Early access opening soon

Quantitative signals you can trace to the source

Stratum turns macro series, market prices, regulatory filings and financial news into scored gold signals: Apache Iceberg tables, an open format you query with plain SQL and safe to re-run without duplicates, whose every build is recorded in a signed, hash-chained lineage that is verified daily in public.

date
2026-10-06
dgs10_dgs2_spread
0.52
is_inverted
false
fed_cycle
"cutting"
t10y2y_ma60
0.47
source_id
"fred"
run_id
"2026-10-06T06:00Z"
schema_hash
"9f3c…a71e"
record_signature
"kms:3045…02b1"
Illustrative values. Signal columns from the yield_curve schema; lineage record simplified.

Built from public data

  • FRED
  • Market prices
  • U.S. Treasury
  • SEC EDGAR 13F
  • CFTC
  • FINRA
  • EIA
  • FEMA
  • News & research

Medallion pipeline

Raw responses in bronze, typed tables in silver, scored signals in gold. Each layer is reproducible and idempotent.

Signals, not just data

Gold tables join sources into domain signals, aligned to release dates and validated walk-forward.

Refreshed on release

When a source publishes, its silver and every gold signal that reads it rebuild automatically.

Bronze → Silver → Gold

Stratum refines raw data through three medallion layers, each one reproducible, idempotent and traceable to its source. How it works →

  1. Bronze Raw responses, exactly as fetched — the audit trail
  2. Silver Cleaned, typed, deduplicated tables, queryable with SQL
  3. Gold Scored cross-source signals as Apache Iceberg tables
Every record stays traceable to the exact source fetch that produced it.

Gold signals, ready to query

Each signal is an Apache Iceberg table: open format, plain SQL, and safe to read while it refreshes. Tables update automatically when their sources publish new data, with no duplicate rows after a re-run.

Read the full signal reference →


Trust, not just data

Plenty of platforms store data. Stratum makes every number explainable: where it came from, which run produced it, and what that run read and wrote. Check today's signed anchor →

  • Tamper-evident lineage

    Every ETL run appends a KMS-signed record to a hash chain: what it read, what it wrote, record counts in and out, the schema hash. Each record links to the one before, so editing or deleting history breaks the chain. A signal isn’t just a number; it’s a number you can explain.

  • Apache Iceberg tables

    Gold signals are stored as Iceberg, an open table format on Parquet, so you query them with plain SQL (Athena, Spark) and are not tied to a proprietary store. Writes are ACID upserts: a reader never sees a half-written table, and re-running a stage or backfilling a source corrects values instead of creating duplicates. Every change is a snapshot, so recent versions of a table can be queried as they were.

  • Verified daily, in public

    A verifier checks every chain each day and publishes the signed chain heads on our transparency page, where your browser can check them. Lineage is also archived to write-once storage, and any record can be exported as an audit bundle that verifies offline.

  • Built for review

    Customer-managed encryption at rest, TLS everywhere, isolated AWS accounts, and infrastructure deployed only through a tested pipeline. Raw source responses are kept as fetched. On the roadmap: tier-based access controls and trusted third-party timestamps.

Get early access

The live signal dashboard is coming. Join the waitlist for early access — plus occasional notes on new gold signals and methodology.

We'll send one email to confirm your address. Your email is stored only for the Stratum waitlist, never shared, and every email we send has an unsubscribe link. Privacy