigation | Design team |
| Data dictionary | Definitions of each data field | One entry per field | Data owner |
NN/g makes the same distinction between a taxonomy and an ontology: a taxonomy relies on parent-child relationships, while an ontology supports many relationship types and is typically a larger project for one knowledge domain. A data dictionary describes fields; a taxonomy organises the values or names.
Types of taxonomy
- Hierarchical. One tree. A product sits in exactly one place: Electronics › Audio › Headphones. Simple, but forces a choice when something fits two branches.
- Faceted. Several small hierarchies, one per attribute (type, price band, colour), combined by filters. A pair of headphones can be "Audio", "Under $100" and "Black" at once.
- Flat controlled list. A single list of allowed values, such as a set of event categories or ticket reasons. No depth, but no free-text drift.
NN/g gives a vehicle example: one hierarchy by how vehicles are shared, personal or cargo, versus a faceted version where each attribute has its own small tree.
Taxonomy examples
| Where | Levels | Example path |
|---|---|---|
| Biology | Domain, kingdom, phylum, class, order, family, genus, species | Animalia › Chordata › Mammalia › Carnivora › Ursidae |
| Website content | Topic › subtopic › theme | Analytics › Events › Naming |
| Traffic sources | Channel › source › campaign | Paid search › google › spring_sale |
| Analytics events | Object › action | cart_item_added |
| Support tickets | Area › issue › cause | Billing › Refund › Duplicate charge |
The common feature is that the same thing always gets the same path. That is what makes totals and filters trustworthy.
A worked example: one action, five spellings
Sample data: a store sent the "add to cart" action under several names, because different developers shipped it at different times. One week of counts:
| Event name as sent | Count |
|---|---|
| AddToCart | 400 |
| add_to_cart | 310 |
| cart-add | 120 |
| Added to cart | 90 |
| add2cart | 30 |
| Total | 950 |
Check: 400 + 310 + 120 + 90 + 30 = 950. An analyst who queries one spelling, the most common (AddToCart), sees 400 of 950 = 42% of the real adds. A funnel built on add_to_cart alone sees 310 of 950 = 33%. Every report built on a single spelling understates the step by more than half.
Now apply a small taxonomy: object, then action, in lower snake case, and decide the five names are one event: cart_item_added. Reports now count 950. The rules that made this possible are a few lines: a naming pattern, a list of allowed verbs, and one owner who approves new names. The decision rule: if two names would be answered by the same question, they should be one name.
How to design a taxonomy in five steps
- Start from questions, not from everything you could name. List the ten questions people ask and the categories needed to answer them.
- Choose the structure. One hierarchy if every item belongs in one place; faceted if items have several independent attributes.
- Write naming rules. Case, separators, order (object before action), singular or plural, maximum length.
- Cap the depth and the width. Keep the tree as shallow as your questions allow, and split any level whose list is too long to scan.
- Name an owner and a change process. A taxonomy without someone who says yes or no to new terms drifts back to chaos within months.
For a deeper treatment of naming for analytics, see the blog on data taxonomy.
A rule sheet can be four lines long. For example: (1) names are object_action, lower snake case; (2) completed actions use the past tense (cart_item_added), intents use the present (checkout_start); (3) only these verbs are allowed: added, removed, viewed, started, completed; (4) new names go through the owner, who also records the retired ones.
Common mistakes
- Copying an org chart. Categories should follow how people look for things, not how teams are arranged.
- Overlapping categories. If an item could sit in two sibling categories, people will file it both ways.
- A catch-all "Other" that grows. When "Other" is the biggest bucket, the list is missing a category.
- No retirement rule. Terms never removed, plus synonyms never merged, makes the list longer and worse. Use a tracking plan to record deprecated names.
- Changing names that data already uses. Renaming splits history in two. Add a new name and map the old one instead.
A taxonomy for events and sources in TapCub
Apply the add-to-cart example in TapCub.
- Set the naming rule. TapCub's event catalog accepts names that are snake_case, start with a letter, are at most 40 characters long and do not start with
$,_ortc_. That already forces lower case and separators; add your own object-then-action order on top. - Review what arrives. New events land in All-platform analytics › Data › Event inbox. You can accept one with a snake_case name or merge it into an existing event instead of naming every variant separately.
- Plan the list. All-platform analytics › Data › Tracking plan records events you intend to collect, with category, platforms, trigger, properties, owner and a status from Proposed to Live or Deprecated.
- Use categories. Each definition has a category (a new project lists Auth, Navigation, Commerce, Engagement, Form, Content and Other), which you can add to, rename and reorder with Manage categories.
- Keep sources tidy. Traffic channels are classified by rules you can configure under Web analytics › Sources; consistent UTM values (all lower case) keep one campaign from appearing under several spellings.
TapCub does not rename old data for you: events already sent keep the name they were sent with.
Sources
Was this explanation clear?
We use your answers to fix and expand this page.
Couldn’t send. Please try again in a minute.
Thanks for the feedback. We’ll use it to improve this page.You’ve already sent feedback on this page. Thanks.

Technical Engineer
Focuses on tracking code, SDKs, and API integrations, with practical guidance on implementation checks and troubleshooting.