The GSDSI Pilot Process: Matched sample → license
A practical evaluation workflow for enterprise and federal buyers: test match rates and delivery specs first, then move to contract terms and production activation.
Since 2018
Privacy-safe enterprise data
VOSB & SDVOSB
U.S. Veteran-Owned, Service-Disabled
SCC Transfers
EU/UK/CH per privacy policy
GDPR · CCPA · CPRA
Consent-first data sourcing
Four steps buyers use to move fast
The fastest “time to value” comes from treating pilots as measurement, not demos. Start with the match-rate and spec checks that predict performance.
1) Scope the use case and success metric
Define the endpoint (activation, measurement, risk, research), the audience/geography, and what “success” means (match rate, lift, precision/recall, coverage).
2) Run a matched sample
Test on your own first-party seeds. Measure match counts and fill rates by field, confidence tier (where applicable), and geography before committing to a full license.
3) Confirm delivery + schema + refresh cadence
Pick SFTP/S3/Snowflake/API, confirm formats (Parquet/CSV/JSON), and align refresh cadence to your activation/measurement window so decay doesn’t dominate results.
4) Lock governance and go live
Document exclusions, permitted uses, retention/deletion SLAs, and downstream-sharing rules in the IO/MSA so production activation is audit-ready.
What to send for a matched sample
- A hashed seed file (10K-100K rows) for match testing (email/HEM, device IDs where applicable).
- Your target geography (country/state/DMA/polygon) and any sensitive-category exclusions.
- Delivery constraints (Snowflake share vs S3 vs SFTP) and preferred formats.
- The metric you’ll evaluate (match rate, lift, coverage, model performance).
Common pilot mistakes (and how to avoid them)
Testing the wrong seed
A pilot only predicts outcomes if the seed matches production reality. Use a representative slice of your CRM or exposure logs, not a “cleaned” demo file.
No success metric
Agree on what “good” means before you run anything: match rate by cohort, lift threshold, or precision/recall for a risk model. Otherwise results get debated, not decided.
Skipping governance
Contracts don’t fix unclear governance later. Lock exclusions, permitted uses, retention, and deletion propagation before scaling volume.
Pilot and license economics (indicative)
GSDSI does not publish a public rate card. Every quote reflects feed mix, geography, refresh cadence, delivery path, and governance scope. The floors below help procurement set expectations before a scoping call; final terms appear only in a signed order or insertion order.
- Matched samples are typically scoped at no charge or a nominal data-prep fee after the use-case call.
- Paid evaluation pilots (when required) are custom-scoped; multi-feed programs are quoted as a bundle, not a sum of list prices.
- Single-feed US production licenses often start in the mid five figures annually; multi-feed enterprise programs scale with volume, refresh SLA, and activation rights.
- Federal, clean-room, and international scopes are quoted separately because delivery and compliance envelopes differ materially.
See pricing for quote intake, the data glossary for shared terms, and evaluation checklists for diligence threads.
Typical success metrics by use case
- Identity / onboarding: match rate by cohort + refresh/decay sensitivity.
- CTV / measurement: dedup-ready exposure joins + stable householding assumptions.
- Location / foot traffic: DUA density by geo + visit-rule stability (dwell thresholds).
- Alt-data finance: ticker-mapping coverage + revision policy + backtest-ready history.