From Statistical Evidence to Executable Data Graphs

Most enterprises don’t lack data.

They lack verified structure.

We’ve all seen relationship diagrams in slide decks. They look clean. They make sense. But they are descriptive — not executable.

In practice, data relationships drift:

· Foreign keys are incomplete

· Naming conventions change

· Cross-system links go undocumented

So the real question becomes:

How do you move from “assumed relationships” to verified, machine-readable structure?

At Arisyn, we approach this from the data itself.

Instead of relying on metadata, we analyze value behavior:

· null_row_num to understand field completeness

· distinct_num to evaluate domain uniqueness

· co_occure and inclusion_ratio to detect structural inclusion

If 90%+ of distinct values in one column appear in another, we don’t treat that as coincidence. We treat it as a structural inclusion signal

From there, relationships are not drawn as diagrams.

They are returned as structured JSON:

· source_table

· source_column

· target_table

· target_column

Each edge is statistically validated.

That JSON graph is executable.

It can generate JOIN paths.
It can support multi-hop traversal.
It becomes infrastructure.

Diagrams explain relationships.

Executable graphs enforce them.

And once relationships are machine-readable, AI stops guessing — and starts operating within constraints.

That’s the shift.