Standards
8 min

Taxonomy Management Software: What to Look For and Who Builds It

Key takeaways

  • Six product categories share this vocabulary. A feature grid laid across them compares things built for different jobs.
  • Governance depth and export fidelity separate products that look identical in a demo.
  • Ask what a term meant on a past date. Versioning, audit logs and approvals often exist without anything joining them.
  • Test with your own vocabulary in a trial. Demos run on data chosen to work.
In this article
Share

What counts as taxonomy management software?

Narrower than the shortlist you will be handed. Taxonomy management software authors and governs a controlled vocabulary: it holds the terms, their labels, their hierarchy and their mappings, and it records who agreed each one and when.

That is a specific job, and most products that appear in a search for it are built for an adjacent one. A graph database stores the vocabulary and does not govern it. A data catalogue describes assets that already exist and does not author the terms. Both are useful. Neither is this.

So the first move in any evaluation is to establish which category each shortlisted product is actually in. A feature grid laid across categories compares things that were never trying to do the same thing.

Which categories get confused with it?

Six categories share the vocabulary of this market. The table below is from our own category research dated 2026-10-27, and taxonomy and ontology platforms compared works through it in detail.

CategoryWhat it is built to doRepresentative products
Taxonomy and ontology managementAuthor and govern agreed meaning, with versioning and a review trailTQ Data Foundation, PoolParty, Synaptica
RDF database / triplestoreStore triples, answer SPARQL, run reasoningGraphDB (Graphwise), Virtuoso, Amazon Neptune
Property-graph databaseGraph storage on a non-RDF modelNeo4j, TigerGraph
Knowledge graph platformVirtualise data where it lives and add semantic contextStardog, Anzo
Data catalogueDiscovery, lineage and policy over existing assetsCollibra, Alation, Atlan
Metrics semantic layerConsistent metric definitions for BIdbt, Cube, AtScale

One note on naming, because it trips up shortlists: Ontotext's GraphDB product pages now redirect to Graphwise, so a comparison written against "Ontotext" is naming a brand in transition. Write GraphDB.

What should you score a product on?

Eight things, and only the first three are where products in this category genuinely differ.

The scoring criteria

  • Governance depth. Can it version a definition, hold an approval against it, and tell you what the term meant on a past date? Version tracking that says something moved is not the same as a review trail.
  • Export fidelity. Does the vocabulary leave with its relationships, its alternative labels and its history intact? Test the round trip, do not read the brochure.
  • Query-time resolution. Can a system resolve a term over an API fast enough to be worth calling, without a person in the loop?
  • Polyhierarchy. Can a concept take more than one parent in the editor, and does export preserve both?
  • Standards support. SKOS at minimum, and whether the implementation is complete or partial.
  • Validation. Can it enforce that concepts have definitions and owners, and detect cycles?
  • Mapping to external schemes. Maintained as data, and not as a spreadsheet beside the product.
  • Workflow latency. How long a term takes to get approved in practice. Slow approval is what makes people route around a vocabulary.

Governance depth and export fidelity are the two that separate products that look identical in a demo. Everything else is table stakes or is easy to verify.

How do you test those claims?

With your own vocabulary, in a trial, in an afternoon. Demos are built on data chosen to work.

Four tests worth running

  • The round trip. Import a real vocabulary, export it, diff the two. Count what was lost.
  • The second parent. Create a concept under two parents. Export. Check both survived.
  • The historical question. Change a definition, then ask the product what the term meant before the change. If you cannot get an answer, there is no review trail.
  • The API call. Resolve a term from outside the product, timed. If it needs a person or a nightly export, it will not serve a retrieval system.

The historical question is the one that most often fails and the one buyers most often forget to ask. A product can have versioning, an audit log and an approvals feature and still fail this. The three are stored separately, and nothing joins them.

Which category do you actually need?

Four questions, in this order.

If the problem is...You need
Disagreement about what terms meanTaxonomy and ontology management
Finding and tracking existing data assetsData catalogue
Querying connected data at scaleA graph database, RDF or property depending on portability needs
Consistent metrics in BIMetrics semantic layer

Most organizations eventually need more than one, and the mistake is buying a product from one category expecting it to cover another. A catalogue with a business glossary is genuinely useful and does not govern a vocabulary. A graph database with an ontology editor bolted on stores the model and does not run the approval.

What does the wrong choice cost?

Three costs, and the first two are usually booked somewhere other than the taxonomy budget.

Where the money goes

  • Re-authoring on replatform. A vocabulary that cannot export with its history has to be rebuilt by hand, and the decisions behind it are lost rather than migrated.
  • Shadow vocabularies. When approval is slow, teams keep local lists. You end up paying for the governed one and using the ungoverned ones.
  • Retrieval that quietly underperforms. A system filtered on inconsistent terms returns partial sets and reports nothing. The symptom looks like a search problem and gets a search budget.

The third is the expensive one now that retrieval systems read the vocabulary directly. A buyer evaluating on editor features and ignoring query-time resolution is optimising for the people who maintain the taxonomy over the systems that consume it, and the consumers now outnumber the maintainers by orders of magnitude.

What this comparison deliberately does not do

It does not score the named products against each other.

Our standard is that every claim about a named vendor is sourced to that vendor's own published documentation, dated and quoted as what they publish. The category placements above meet that standard and come from research dated 2026-10-27. A per-product scoring table against the eight criteria does not exist yet, because it requires re-reading each vendor's current documentation and that has not been done.

A scored grid assembled from memory or from last year's reading would look more useful and be worth less. We will publish one when it is sourced.

So what should I take away?

Decide the category before you compare products, score on governance depth and export fidelity, and test both with your own vocabulary instead of accepting a demo.

TopQuadrant builds in the taxonomy and ontology management category, which means we are one of the options and not the whole list. The honest version of our pitch is narrow. We built for the agreement and the record of who made it, on open standards. The vocabulary and its history can leave if you decide to go elsewhere. Test that claim the same way you test everyone else's. Run the round trip. Taxonomy management covers the discipline the tooling is meant to serve.

Give your AI the context it's been missing

See how the TQ Data Foundation turns your enterprise knowledge into trusted, Al-ready context.