How funds evaluate alternative data vendors

The checklist data sourcing teams run before licensing a dataset: history, coverage, point-in-time, delivery and compliance.

History and survivorship

Backtests need depth — typically 3+ years — and they need honest history. A dataset that only keeps live records overstates past performance (survivorship bias). Ask when coverage started, whether deleted entities are retained, and whether you can reconstruct the feed as it looked on any past date.

Point-in-time correctness

Point-in-time means every record carries the timestamps needed to know what was observable when: when a job posting opened and closed, when a transaction settled, when a filing became public. Without it, backtests silently leak the future into the past.

Coverage, delivery and compliance

Check geography, entity resolution (can the data be mapped to tickers or companies?) and refresh cadence. Delivery should fit your stack: API for exploration, S3 or Snowflake shares for production, Parquet for research. Finally, ask for the data provenance story — consent, PII handling, licensing terms — because compliance will gate every serious buyer.