# Compare analytical workloads

Compare operational analytics using matched data, query semantics, hardware and timing conditions.

Source: https://particledb.ai/benchmarks/olap/

Use application-shaped datasets to complement broad analytical suites. Examples include rating summaries, event aggregation and grouped operational reports.

## Prepare a comparison

Use the same record count, data distribution and query semantics on every system. Record types, indexes and ingestion settings. Compare the returned values before comparing timings.

## Measure

Run systems sequentially on the same idle machine. Use the suite's two warmup iterations, then collect repeated timed runs. Report individual query results and the selected summary statistic.

## Interpret

A fast count over one dataset does not establish that every query or data model is faster. Report memory use, load cost, failures and query shapes where another system performs better.

## Publication

A comparison needs raw timings, the complete query set and exact versions. Until those inputs are available and reproduced, avoid retaining numerical speedup claims from an older environment.

[Shared methodology](/benchmarks/overview/) · [Query tuning](/guides/query-performance/)
