Fourteen teams produce architecture diagrams nobody can compare. The same rectangle means a single service in one diagram and an entire platform in another; arrows sometimes mean data flow and sometimes deployment. Leadership cannot answer which applications support the claims capability or which run on the estate leaving support next year. The CIO asks which EA framework to adopt. Which move addresses the actual problem?
Show the full answer Hide the answer
The deciding property
The complaint is about meaning, not about process. Nothing in the stem says decisions are made badly or in the wrong order; it says two diagrams cannot be put side by side, and two cross-layer questions cannot be answered. That is a notation problem: the estate has no agreed vocabulary for what a box is, what an arrow asserts, and how an application relates to a capability or a node.
The three families of EA framework answer different questions, and conflating them is the most common wasted year in this discipline:
| Family | Example | Answers |
|---|---|---|
| Method | TOGAF's ADM | in what order do we do the work |
| Taxonomy | Zachman | what kinds of artefact could exist and for whom |
| Notation | ArchiMate | what does this symbol mean |
ArchiMate is the one with defined element types across business, application and technology layers and defined relationships between them, which is exactly what makes two diagrams comparable and what lets a question cross layers. Adopt the layer of the problem and no more.
What it costs
Notation is not free: expect a few weeks of friction, a one-page subset (most estates use 15 to 20 element types out of the full language), and a rule that anything drawn for a non-technical audience may stay informal. A notation mandated in full is abandoned in full.
Why the other options fail
TOGAF and the ADM would be right if the symptom were decisions made in the wrong order or target states with no migration plan. Adopting a method to fix a vocabulary problem adds phases and deliverables while the fourteen teams keep drawing incompatible boxes inside them.
Zachman is a classification of what artefacts could exist, cut by six interrogatives and six perspectives. It is a genuinely useful completeness check — it tells you that nobody has written down the motivation view — but it defines no symbols, so two teams using it still draw whatever they like.
An EA repository tool is where teams go to fight about semantics, not where semantics come from. Without an agreed meaning for a box, the tool fills with 14 teams' private models and a licence renewal. Buy it after the notation is settled and there are questions worth querying.
Mandating C4 is the strongest distractor and is right in a different estate. C4's levels are precise about software structure — context, container, component, code — and its explicit level numbering fixes the "box means a service or a platform" confusion better than anything. What it does not model is capabilities, business processes or physical nodes, so it cannot answer either of the two questions leadership asked. The flip is clean: if the diagrams in dispute are all inside software, choose C4 and skip ArchiMate entirely.
When this is the wrong answer
Below about six teams, a one-page drawing convention on the wiki — what a box is, what an arrow means, which three levels exist — gets 80% of the benefit with none of the training. The threshold is whether diagrams are read by people who did not attend the meeting where they were drawn. If everyone was in the room, the convention can stay informal.