Every SEC filing, traced to its source.
We pull balance sheets straight from EDGAR, verify each one against the accounting identity (A = L + E), and tell you exactly what we found. When a filing reconciles, you get the figure. When it doesn’t, you get the reason — never a silent wrong number.
Live from the filings
| Deposits with banks | $285.1B6% |
|---|---|
| Investment securities | $536.0B11% |
| Loans | $1.52T30% |
| Goodwill | $52.7B1% |
| Other assetsincludes 3 line items this filer does not report separately | $2.62T52% |
| Customer deposits | $2.71T54% |
|---|---|
| Short-term borrowings | $692.1B14% |
| Other liabilitiesincludes 1 line item this filer does not report separately | $1.23T25% |
| Shareholders' equity | $374.6B7% |
Both columns are the same height because they are the same money, counted twice: once by what it is, once by who it belongs to. Open the full page for JPM.
Data is for informational purposes only. Not financial advice.
Data & method
Source
SEC's own Financial Statement Data Sets, plus the XBRL frames API for anything filed since the last quarterly dataset published. Nothing is scraped, bought, or estimated.
Freshness
The loader decides it is behind by asking the database, not by watching a clock, so a container that was down for a week closes the gap on its next tick instead of waiting for a schedule. Most recent filing in the load: 2026-09-11.
As reported, never restated
Where a figure has been filed more than once, the earliest filing wins. You see what the company said at the time, not what it said later about the same quarter.
The check
All 5,809 companies with a complete balance sheet are checked against assets = liabilities + equity; any that do not balance are flagged with the reason. The rest are drawn with what is missing named, rather than quietly balanced.
One row, not twenty-three
SEC's data carries the same figure many times per filing — by segment, by geography, by legal entity. Exactly one of them is the consolidated company, and isolating it is most of what this does.
What it doesn't do
It draws what was filed and says so when it cannot. Nothing here is an opinion about what a company is worth, or about what it is going to do next.
Checked against the accounting identity, not sampled. The failure this avoids is the SEC duplicate-tag problem — JPMorgan reports “Total Assets” 23 times in one filing, once per segment and subsidiary — solved by isolating the consolidated row. Why XBRL data is wrong one time in five, and get the data.
Every valid SEC filing we ingest reconciles to the accounting identity. When a filing doesn’t balance, we flag the exact reason — noncontrolling interests, mezzanine equity, rounding, or a broken filing — never silently fudged. How the check works.
Weighing this against something else? The SEC filings APIs worth considering for quant work is the roundup, written to be useful even where it does not conclude in my favour.
Same scale rules, other companies
Each drawing is that company's own balance sheet at its own proportions. They look nothing alike because they are nothing alike.
Pricing
The drawings are free and always will be. The machine-readable version is what costs money — and it comes two ways. The dataset is a photograph: one CSV, downloaded once, fixed forever. The API is a window: live data, current every time you call it. Full comparison and FAQ.
$0
Unlimited — look up as many companies as you like
- Every company, every balance sheet, no key and no account
- The live demo is uncapped too — try the real endpoint on anything
- 10 keyed API calls a month if you want JSON in bulk
- Best for: reading the site, and evaluating the API
$79.99 once
Static snapshot of all company data as of 12 September 2026
- One-time download — no updates
- 1.6M rows as one CSV file
- Best for: one-time analysis, research, Excel work
- Data is fixed — it does not change
$49/month
Live, up-to-date data
- Programmatic access — query any company anytime
- 10k API calls per month
- Best for: algorithmic trading, dashboards, ongoing research
- Data updates daily — you always get the latest filings
$490/year
The same Pro access, paid yearly — save $98
- Everything in Pro
- 10k API calls per month
- Two months free against the monthly price
- Best for: a workflow you already know you are keeping
Paid plans go through Stripe. Your card details are entered on Stripe's page and never reach this site. What is the difference between the dataset and the API?
Live demo
The real endpoint, the real data, no key needed. As many companies as you like — there is no limit on looking.
What you're looking at
Both columns are the same height because they are the same money, counted twice. The left column is what the company owns, sorted by what it is. The right column is who has a claim on it — lenders and suppliers first, then whatever is left over for the owners. Every band is drawn at the size the company reported, so a bar twice as tall is twice the money.
An example with round numbers, not a real company.
The numbers behind it
Why it's harder than it looks
Filings don't say things once
SEC's data carries the same figure many times per company per quarter — broken out by segment, by geography, by legal entity, by fair-value level. JPMorgan reports “total assets” twenty-three separate times in one filing. Exactly one of those is the company. Take the wrong one and you get $641 billion instead of $4.4 trillion, and nothing about it looks wrong.
Balance sheet items and income items are different kinds of fact
One is a photograph, the other is a film. A balance sheet figure is an instant — what was there on one day. Revenue is a duration — what happened over three months. Read a duration where you needed an instant and you get the change in assets rather than assets, which is how a company ends up with a negative total.
Every industry files differently
A bank doesn't report inventory; it reports loans and deposits. Look for the retail tags on a bank and you get a grey rectangle. The tag names move too — the short names most people use are deprecated, and the modern bank tags carry an “ExcludingAccruedInterest” suffix from a 2020 accounting standard.
Every tag was confirmed against the raw filing data before being used. Nothing was mapped on the strength of it sounding right.
What this doesn't do
- No predictions.
- No scores.
- No recommendations.
- Nothing estimated — where a company doesn't report something, the page says so rather than showing zero.
- Every figure traces to a filing, with the date it was filed.