Quality Rules

Quality is not optional. Every data product declares quality checks in quality.yaml that run after every materialization. Checks with severity: error block downstream propagation — bad data does not reach consumers. This guide covers rule types, threshold configuration, SLA management, and the Quality Oracle system.

How Quality Enforcement Works

Materialization completes
    |
    v
Quality checks execute (platform quality checks)
    |
    +-- All pass --> Write to serving stores, notify downstream
    |
    +-- Warning fails --> Log warning, continue serving writes
    |
    +-- Error fails --> Block serving writes, enter retry/DLQ
                        Downstream assets are SKIPPED

The platform translates your quality.yaml declarations into platform quality check functions. No external quality frameworks are involved at runtime.

Note

When a blocking quality check fails, the platform’s asset graph structurally prevents downstream materialization. This is an architectural guarantee, not an application-level check.

---

Quality Rule Types

Akili supports three tiers of expressiveness, from declarative YAML to custom Python.

Tier	Defined In	Use Case
Declarative	YAML type + config	Standard checks: completeness, freshness, volume, uniqueness, range
Custom SQL	Inline sql block	Complex business logic, multi-table assertions
Custom Python	External .py file	Statistical analysis, ML drift detection

---

Continue Reading

• Built-in Checks — The 10 declarative quality check types
• Custom Rules — Writing custom SQL and Python quality checks
• SLA & Scoring — SLA configuration, quality scoring, and breach handling