Skip to main content

Appendix: Build the DataHub Model in Your Own Trial

In the shared workshop account, DataHub is pre-built and read-only — so the room shares one trusted set of golden records and nobody can break it for everyone else. If you want the full hands-on experience of building the deterministic master-data layer yourself — model, sources, match rules, survivorship, deployment, and a live golden-record load — do it in your own free Boomi trial, where you have your own isolated account.

This appendix is optional and self-paced.

1. Get your own trial​

  1. Go to boomi.com and start a free trial, or use train.boomi.com.
  2. Sign in to your trial account at platform.boomi.com with username/password (not SSO).

2. Build the model with Companion​

Use the exact Companion prompts from Part 2 — they work the same in your trial:

  1. Create the model:

    Create a DataHub master data model called CustomerGolden360 for a golden customer record. Fields: legalName (required, string), dbaAliases (string), parentLegalName (string, the corporate parent's legal name for hierarchy), country (string), taxId (string), contractValueUsd (float/number), paymentTerms (string), renewalDate (date), products (string), primaryContact (string), plus a recordId natural-key string field (the model root reserves id), plus source-key fields acmeCustomerId (string) and betacorpAccountId (string). Use legalName as the record title.

  2. Add the three sources (ACME_SFDC, BETACORP, AGENT_CANDIDATE) and the three-tier match rules, exactly as in Part 2.

  3. Publish the model as version 1.

3. Deploy a repository​

  1. Create the repository and deploy the model:

    Create a new DataHub repository named Customer360MA on the USA East Hub Cloud and deploy the published CustomerGolden360 model to it.

  2. Confirm the deployment status is SUCCESS.

4. Mint the Hub Authentication Token​

This step exists only in your own account — it is the step you must never touch in the shared environment.

  1. Open your repository → Configure → Hub Authentication.
  2. Generate a Hub Authentication Token. You will use it (with your Hub username) as basic-auth credentials for the Repository API.
No API for this step

There is no platform API to mint the Hub token — it is generated in the DataHub UI only. This is exactly why, in the shared workshop account, the token is pre-provisioned and off-limits.

5. Load the Acme golden records and run a match preview​

  1. Load the six seeded Acme golden records — Northwind Trading Company, Globex Corporation, Initech LLC, Hooli, Inc., Stark Industries, and Starke Industrial GmbH — into the ACME_SFDC source. The sample-records bundle ships them ready-made in the DataHub batch upsert shape: acme-golden-records-batch.xml. POST it to your repository's Repository API records endpoint, using your Hub username and token as basic-auth credentials.

    Initial-load lifecycle

    A brand-new source starts in SOURCE_ATTACHED, and a plain update returns HTTP 400. Run the source through its lifecycle first: enable initial load → load the batch → finish initial load. Only then does the source accept ongoing updates. (This is the same lifecycle the BETACORP and AGENT_CANDIDATE sources need.)

  2. Publish the six candidates — Northwind, Globex, Initech, Wonka, Hooli, and Pied Piper — and watch DataHub match and survive them. Confirm the engineered outcomes from Part 6 reproduce in your repository:

    • Northwind, Globex, Initech → match and survive into their Acme golden records (Globex via the taxId rule).
    • Wonka → clean net-new insert.
    • Hooli stays the parent; Pied Piper is inserted as a child linked to Hooli.

  3. Optionally, publish the ambiguous "Starke Ind." record as an experiment. It matches neither Stark Industries (US) nor Starke Industrial GmbH (Germany) — the country gate blocks the German record and the 0.92 Jaro-Winkler threshold rejects the US one — so it lands in quarantine for stewardship: the DataHub-side equivalent of the Part 6 steward exception (no agent is involved here).

You built the deterministic core yourself

You now have your own CustomerGolden360 repository with live golden records and a tuned match engine — the same deterministic system of record the shared lab provides, but entirely yours to experiment with. Loosen the Jaro-Winkler threshold, add a source, or feed a new artifact and watch how matching changes — without affecting anyone else.