Marketing Taxonomy: How to Build One That Survives

A marketing taxonomy is the agreed structure for classifying and naming marketing data — the dimensions you record about every campaign, asset and placement, and the permitted values for each.

Most organizations have built one. Far fewer are still using the one they built. A taxonomy dies the same way each time: it is agreed in a workshop, published as a document, and then applied by people who were not in the workshop, in systems that do not check.

What is a marketing taxonomy?

Both words in “agreed structure” are load-bearing, and taxonomies fail on the first one. The structure is the set of dimensions you record: channel, market, brand, objective, audience. The agreement is that everyone records the same ones, with values drawn from the same lists.

A taxonomy without the structure is a naming convention with no rationale. A taxonomy without the agreement is a personal preference that happens to be written down. The business case for having one is well made elsewhere (Rapp, “Transforming Your Business with Marketing Taxonomy”, accessed 2026-09-11); this page is mostly about the part that comes after everyone agrees.

What a taxonomy classifies

Channel, campaign, audience, creative, market, objective — and the allowed values for each.

The “value type” column is the one that determines whether the taxonomy survives. Closed lists can be enforced; patterns can be validated; free text cannot be either. Every dimension you leave as free text is a dimension that will be inconsistent within a quarter.

The “owner” column matters for a different reason: these owners are in different teams, and several of them do not report to whoever is running the taxonomy project. Nielsen’s treatment of why an organization needs a data taxonomy makes the cross-functional case plainly (Nielsen, “Why You Need a Data Taxonomy”, accessed 2026-09-11).

The four types of taxonomy

Flat, hierarchical, network and faceted.

• Flat. One level, no parents. A simple list of channels. Easy to apply, impossible to roll up.
• Hierarchical. Parent-child levels: channel → sub-channel → placement type. Supports roll-up reporting and is the most common structure in marketing.
• Network. Items relate to multiple parents in no fixed order. Powerful, and rarely worth the cost outside a content or product graph.
• Faceted. Independent dimensions applied in combination, so a campaign is simultaneously a market, a brand, an objective and a channel, in any order.

A note on that list: the four-way split is an information-architecture convention rather than a formal standard, and different practitioners draw it slightly differently. The distinction that carries real weight — and the one that decides how a marketing taxonomy behaves — is hierarchical versus faceted. Nielsen Norman Group puts it precisely: a hierarchical taxonomy “allows for only one ‘lens’ or organizing principle,” while a faceted one “has a separate small hierarchy for each facet or attribute” (Nielsen Norman Group, “Taxonomy 101: Definition, Best Practices, and How It Complements Other IA Work”, Page Laubheimer, 3 July 2022).

That second definition is exactly what a working marketing taxonomy is. Most are faceted with hierarchy inside individual facets. Channel rolls up; market rolls up; but market is not a parent of channel, and forcing either into a single tree is what produces the six-level naming conventions nobody can complete from memory.

Structure: flat, hierarchical, faceted

Depth buys precision and costs adoption.

Every additional level makes the data more precise and makes the person filling the field more likely to guess. That trade-off is the single most consequential design decision in the exercise, and it is usually made by whoever most enjoys designing taxonomies — which biases it toward depth.

A workable heuristic: a dimension earns another level when a real reporting question requires it, and not before. “We might want to slice by that someday” is not a reporting question. Levels are cheap to add later and expensive to remove, because removing one invalidates every value already recorded against it.

The same trade-off governs breadth. Each additional dimension is another field somebody completes on every record, and completion rates fall as the form gets longer — which means an ambitious taxonomy can produce less usable data than a modest one, because the fields that matter get rushed alongside the fields that do not.

There is a practical ceiling worth respecting: if a person cannot hold the required dimensions in their head while setting up a campaign, they will consult the document once, approximate thereafter, and the approximations will not be random. They will cluster on whichever values appear first in the list.

Taxonomy, metadata and standards

The taxonomy is the structure; metadata is what it produces; the standard is what enforces it.

The three words get used interchangeably and name three different things.

• The taxonomy decides that channel is a dimension worth recording, and that it has eight permitted values.
• The metadata is the actual value on an actual record: this placement’s channel is paid_social.
• The data standard is what makes the record refuse anything that is not one of the eight.

Teams that have the first and the second but not the third have a taxonomy that describes their intentions and metadata that records their practice, with no mechanism holding the two together. The data dictionary is where the definitions and owners are written down.

The standards layer: Read: what are data standards? — the mechanism that makes a taxonomy hold.

How to build one

Start from the questions the data must answer, then derive the dimensions.

