A bank rating you can check by hand: how Bank Index scores 27,973 institutions, and why a one-office bank beat a 9.6 billion one

On 23 September I wrote here that three active US banks share the name Eagle Bank, and that the directory I was reading listed only one of them. I went back to it this morning. All three are there now, and so is EagleBank in Bethesda, written as one word. Every card carries a score out of 10.

I had assumed the scores would follow size. They do not, and the order surprised me. Eagle Bank in Polson, Montana has one office and 179.5 million dollars of assets, and it scores 5.8. EagleBank in Bethesda has 14 offices and 9.62 billion and scores 5.4. The Everett bank, open since 1889, scores 4.8.

That is either a rating telling me something I did not know, or a rating making a mistake in public. I wanted to know which before pointing anyone at it. So I did what I would do with any bank rating before trusting it with a payroll account, which is to open the cards one at a time, copy every line into a notebook and redo the arithmetic by hand, using nothing but the numbers the page itself prints.

The bank rating nobody is allowed to show you

Every insured bank in the country already has a grade. Examiners give each one a composite CAMELS rating from 1 to 5, built from capital adequacy, asset quality, management, earnings, liquidity and sensitivity to market risk. It is the best informed judgement anyone makes about a bank. The people making it have read the loan files.

You will never see it, and neither will I, because an interagency advisory the FDIC sent out as FIL-13-2005 reminds banks that they are prohibited by law from disclosing CAMELS ratings to outside parties without the regulator's permission. Anyone who discloses or uses that information without permission may face criminal penalties under section 641 of title 18 of the US Code. That is not a polite request.

The letter was written because insurers had started asking banks for their ratings when pricing cover for directors and officers. The agencies told them to stop asking. I find that detail oddly clarifying. If an insurer writing a policy for the bank's own board cannot see the grade, a small company opening a checking account has no chance at all.

The rules may loosen a little. On 19 May 2026 the FFIEC asked for comments on the first comprehensive revision of CAMELS in 30 years, and its chair, Michelle Bowman, said the revised framework “marks a decisive shift toward transparency, quantitative factors, and predictability of supervisory oversight.” I read the release looking for a sentence that would let a depositor see the grade, and I have not found one. As far as I can tell, the transparency is aimed at the banks being graded.

So every public bank safety rating is a proxy. It has to be assembled from what regulators do publish: call report ratios, capital, earnings, enforcement orders, the licence register. Agency credit ratings help at the very top, and the JPMorgan Chase card quotes bank-level ratings of Aa2, AA- and AA. The Everett card quotes no agency rating at all. It has the register, and that is it.

What the index holds, by its own count

The front page of Bank Index, last updated at 18:15 UTC on 23 September, counts 106,244 financial companies, of which 45,142 are banks and account providers. It keeps two groups outside that total and says so on the page: 4,239 non-financial companies and 26,901 US money-transfer agents. I like that it says so, because padding the headline number with them would have been easy and nobody reading the front page would have been any the wiser.

The banks section runs to 14,232 banks in 165 countries. For the United States the index lists 24,384 companies, with 4,509 banks and 4,641 credit unions. The biggest single American group is not banks at all but 7,677 financial advisers and accounting firms, followed by 5,058 fund managers and 2,271 brokers, most of whom a company looking for a checking account will never need, though a founder raising money probably will. That is what you get when a directory is built from public registers and reads every one of them.

Not everything listed carries a score. The ranking covered 27,973 rated institutions averaging 4.3 out of 10, and the page counts 20,830 licences confirmed on a regulator's own list. The regional split is where it gets strange. The 9,541 rated in the United States average 3.9, against 4.5 for Europe, 4.8 for the United Kingdom and 3.4 for Canada.

I do not know why the US sits that low. Nothing on the page explains it. I would not read it as a verdict on American banking either. The same cards give the United States 87 out of 100 on the World Bank governance indicators, the measure the index uses for how strong a country's regulation is, while Finland and Denmark top that list at 100.

How one line becomes one score

The method fits on a postcard, and the site prints it under the ranking. Each card scores a company from 0 to 10 on each need, in steps of 0.5. Every score gets one sentence of reason and a link to where the fact came from. The overall is the plain average of the needs that are not about geography, so a Swiss bank is not punished for having no American branch. Ties go to the bigger balance sheet.

A pair of rules sits on top of the arithmetic. An overall score needs at least three scored needs, because one good number is not a verdict. And a company whose licence no regulator's own list confirms cannot rank above one that is confirmed. The site says an automatic check reruns both rules before each update and refuses to publish if the top of the ranking breaks them, and while I have no way to audit that check from the outside, the rules themselves are printed where anyone can hold a card up against them and see whether it fits.

