One seeded graph, projected into every vendor's schema.
The account behind a Salesforce opportunity is the account a Zendesk ticket and a
Gong call reference, because each system is a projection of one company graph rather than a table
of its own.
The graph is generated from a seed, a clock and a day-state - 730 days of trading, hiring and
filing - and the same three inputs rebuild it down to the ids.
What the generator does that a row filler does not.
Projection, not a generator per system.
The entities - people, customers, deals, tickets, calls, files, chat - reference each other in the graph. A connector adds no data of its own: it maps those entities onto its vendor's resource shapes, pagination and error codes, which is why the ids still line up once the systems are apart.
The dashed ring is the part a per-system generator cannot have: neighbours holding each other's ids, so a cross-system join resolves.
The requester on a Zendesk ticket is a contact at the account the Salesforce opportunity names, and the agent who answered is the employee HiBob has on that team. The ids match because there is one company underneath, not three.
Two years of it, at the scale you ask for.
History runs 730 days back, because a year-over-year question and an annual renewal both need last year to exist. Scale is a choice at build time - the vocabulary this console builds in - and the rows below are measured from the systems rather than modelled on this page.
| Tier | The company | Rows, four measured systems |
|---|---|---|
| Measuring… | ||
The same company, a quarter later.
Generating again with another seed gives a different company, which is no use to anything that syncs. Era holds three states of the same one, and the quarter is derived once for the graph so every system moves by the same thirteen weeks.
Scored by a scorer, checked by the vendors' own clients.
Every tenant is scored against reference distributions and swept by an adversarial detector looking for the tells of generated data, and the baseline a CI gate holds the generator to is published from the scorer's own output.
The shape of an API is checked the same way, with code nobody here wrote.
zenpy's Zendesk suite runs with its cassettes flipped to record mode, putting requests it
once recorded against a real account back on the wire. psycopg 3's suite is written against a server
rather than the driver, so it can be pointed at the Postgres simulator. Then stripe-python, the Jira
client, elasticsearch-py, minio-py, pymysql, clickhouse-connect and the vitest suites from
vercel-labs/emulate. Every upstream is pinned, and every connector carries an audit file
saying what was probed, what was broken and which test pins the fix.
/_meta