1. Write down the reporting questions. “What did we spend by channel and market last quarter?” Every dimension in the taxonomy should be traceable to a question someone actually asks.
2. Derive the dimensions from those questions, not from the systems. Platform field lists are an inventory of what is available, not a decision about what matters.
3. Decide the value type for each dimension. Closed list, pattern, or identifier. Resist free text; each instance is a future reconciliation project.
4. Name one owner per dimension. Someone who can approve a ninth channel value without convening a committee.
5. Agree the values with the people who will enter them. Including agencies. A list agreed without them is a list they will approximate.
6. Write the naming convention last. The convention assembles taxonomy values into a string; it cannot be designed before the values exist. Naming conventions are their own discipline.
7. Put it where the work happens. Into the campaign builder, the intake form, the setup workflow — not into a document that people are asked to consult.

Step 1 is the one most often skipped, and skipping it is why so many taxonomies have dimensions nobody reports on. Every unused dimension is a field someone fills in for no reason, which is how a taxonomy loses credibility with the people it depends on.

Naming conventions in detail: Read: naming conventions — how taxonomy values become a campaign name.

Who owns the taxonomy

A named owner per dimension, usually in marketing ops or ad ops.

Ownership is the part most taxonomy projects under-specify. There are really three roles.

The taxonomy owner is accountable for the structure as a whole: which dimensions exist, when one is added or retired, how change is communicated. This normally sits in marketing operations.

Dimension owners approve new permitted values within their dimension. Media ops owns channel; the brand team owns brand; procurement often owns agency. Their authority is narrow and real.

The enforcement owner is whoever controls the systems where values are entered. This is frequently a different person from either of the above, and their absence from the project is the most common structural cause of failure. A taxonomy agreed without them is a taxonomy that cannot be applied.

Stakeholder alignment on naming conventions and taxonomy governance is the second most common problem raised with us, across 80 enterprise accounts — and it is consistently described as harder than deciding the values themselves.

Why taxonomies get abandoned

Agreed in a workshop, applied by people who were not there, in systems that do not enforce it.

The abandonment pattern is consistent enough to describe as a sequence.

Month one. The taxonomy is agreed and documented. Compliance is high, because the people who designed it are the people using it.

Month three. New team members and a new agency start work. They receive the document. They apply it approximately, because a document cannot tell them that Paid Social should have been paid_social.

Month six. Reporting shows variants. Someone builds a mapping table to reconcile them, which works and becomes permanent maintenance.

Month twelve. The mapping table is the real taxonomy. The document describes something that no longer matches the data, and the next project proposes building a new taxonomy.

Cross-agency governance and taxonomy compliance failures come up across 74 enterprise accounts, which is the same pattern viewed from outside: the people least able to see the taxonomy are the ones generating a large share of the records.

Colgate-Palmolive’s ad operations team described the alternative — a taxonomy that reaches agency teams through the systems they work in rather than through a document.

“With these integrations in place, our agency teams can confidently set up campaigns, knowing that every element aligns seamlessly with Colgate’s taxonomy requirements.” — Eric Kirtcheff, Global Head of Ad Operations, Measurement, and Data Integrity, Colgate-Palmolive

The difference is not that the agency teams became more careful. The setup step stopped accepting values that broke the taxonomy.

Taxonomy inside governance

The governance framework says a taxonomy must exist; the taxonomy says what the values are.

Data governance is the wider program: ownership, policy, quality expectations and the mechanism that enforces them. The taxonomy is one of the artifacts it governs, and usually the one with the most day-to-day contact with working teams.

The relationship runs both ways in practice. Governance without a taxonomy has nothing specific to enforce. A taxonomy without governance has no route to resolve the argument about whether paid_social and social_paid should both exist, which is the argument that eventually decides whether anyone keeps using it.

Vanguard’s marketing technology team named the step that precedes all of it.

“We had to come to an agreement on what the taxonomy would be for how we measure campaigns across different channels.” — Kimberly Whitehead, marketing technology manager, Vanguard

Agreement is the prerequisite and the hardest part. It is also the part a tool cannot do for you.

What a tool can do is make the agreement stick after the meeting ends, which is a narrower claim than most vendors in this space make and the only one worth relying on. The workshop produces the taxonomy. The systems decide whether it survives contact with the next quarter.

Frequently asked questions

Sources

Outbound citations, named and dated:

• Rapp, “Transforming Your Business with Marketing Taxonomy” (accessed 2026-09-11) — the strategic business case for taxonomy.
• Nielsen, “Why You Need a Data Taxonomy” (accessed 2026-09-11) — the cross-functional case for a shared classification structure.
• Nielsen Norman Group, “Taxonomy 101: Definition, Best Practices, and How It Complements Other IA Work” (Page Laubheimer, 3 July 2022) — the hierarchical-versus-faceted distinction, quoted directly.