2. The ontology spectrum

In chapter 1 you built the definition: a formal, explicit specification of a shared conceptualization.

Now a practical problem. Somebody shows you a diagram and calls it an ontology. How do you tell whether it is one?

You could use the four words from chapter 1 and answer yes or no. That works, but a flat no sounds like an insult, and the other person stops listening. There is a better way.

One example, all six steps

Say you run a shop and you want to sort your products into groups. Here is the same job done six ways.

Step 1. Controlled vocabulary

In plain words: a fixed list of the words you are allowed to use.

Electronics, Audio, Headphones, Kitchen

That is all of it. No definitions. No structure. You just cannot type anything else.

What it buys you: everyone uses the same word. No more Electronics, electronic, and elec showing up as three different things in your reports.

Step 2. Glossary

In plain words: the same list, plus a sentence saying what each word means.

Audio: anything that produces or records sound.

What it buys you: everyone now means the same thing, not just types the same thing.

Notice the definition is written for a human to read. A machine cannot use it.

Step 3. Taxonomy

In plain words: the words arranged in a tree of parents and children.

Electronics
  └─ Audio
       └─ Headphones

What it buys you: one kind of link, "is a kind of". You can roll up. Ask for Electronics sales and headphones are counted automatically.

This is the step nearly everything sits on. When somebody says "we built an ontology", this is usually what they built.

Step 4. Thesaurus

In plain words: the tree, plus links for words that mean the same thing, or are just related.

It is the same word as the book you used at school to find another word for "big". It means the same thing here.

Headphones   same meaning as   Earphones
Headphones   related to        Music Player

What it buys you: words that mean the same thing, and words that are loosely related. Search gets much better, because somebody typing "earphones" now finds headphones.

Step 5. Conceptual model

In plain words: things with types and rules, instead of just words.

Product    has price (number), weight (kg)
Product    belongs to   Category
Supplier   supplies     Product

What it buys you: rules a machine can check. It can tell you that a product with a price of "blue" is wrong, or that a product has no supplier.

Step 6. Axiomatized ontology

In plain words: rules a machine can reason with.

An axiom is just a rule you state as true. So "axiomatized" means the rules are written down as logic, with a reasoning engine behind them.

Anything that needs a battery is an ElectronicItem.

You never tagged the torch as electronic. The reasoner works it out on its own and adds it.

What it buys you: the machine finds facts you never wrote down.

Place some real ones

Reading the ladder is easy. Using it takes practice.

Three things to remember

1. Each step adds one thing, not many

Step 3 adds the parent and child link. Step 4 adds other kinds of link. Step 5 adds types.

One new ability each time. That is why they are steps in order, and not six choices you pick from.

2. Higher is not better

Start with the question you cannot answer today. Then pick the lowest step that answers it.

A simple word list that every team really uses is worth more than a step 6 ontology that one person built and then left. Each step up costs more time to build, and more time to keep correct after that.

3. It stops an argument nobody can win

Somebody will say "that is not a real ontology." They are usually correct. It also helps nobody.

Say this instead:

That is a taxonomy. Step 3. It is doing its job.

Now you have said something true, you have not insulted anyone, and you can ask the question that actually matters: do we need to go higher, or not?

How to place anything in ten seconds

Do not ask what the model is called. Ask what it can do.

That last row is the real dividing line. If the system only ever gives back what you put in, you are on step 5 or below, whatever anybody calls it.

Where Fabric IQ sits

Start with what the ontology item clearly has. All of this is in the documentation, checked August 2026.

Entity types are real types, not words in a list. Properties carry a stated data type, so the system knows a temperature is a number. Relationship types have a direction and join two named entity types. Data quality rules can be set at the concept layer, on things like required values, allowed ranges, and uniqueness.

Run those against the ten-second table above. Things with types, links with types, and rules a machine can check. That is step 5 behaviour, and it is comfortably above where most business models ever get.

The one trap

Do not aim for step 6 by default. Ask which question you cannot answer today, then find the lowest step that answers it.

Most of the value in most organizations is getting from step 1 to step 3 and making it stick.

Next, chapter 3. You can now say what an ontology is and place any model on the ladder. Chapter 3 covers something stranger. Whatever step you are on, your model is quietly claiming that some things exist and others do not. Most modelers have never noticed they are making that claim.