Duration: 5

This guide shows how to deliver a governed data product in a single sprint using Cortex Code (CoCo) and a machine-readable data contract as the single source of truth. One document captures the requirements; an agent reads it and generates, deploys, and monitors every artifact automatically.

The example uses a Retail Customer Churn Risk data product in a retail banking context, but the approach works for any data product in any industry.

In one sprint you will produce a complete, production-ready data product. Source tables are pre-loaded — everything below is generated and deployed by the Cortex Code agent:

Duration: 10

Clone https://github.com/srini86/data-products-lifecycle-fsi-example and cd into the repo root.

Open 00_setup/setup.sql in Snowsight and run each step in order:

Source tables loaded by Steps 3–4:

Confirm Step 6 shows all five tables before continuing.

Duration: 5

Open 01_discover/data_product_canvas.png. This one-page canvas captures the agreed requirements:

Deliver RETAIL_BANKING_DB.DATA_PRODUCTS.RETAIL_CUSTOMER_CHURN_RISK — governed, tested, and monitored — by end of sprint. The canvas is the only input needed to produce every artifact in the phases that follow.

Duration: 10

A data contract (ODCS v2.2 YAML) is the sprint's single source of truth — every artifact in the Deliver phase is generated from it.

Open 02_design/_example/churn_risk_data_contract.yaml. The contract has three key sections:

Schema — every output column with type, description, derivation logic, and a pii flag. Columns marked pii: true (email, phone, date of birth) automatically trigger masking policy generation in the Deliver phase.

Quality rules — null checks on required columns, a value range rule (churn score must be 0–100), and a freshness rule (max 24-hour age). These drive the DMF setup generated in Deliver.

SLAs and governance — daily refresh schedule, 99.5% availability target, data steward contact, retention period, and regulatory tags (GDPR, PSD2). These drive the monitoring alerts and access control setup.

Use Cortex Code to generate a contract from your own canvas:

"Generate an ODCS v2.2 data contract for <your data product>. Use 02_design/_example/churn_risk_data_contract.yaml as a reference."

Duration: 25

This phase uses the dplc-accelerator skill — an interactive orchestrator that detects your current phase, displays a progress tracker, and presents each prompt one step at a time.

From the repo root, run cortex to start a session, then invoke the lifecycle accelerator with $dplc-accelerator (or type start lifecycle). The skill detects your current phase and presents the steps one at a time. Keep this session open throughout the Deliver phase.

"Read the data contract at 02_design/_example/churn_risk_data_contract.yaml and generate a complete dbt project — model SQL, schema.yml, and tests."

Output: 03_deliver/dbt_project/models/ — model SQL joining all five source tables, schema.yml with not_null/unique/accepted_values tests, and singular tests for range and referential integrity rules.

"Generate masking policies and DMF setup SQL based on the governance rules in the data contract."

Output: masking_policies.sql — one conditional masking policy per pii: true column, granting unmasked access to privileged roles only. dmf_setup.sql — attaches Snowflake Data Metric Functions for each quality rule, scheduled to match the contract's refresh schedule.

"Generate monitoring and observability SQL — freshness SLAs, quality checks, usage tracking, and alerts."

Output: 04_operate/_example/monitoring_observability.sql — freshness check, DMF quality dashboard (PASS/FAIL per rule), usage tracking by role/user, and a Snowflake Alert that fires on SLA or quality breach.

"Deploy the dbt project to Snowflake using snow dbt deploy and run it."

The agent runs snow dbt deploy targeting RETAIL_BANKING_DB.DATA_PRODUCTS, then snow dbt run to materialise the output table. It then applies masking policies and DMF attachments from Prompts 1 and 2.

"Validate the deployment — run tests, check row counts, verify masking is applied."

The agent runs dbt test, checks row and null counts, then queries the table as the PUBLIC role to confirm PII columns are masked (email → *****@domain.com, phone → ***-***-XXXX).

Duration: 5

The monitoring SQL was generated in Prompt 3. Open 04_operate/_example/monitoring_observability.sql in Snowsight to review or re-run any check independently.

RACI Template: 04_operate/raci_template.md provides a reusable responsibility matrix — copy and fill in for your team.

Duration: 3

The repo includes a v2 contract (05_refine/_example/churn_risk_data_contract_v2.yaml) adding two new columns driven by consumer feedback, and evolution SQL in 05_refine/_example/evolution_example.sql. Update the contract and re-run the agent to regenerate only the affected artifacts — no full rebuild required.

Duration: 2

Run 06_cleanup/cleanup.sql in Snowsight. It drops RETAIL_BANKING_DB and DATA_PRODUCTS_WH, removing all objects created by this guide.

Duration: 2