Start with a real decision.
Which customer relationships cover their costs? Sales has orders, Operations has service usage, Finance has costs. Each team sees only part of the answer.
People & decisions: Agree which decision should improve, and who will use the result.
Data & systems: Identify the existing sources. Nothing is connected yet.
Give people room to act.
Named owners get time and authority to improve their data. The teams agree what “customer” and “cost” mean. Responsibility stays with the domains.
People & decisions: Assign owners, clarify decisions and make time for the work.
Data & systems: Record the responsible owner and shared definitions for each source.
Connect the sources.
A logical layer provides shared views of orders, usage and costs. Teams can combine them without first moving everything into one database.
People & decisions: Agree who may provide which data, for what purpose.
Data & systems: Connect the three systems and map their customer IDs.
Make the promise explicit.
Start with Customer value. Its data contract C01 defines structure, values, delivery and freshness. Quality gates check those promises. A record without a customer ID is stopped and its issue returned for correction.
People & decisions: Producers and users agree the contract and how changes are handled.
Data & systems: Record the contract as rules. Quality gates check records and control their release.
Detect violations. Fix the cause.
A record may contain invalid values, arrive too late, or belong to a stale data view. The relevant check shows the violation and prevents the affected use. Owners fix the cause, then the data is checked again.
People & decisions: Agree ownership, response and release decisions for each violation.
Data & systems: Check values, delivery delay and data age separately. Return issues and withhold affected data.
One product becomes part of a larger system.
The detailed contract C01 contracts into a small badge beside Customer value. Service cost follows with C02, combining usage and cost. Account overview builds on both products and receives C03. Each product has its own rules; the lines show its inputs.
People & decisions: Agree the meaning, support and intended uses of each product.
Data & systems: Connect each source to its product views. Compose products and define their individual schemas, quality rules and time limits.
Find a product, then arrange access.
The catalog makes products discoverable through metadata: What can a product be used for? Who maintains it? How current is it, and where does it come from? Read the description and select a product even without access to its data.
People & decisions: Write useful product descriptions and agree who may discover their metadata in the catalog.
Data & systems: Index descriptions, ownership, schemas, quality rules and lineage. Control metadata discovery separately from data access. Results show the owner, delivery time, freshness, frequency and quality metric from each Data Contract.
The same product, different access.
An analyst sees a masked customer name. The assigned account owner can see the name. An external user receives no data. The rules follow the product into its uses.
People & decisions: Agree purpose, permissions and who approves access.
Data & systems: Apply authorization and masking before returning data.
Build on what already works.
An app, a forecast and an AI assistant use the same products. More teams add use cases and reuse the definitions and checks. The products become a base for further work.
People & decisions: Help teams find existing products and contribute improvements.
Data & systems: Connect six uses to three products instead of rebuilding the data logic each time.
Let everyday use improve the system.
An application reports an error. Lineage helps the team find the cause and fix data capture. Several data products benefit, including products built on top of them. In later measurement windows, more incoming records pass their contract checks.
People & decisions: Review feedback with the responsible team, fix the cause and follow up on the improvement.
Data & systems: Propagate the correction through dependencies. Measure quality again: in this example, from 99.4 / 99.1 / 99.8% to 99.8 / 99.6 / 99.9%.