Explain event schemas and why schema registries matter at scale

WHAT THIS TESTS: This question probes your understanding of data contracts and schema evolution in event-driven architectures. Interviewers want to see that you treat event schemas as production infrastructure, not afterthoughts, and that you understand how a registry prevents downstream data quality decay.

A GOOD ANSWER COVERS: First, define an event schema as a versioned contract that specifies field names, data types, nullability constraints, and semantic meaning for every event in a stream. Second, explain the registry purpose: it acts as a centralized catalog where producers register schemas and consumers retrieve them, enforcing compatibility rules such as backward, forward, or full compatibility before allowing new schema versions into production. Third, connect governance tools like Segment Protocols or Confluent Schema Registry to organizational scale by describing how they block breaking changes, automate documentation, and standardize naming conventions across teams. Fourth, enumerate specific scaling problems without governance: pipeline parsing failures due to type mismatches, exploding null rates in the warehouse when optional fields are added inconsistently, divergent analytics definitions when teams interpret the same event differently, and the inability to replay or reconstruct streams because historical events lack schema context.

COMMON WRONG ANSWERS: A weak answer describes schemas as nice-to-have documentation or confuses a schema registry with a simple message format like JSON. Another red flag is blaming data quality issues solely on bad producers without mentioning the lack of an enforceable contract. Saying that schemas are only for static databases and not for streaming events also signals a gap.

LIKELY FOLLOW-UPS: The interviewer may ask how you handle schema evolution when requirements change, such as renaming a field or changing a type from integer to float. They might also ask for a comparison between schema-on-write and schema-on-read, or how you would migrate an unregulated event stream into a governed one without breaking existing consumers.

ONE CONCRETE EXAMPLE: Imagine a checkout event emitted by a payments service. Without a registry, one team adds a discount_amount field as a string while another team expects a decimal; the analytics pipeline silently coerces the string to zero, causing revenue reports to understate discounts by 12 percent. A schema registry with compatibility checks would reject the string-typed producer change and force a typed decimal contract before deployment.