SEC EDGAR API for Point-in-Time Fundamentals: What Free Actually Costs

If you searched for "sec edgar api fundamentals point in time free," the honest answer is: yes, it exists, it's free, and it's the same source Tradevo Data is built on. SEC EDGAR's companyfacts API gives you every XBRL fact every US public company has ever filed, including the original tag, the filing date, and every amendment. There is no paywall. There is no reason to trust a third party's numbers over the primary source, because the third party's numbers are the primary source, reshaped.

So this page isn't a takedown of EDGAR. It's a parts list of what you're actually building if you go direct, so you can decide with real information whether the weeks are worth it to you.

What you get by parsing companyfacts yourself, for zero dollars

Hit https://data.sec.gov/api/xbrl/companyfacts/CIK##########.json for any filer and you get, per concept (Revenue, NetIncome, Assets, etc.), an array of every reported value with its start, end, filed, form, fy, fp, and frame metadata. That's genuinely everything: original filings, 10-K/A amendments, every restatement, every dimensional breakout the company chose to tag. It's public domain. No rate-limit fee, no seat license, no vendor lock-in. If you only need a handful of tickers and you're comfortable writing a parser, you can be pulling real numbers in an afternoon.

What it actually costs you

The data is free. Reconstructing point-in-time correctness from it is the part that eats weeks, not hours. Specifically:

XBRL tag drift. Companies change which us-gaap tag they use for the same economic concept across years (e.g., NetIncomeLoss vs ProfitLoss, Revenues vs RevenueFromContractWithCustomerExcludingAssessedTax). Building a stable multi-year Revenue series means mapping and reconciling tag aliases per company, per era — and re-checking that mapping every filing season, because companies keep changing tags.

First-filed vs. restated reconciliation. EDGAR gives you every value ever reported for a period, but it doesn't hand you a clean answer to "what did the market know on this date." You have to sort by filed date yourself, determine which value was public as-of any given as_of date, and separately track whether a later filing revised that same tag/period by a material amount. Get this wrong and you've silently reintroduced the lookahead bias you were trying to eliminate.

Q4 derivation. Many companies don't file a separate Q4 10-Q (there isn't one) — Q4 has to be derived as FY minus Q1+Q2+Q3, and only when all four period-tags are consistent and present. Doing this wrong produces phantom quarters or silently wrong numbers. (Even done carefully, some things — like Q4 EPS or diluted shares — aren't reliably derivable this way, which is why we don't offer them.)

Dimensional members. A lot of fundamentals are tagged with xbrli:explicitMember segments (by business unit, geography, currency). Naively summing a concept across a filing without filtering members produces double-counted or garbage totals.

Amendments and qa checks. A 10-K/A three years later can quietly revise a number you've already baked into a backtest. Nothing in the raw feed flags this for you — you have to build your own restatement detector (a >0.5% same-tag delta is a defensible threshold, but you have to choose one and apply it consistently across every company you track).

Rate limits and scale. EDGAR's fair-access policy is workable for a few hundred companies. At scale — thousands of companies, up to 12 years of annual history, plus quarterly detail — you're managing retries, backoff, and a local cache — plus a refresh job every filing season, indefinitely, because EDGAR doesn't stop generating new 10-Ks the day you finish your script.

None of this is exotic engineering. It's just time. Multiple person-weeks, typically, to get from "I can fetch companyfacts" to "I trust this as-of any date without silent lookahead bias."

Side-by-side

See the lookahead leak on real filings — 40 companies, 6,969 rows, CC0, no signup. Three minutes in a notebook, before you decide anything.

This table is deliberately not trying to make EDGAR look bad — it can't, it's free and it's correct. The comparison is about who spends the engineering hours: you, once, or us, continuously, for $29/mo.

When you should build it yourself

Be honest with yourself about this list, because a lot of it applies to more people than vendors like to admit:

• You only need a small, fixed set of tickers (say, under 30) and can hand-verify tag mappings once.
• You want non-US, delisted, or private-company coverage, which EDGAR-only tools (ours included) don't have.
• You need the raw dimensional/segment-level detail rather than the flattened, reconciled concepts.
• You have the engineering time and actually enjoy owning this kind of pipeline — some people do, and it's a legitimate choice, not a failure to "just buy the product."
• You want Parquet natively, or a data warehouse-native format, right now — we don't include that.

If any of that is you, go build it. Start with our free sample repo (github.com/christianpichichero-max/pit-fundamentals — 6,969 rows, 40 companies, full methodology documented, no signup) to see exactly how we handled tag mapping, restatement flags, and Q4 derivation, and take whatever's useful. A competent engineer can absolutely replicate this. We're not claiming otherwise.

When Tradevo (or another vendor) is the better choice

If you'd rather spend your weeks on the strategy instead of the pipeline — if you want 5,170 US companies, up to 12 years of annual history, first-filed vs. latest values, and restatement flags already reconciled, queryable by as_of date on day one — that's what the $29/mo buys: not data you couldn't get, but the weeks you'd otherwise spend building the parts list above.

But Tradevo isn't the only, or automatically the right, paid option. If your priority is longer history, non-US or delisted coverage, a broader concept set, or a different data format or research-stack integration, Sharadar, Tiingo, or QuantConnect may genuinely fit better — they solve overlapping problems with different tradeoffs and coverage. Check their pricing pages and docs directly and compare against what's above; we won't quote numbers we don't control.

Further reading on the mechanics: /blog/lookahead-bias-fundamental-backtests and /blog/point-in-time-fundamentals-data walk through the first-filed/restated distinction in more depth.

Try the raw approach first — it costs nothing but time: github.com/christianpichichero-max/pit-fundamentals. If you'd rather skip straight to a working as_of API, the whole thing is at https://tradevodata.com/?utm_content=alt-sec-edgar-direct for $29/mo, cancel anytime.

Not investment advice. Verify any competitor's pricing and coverage yourself on their own site.