Data Coverage & Completeness

Updated: September 13, 2026 · Author: Jennie Chu · Reviewed by: Gemral Research Desk · Editorial Policy

Direct Answer: How large is the Gemral Edge dataset?

Direct Answer: Gemral Edge holds 1,305,395 public records across 20 datasets in two feeds: Government & disclosure records (875,154 records) and AI-infrastructure physical footprint (430,241 records). Every record is tied to the company and ticker that were valid on the day it became public.

What does the Gemral Edge data coverage page measure?

This page reports how many public records each Gemral Edge dataset holds, how far back each one goes and how the records spread across years. The counts are produced by a script that queries the production database and include only records that were already public on the update date. Defense contract actions, for example, become public about 90 days after signing, so actions signed in the last three months are not counted until their publication date passes. The figures below were measured on 2026-10-28.

Feed 2 — AI-infrastructure physical footprint: which datasets are included and how far back do they go?

AI-infrastructure physical footprint holds 430,241 public records across 8 datasets, refreshed daily to monthly by layer.

How are records dated and attributed to listed companies?

Each record keeps the date of the underlying activity and the date it became public. Research joins use the publication date, which removes look-ahead. Records are mapped to the listed parent that controlled the issuing or receiving entity on the publication date, so a contract signed before an acquisition stays with the original company, delisted and acquired companies remain in the history under the ticker they had at the time, and no ticker is attached when no public document establishes the link.

Where do the records come from?

All records come from public sources, described by category: U.S. federal procurement and contract-award records, defense contract announcements, federal arms-transfer notices, congressional and insider disclosure filings, lobbying and campaign-finance disclosures, company filings with the securities regulator, state and county environmental permit and land-use records, federal energy and trade statistics, open map data and public satellite imagery. The data contains no material non-public information. Field definitions, known gaps and point-in-time rules are documented in the Data Due Diligence Questionnaire, sent with every sample pack.

What do the coverage counts include and exclude?

A record is counted once it has a publication date on or before the measurement date and belongs to a dataset that is live in the data dictionary. Records whose publication date is still in the future are left out, which matters most for defense contract actions with their statutory delay. Derived chains that combine several records into one event are not counted separately, so the totals reflect underlying public records rather than analytics built on top of them. Identity tables such as companies, people and corporate relationships are not counted either: they support entity resolution but are not events. Counts are refreshed every day from the production database, and each year column shows how dense a dataset is in that year, which helps a research team judge whether a backtest window is long enough before requesting data.

How should a fund read the history depth?

History depth differs by dataset because each public source began publishing in machine-readable form at a different time and because older history is loaded in backfills. Backfilled rows carry a flag and a first-seen timestamp later than their publication date, so a fund can test them separately from records captured live. Sparse early years usually mean the source itself published little in machine-readable form, not that records were dropped. Where a gap is known, it is listed in the Data Due Diligence Questionnaire together with the reason and the plan to close it.

To see what individual rows look like before requesting data, open the data preview, which shows up to 20 real rows per table.

How is accuracy checked before a dataset ships?

Every record type is checked against the original public documents on a graded sample before it ships, and a record type that fails the accuracy gate is left out of the delivery and listed in the manifest. Funds can request a sample pack of Parquet files with a manifest of row counts and checksums, the data dictionary, a methodology note, the license and the DDQ by emailing support@gemral.com.

What related alternative intelligence streams are available?

Cross-reference correlated public records across our alternative intelligence directories, statutory trackers, and quantitative research libraries:

• Data Preview
• Verification Methodology
• Track Record
• Public Datasets & Machine Feeds
• API for Funds & Fintech
• Subscription Plans & Pricing