Research-grade SEC data — without the $24K terminal.
Point-in-time, survivorship-free, and traceable to the filing — with a published, re-derivable accuracy baseline. The catastrophe this removes: the result that doesn't replicate because the vendor quietly revised the data under your paper.
- Point-in-time, as-first-reported data — kill look-ahead bias before it inflates your results.
- Full survivorship-free universe including delisted, bankrupt, and merged entities.
- Every fact traceable to its source filing for citation and peer review.
- A published accuracy baseline — 19,607 S&P 500 annual filings pass all 35 accounting-identity checks, 0 failures, CI-gated and re-derivable from one DuckDB script.
Built for
Researchers
- Point-in-time accurate
- Survivorship-bias-free
- Every number cited to its filing
Works where you do
The pain points we remove
Rigorous research needs clean, point-in-time, reproducible data — but the standard sources are expensive, gated, and quietly mutable. Valuein is built for the opposite.
The cost wall
Bloomberg is ~$24K+/user/yr; Compustat and CRSP come through WRDS, gated to whoever holds a university subscription. Independent researchers are locked out.
Look-ahead bias baked into vendor data
Look-ahead bias is present in common Compustat products — using the wrong vintage silently inflates results in studies of fundamentals and returns.
Survivorship bias
Testing on current constituents overstates returns because the underperformers dropped out. You need the delisted and bankrupt names present.
Reproducibility broken by silent revisions
When a vendor readjusts its time series after the fact, the dataset under your published paper changes — and replication breaks.
EDGAR is free but not usable
EDGAR is free but not trivial to scrape, and raw XBRL needs heavy processing. DIY normalization eats months you'd rather spend on the research.
The grind we take off your plate
From the daily check-ins to the month-end scramble — this is the recurring work Valuein automates so you spend your hours on the thesis, not the data.
Every day
- Write and run analysis code
- Clean and normalize raw data
- Debug coverage gaps and tag mismatches
Every week
- Construct datasets and factor panels
- Run regressions and backtests
- Validate against look-ahead and survivorship traps
Month-end & earnings
- Refresh panels with new filings
- Version data for reproducibility
- Document provenance for submission and peer review
What you can do with Valuein
Each job you need done, mapped to the exact capability that delivers it.
Affordable research-grade access
Free sample + S&P500 tiers, then Pro at $49/mo — no $24K terminal, no university WRDS gate.
Point-in-time, as-first-reported data
accepted_at on every fact and as_of PIT enforcement in the SDK kill look-ahead bias.
Full survivorship-free universe
The complete SEC population keyed on CIK — active plus inactive — back to 1993.
Reproducible, provenance-tracked results
verify_fact_lineage traces each number to its filing; versioned Parquet vintages plus deterministic, typed tools mean the same inputs give the same output — years later.
A data-quality claim you can check
The accuracy baseline (19,607 S&P 500 annual filings passing all 35 published accounting identities, 0 failures) is published and re-derivable from one DuckDB script — cite it, or reproduce it.
Pre-normalized XBRL
~11,966 raw tags mapped to 292 canonical concepts — comparable out of the box.
Works where you do
One Bearer token reaches the same point-in-time data from your AI agent, your notebook, or your browser. Use the surface that fits the job.
Research-grade SEC fundamentals without a $24,000 terminal or a WRDS login.
Point-in-time and survivorship-free by default — kill look-ahead bias before it inflates your Sharpe.
Reproducible by design: same inputs, same output, and every fact traces back to its filing.
Frequently asked
Can I cite Valuein data in a paper, and is it reproducible?
Yes. Every fact resolves to its source filing via verify_fact_lineage, and the Parquet schema is versioned so a given vintage is immutable — you can re-run the exact dataset that backed your results.
How accurate is the standardization — and can I verify the claim?
We publish a measured baseline: all 19,607 S&P 500 annual filings pass every one of 35 active published accounting identities, with 0 failures. It's CI-gated and re-derivable from one DuckDB script, so you can check it rather than take it on faith.
Do you offer academic or student access?
The sample and S&P500 tiers are free (the S&P500 tier is full history, 1993-present, for the index). Pro at $49/mo opens the full 19,000+ universe — a fraction of a WRDS seat. Reach out for classroom or research-group needs.
How do you handle look-ahead and survivorship bias?
Point-in-time acceptance timestamps prevent look-ahead, and the universe includes delisted/bankrupt/merged entities so it's survivorship-free — the two biases most likely to invalidate an empirical finance result.
What's the difference from raw SEC EDGAR?
EDGAR is free but raw — inconsistent XBRL tags, no standardization, painful to scrape at scale. We normalize ~11,966 raw tags into 292 canonical concepts and serve them point-in-time as columnar Parquet.
Research-grade SEC fundamentals without a $24,000 terminal or a WRDS login.
111M+ standardized SEC facts across 19,000+ companies, 1993–present. Free to start — no credit card.