Learn
Where Our Data Comes From
Published · 10 min read
Every number on Moneyta traces back to a source you can name, and most of them to a source you can download yourself. This page lists each dataset we ingest: what it is, how often it updates, how far it inherently lags reality, and what we check before we let it near a page you read. We publish this for the same reason our research pages cite their filings. A number without a source is an opinion wearing a costume.
One principle runs through all of it: when data is missing, wrong, or too thin to trust, we say so on the page rather than papering over it. You will see dashes, coverage percentages, and notes like "2 events so far, honest rates appear at 30" across the product. Those are this page, applied.
SEC quarterly holdings filings: what large managers hold
Any institution managing over $100 million in US-listed stocks must disclose its holdings to the SEC every quarter. We load these two ways. The SEC's structured quarterly data sets give us every filing in bulk, and we load full history from 2020 onward for the 500 largest managers plus complete current quarters for every filer, roughly 9,500 institutions. Between quarterly drops, we watch EDGAR's real-time feed so new filings appear within hours of acceptance, and the bulk data supersedes the real-time copy per filing once it arrives.
Update cadence
Real-time feed checked continuously; bulk quarterly data sets loaded when the SEC publishes them
Inherent lag
Filings arrive up to 45 days after quarter end, by law. Every page that shows this filing data says so
How we validate it
Row totals are reconciled against each filer's own cover-page summary, and pre-2023 files carry a units check because older filings reported values in thousands
SEC insider transactions: Forms 3, 4 and 5
Company officers, directors and large owners must report their own trades, usually within two business days. We parse the filings themselves, including whether a trade rode a pre-scheduled 10b5-1 plan, and we load the SEC's structured insider data sets for bulk coverage. Our stored history begins in 2026, so insider base rates are young, and the evidence pages refuse to show a rate until at least 30 closed events exist. History accrues nightly.
Update cadence
Real-time feed checked continuously; bulk quarterly data sets as published
Inherent lag
Roughly two business days from the insider's trade to the filing
How we validate it
Amendments are linked to what they amend; duplicate and superseded rows are resolved by accession, never by guesswork
SEC XBRL company facts: the fundamentals
Revenue, income, shares outstanding, debt and every other fundamental figure come straight from the structured XBRL data companies file with the SEC. We store facts point-in-time, meaning each figure is tied to the date it was filed, so a backtest never sees a number before the market could have. When a company restates a figure, the latest filing wins and the history of what was previously reported remains.
Update cadence
Refreshed as companies file; quarterly rhythm in practice
Inherent lag
None beyond the company's own filing schedule
How we validate it
Values are used as filed, never retyped or estimated. Derived figures such as trailing-twelve-month EPS require four complete quarters or they render as a dash
SEC Form ADV: who the advisers are
Registered investment advisers file Form ADV, which carries the firm's own account of itself: office location, website, employee count, and total regulatory assets under management. We load the full registry monthly, about 23,600 firms, and join it to quarterly holdings filers where the adviser reported its EDGAR identifier. This is why a manager page can show two different asset numbers side by side: the holdings report shows the disclosed US stock book, Form ADV shows everything the firm manages. Banks, insurers and operating companies file holdings reports without filing ADV, so their pages simply omit the adviser facts rather than inventing them.
Update cadence
Monthly, from the SEC's public compilation drops
Inherent lag
Adviser-reported, refreshed on the SEC's monthly schedule
How we validate it
Two independent drops are cross-joined by the adviser's registry number; fields we cannot source stay empty
Market prices: Tiingo
Daily prices, volumes, and split- and dividend-adjusted closes come from Tiingo, a licensed market data provider. Adjusted closes are what make total-return math honest, because they fold dividends and splits into one series. The bid and ask on ticker pages come from Tiingo's IEX feed and are labeled as one venue's top-of-book, not the consolidated market, with a dash when the venue is quiet. Fund fee data rides the same pipeline.
Update cadence
End-of-day bars nightly; the bid/ask quote is cached for seconds, not hours
Inherent lag
End-of-day data is exactly that; nothing pretends to be intraday except the labeled quote
How we validate it
Bars are checked for gaps nightly and backfilled; a signal computed on thin history says so or does not compute
Company profiles and analyst ratings
Company descriptions, sector classifications, executive lists and third-party analyst consensus come from a commercial market data provider. We treat these as context, never as evidence: analyst ratings appear clearly labeled as third-party opinion and never enter Moneyta's own sentiment score. Executive lists link to public LinkedIn searches, not to purchased profile data.
Update cadence
Cached and refreshed on a rolling schedule
Inherent lag
Vendor-dependent; the as-of date travels with the data
How we validate it
Displayed for context with attribution; where the vendor and the consensus disagree with each other we show the discrepancy rather than reconciling it silently
Interest rates: FRED
The risk-free rate behind risk-adjusted return math comes from the Federal Reserve's FRED service, specifically the three-month Treasury series. It is public data from the primary source.
Update cadence
Daily
Inherent lag
One business day
How we validate it
Used as published; the observation date is stored with the rate
What we deliberately do not do
- We never estimate a missing number to fill a gap. Absence renders as absence.
- We never backdate. Historical analysis only sees data that was public on the date in question, including the 45-day lag on quarterly holdings filings and filing dates on fundamentals.
- We never let public pages trigger vendor calls. Public pages are served from our own store, so a traffic spike cannot burn through rate limits and quietly degrade the data.
- We never show a statistic without its base: event counts, universe sizes and coverage percentages travel with every rate we publish.
Where to see this in the product
The evidence pages show what historically followed every signal we surface, computed from the price and filing history described above. The fund manager directory renders the holdings and Form ADV data directly. And every research report cites the specific filing behind each claim, with an accession number you can open on EDGAR yourself.
Found something that looks wrong? Tell us at [email protected]. A corrections process is part of publishing computed claims, and we would rather fix a number than defend it.