Repair customer identity and star-schema filtering
A hands-on PL-300 lab. You produce the real artefact and 9 automated checks verify it behaves the way the exam expects.
Try this labAll PL-300 practice
- Certification
- PL-300
- Format
- Structured configuration
- Difficulty
- medium
- Estimated time
- 35 min
- Automated checks
- 9
The brief
Edit star.model.json. keys must define customers and sales model_customer_key using {kind:tuple,output:model_customer_key,columns:[unit,customer_no]} or {kind:copy,output:model_customer_key,column:customer_id}; other existing source columns can be used but must satisfy the result contract. Tuple means JSON-encode the ordered component array as one scalar key, so separators inside components remain unambiguous. relationships contain manyTable, manyColumn, oneTable, oneColumn, cardinality (many-to-one/many-to-many/one-to-one), direction (single/both), active Boolean. Require exactly active many-to-one sales.model_customer_key to customers.model_customer_key and sales.product_id to products.product_id, both single. measure contains table, column, aggregation (sum/count); the requested measure is sum of sales.amount at fact grain. Customer, product and simultaneous selections must retain the declared fact IDs and totals in all populations. Product selections must leave original customer choices available, including no-sales products. Unfiltered facts, including unmatched customer rows, remain in the measure. Read source tables and probes, run tests, inspect materialized keys and actual retained IDs, and repair without Reset.
What the checks verify
Your work is graded on 9 independent properties, not on matching one reference answer.
- Use supported typed derived-key, relationship and measure records.
- Every derived customer key identifies exactly one source dimension row in every population.
- Use exactly the two supplied active many-to-one relationships with single dimension-to-fact filtering.
- Customer and region selections retain precisely their own fact identities, including no-match selections.
- Tea and no-sales product selections retain only their actual facts.
- A simultaneous region and Tea selection retains facts satisfying both independent dimensions.
- Sum each retained sales amount once at fact grain under every query and population.
- The unfiltered measure retains every original sale identity, including unmatched customers.
- Product selections leave the unselected customer dimension's original choices available, even when no facts match.
Where this sits in the PL-300 blueprint
- Domain
- Model the Data
- Objective
- Design and Implement a Data Model
- Skill
- Cardinality and Filter Direction
Part of PL-300 preparation
Labs are written by ExamNova to teach the decisions the exam tests. They are not reproductions of vendor lab content.