Ask the Experts: 10 Common Mistakes Teams Make When Adopting Serverless Querying

We asked ten platform engineers and data leaders to share the most common mistakes they saw when teams move to serverless query models. Their answers reveal recurring themes around data layout, governance, and culture.

1. Skipping Data Layout Work

Many teams lift-and-shift JSON or CSV files into object storage and expect good query performance. The experts unanimously recommend converting to Parquet/ORC and partitioning early.

2. No Baseline Monitoring

Without accounting for bytes scanned and cost per query, teams are blind to expensive patterns. Instrument query logs from day one.

3. Allowing SELECT * in Production

SELECT * is convenient, but costly—especially when schemas evolve. Enforce explicit projections for heavy workloads.

4. Not Using Materialized Views for Dashboards

Dashboards are repeatable; materialize their computations and you’ll save money and latency.

5. Ignoring Small Files Problem

Small files increase metadata overhead and slow queries. Coalesce small objects during ETL.

6. Poor Partition Strategy

Partitioning on low-selectivity columns or creating too many tiny partitions hurts more than helps—plan partitions based on query patterns.

7. Overlooking Caching Opportunities

Result caching or query result sharing dramatically reduces repeat scans for analysts.

8. Lax Governance on Ad-hoc Environments

Sandboxes are necessary but should have cost guards and data redaction to avoid surprises.

9. Expecting Serverless to Fix Everything

Serverless removes the need to manage clusters, but it does not eliminate the need for good data engineering and query optimization.

10. Not Modeling Pricing Scenarios

Assume different pricing models and run cost simulations to inform commitments or reserved purchases.

Expert Recommendations

Across contributors, the highest-impact actions were:

• Convert to columnar formats and partition early.
• Implement query logging and cost estimation in the UI.
• Provide prebuilt datasets for common analysis to reduce exploratory scans.

Final Thought

Adopting serverless querying requires a mix of technical work and operational discipline. Avoid these common mistakes, and you’ll unlock powerful analytics benefits without the surprise bills.