Data Dictionary Overview (Fields, Joins, Delivery)
A buyer-facing overview of how GSDSI datasets are structured: common join keys, refresh cadence, and delivery formats. Request a field-level dictionary pack for evaluation.
Why procurement asks for dictionaries first
Enterprise buyers rarely license a feed on record counts alone. Legal wants permitted-use boundaries; data science wants join keys and decay fields; platform teams want partition conventions and manifest checksums. A dictionary pack answers those questions before a matched-sample pilot consumes engineering time.
What a data dictionary includes
- Field names, types, and nullability expectations.
- Entity keys and recommended join paths (household, device, location, firmographic).
- Refresh cadence and change control expectations.
- Delivery format conventions (Parquet partitioning, CSV delimiters, manifests).
What buyers should validate in a pilot
- Join yield on your own seed (match rate by cohort/geography).
- Accuracy floors and filtering rules for noisy signals.
- Sensitive-category exclusions and permitted-use boundaries.
- Schema stability and backward compatibility for production pipelines.
Join keys by product family
Dictionary packs are scoped to the feeds you plan to pilot. The table below is a starting point; final columns ship in the pack tied to your order.
| Family | Typical keys | Deep dive |
|---|---|---|
| Identity | MAID, HEM, household_id, ctv_id, link_type | Identity graph |
| Location | poi_id, geohash, device_hash, visit_date | POI data |
| Finance / alt data | ticker, brand_id, geography, event_date | Alternative data hub |
Downloadable field-level PDFs
Sales maintains downloadable dictionary PDFs per SKU. When those assets are cleared for public hosting, they will link from the sales toolkit. Until then, request a scoped pack through contact and cite this page in procurement memos.
Start with the pilot process for matched-sample evaluation, or browse the product catalog for delivery format details per feed.
Frequently asked questions about GSDSI data dictionaries
What is included in a GSDSI data dictionary pack?
A dictionary pack documents field names, types, nullability, entity keys, recommended join paths, refresh cadence, partition conventions, and delivery manifests for the feeds in scope. Enterprise buyers use it to wire warehouse schemas and QA scripts before production.
Which join keys are most common across GSDSI feeds?
Identity feeds emphasize MAID, SHA-256 HEM, household or device cluster IDs, and CTV identifiers. Location feeds pair POI IDs with mobility device hashes at privacy-safe cohort grain. Finance-oriented feeds add ticker, brand, and geography mapping columns. The pack calls out the keys required for your pilot.
How do I request field-level documentation for a pilot?
Use /contact or the Request dictionary pack CTA on this page. Specify the product families (identity, POI, mobility, CTV/ACR, clickstream, property, B2B) and delivery path (S3, SFTP, Snowflake). The solutions team typically returns a scoped pack within 2 to 3 business days after scoping.
How often do schemas change?
Core identity and POI schemas are versioned with backward-compatible change logs. Breaking changes require notice per contract. Pilots should include a schema-stability checklist and a sample change-delta file when available.
Where does provenance documentation live?
Pair the dictionary pack with /sourcing-methodology for source categories and exclusions, and /trust/sub-processors for hosted processing. Executable permitted-use language remains in the signed order or data processing addendum.