That is the whole method as published. What interested me was what happens when those rules meet a bank nobody on the team has read by hand yet.

The Eagle arithmetic

Here is the Polson card, line by line, in the notebook, which felt old-fashioned and turned out to be the right call.

Lowest fees scores 7.5, because the bank's own business checking page says “No monthly service charge”. The reason line explains that the ceiling is 7.5 rather than 10 because transfer and card prices were not read. Financial strength also scores 7.5, from FDIC filings for 30 June 2026: equity at 11.5 per cent of assets and a return on assets of 2.29 per cent.

Cross-border payments get 5.5 on this card. Interest and trade get 5.0 each, and onboarding 4.5. Each of those carries the same reason, word for word: “Holds a banking licence, so it can hold deposits and run payments; the actual terms have not been checked yet.” The United States line scores 6.0 but is geographic, so it stays out. Six lines, 35 points, an average of 5.83, shown as 5.8.

EagleBank in Bethesda earns the same 7.5 for strength, on equity of 12.2 per cent, a tier 1 capital ratio of 14.9 per cent and a return on assets of 0.52 per cent. Its fee line is lower, at 5.0. Its own page says the “Monthly maintenance fee of $15 is waived if you maintain the balance requirement”, which is still a fee, even a waivable one. The other lines match Polson exactly. Six lines add up to 32.5 points, or 5.42.

So the whole gap between a one-office Montana bank and a Maryland bank more than 50 times its size comes down to one sentence on each bank's fee page. I find it hard to read past that. But I think it is also the honest way to show it. The card tells you exactly which sentence moved the number, and you can go and read that sentence yourself.

A monthly fee is a real cost for a company with two staff. A return on assets of 2.29 per cent at a single office is real too. What the overall does not tell you on its own is that most of the lines on both cards are defaults. The card does tell you, in plain words, if you scroll.

The Everett card shows the defaults with nothing on top. Where the other two point to filings, it says “Scores set by its kind of institution”, and lower down, “We have not checked this company by hand yet.” Strength is 4.0, placed by total assets from the register rather than by ratios. Fees sit at 5.0 because no fee page has been read. Six lines make 29 points, which is 4.83.

I had missed this the first time through. The lowest of the three Eagle Banks is the one whose filings nobody has read yet. That says nothing so far about whether it is the weakest bank of the three. Still, a reader who stops at the overall, which on a phone screen is the only number big enough to notice, would take that 4.8 for a judgement on a bank that has been open since 1889, and that bothers me more than the Montana result does.

Big bank, small fintech, 0.1 apart

The hand-checked cards read very differently. JPMorgan Chase scores 6.1 across 13 needs, every one backed by filings or its own pages. Mercury, which its card calls a San Francisco fintech and not a bank, scores 6.0 on 13 needs too. On the overall they are neighbours.

In every other way they are nothing alike. JPMorgan gets 10 for financial strength, on 4.42 trillion dollars of assets and those agency ratings, and 9.0 for trade finance. It gets 4.5 for fees, because incoming wires cost 15 dollars and outgoing ones 25 to 50. Foreign-owned companies get 4.0, because Chase wants an SSN or ITIN from every owner.

Mercury is close to a mirror image. Fees score 9.5, with no monthly charge on the base plan and free domestic wires, and founders living abroad get 8.5. Strength gets 4.5, since it is not a chartered bank. The card adds that in April 2026 the OCC gave preliminary conditional approval to a Mercury Bank charter that needs at least 300 million dollars of paid-in capital. Trade gets 1.5, because there are no letters of credit.

Read only the overall and you would call them equivalent. In practice they are equivalent for nobody. Picture a software company whose co-founder lives in Lisbon and who wants free wires and an online application, next to a furniture importer paying a supplier in Vietnam by letter of credit and needing a bank that will still be standing when the container lands. Each would find one of these close to useless. They would pick opposite ones, and both would be right.

How to compare business bank accounts with it

The front page is built for exactly this, and it is where I would start if I had to compare business bank accounts this week. You pick one or more needs from three groups the page calls Getting in, Money, and Scale and crypto, and the ranking, the map and the counts all recalculate for that choice, right down to the average score printed at the top of the page. On my reading the page reported the recount at 308 milliseconds, which is quicker than I can decide what I need.

It also shows a top three for the needs people ask about most. With no filter, Lowest fees was led by Starling Bank and Mercury at 9.5, with Wise at 8.5. Fast onboarding went to Revolut at 9.0, then Mercury and Tide. Non-resident owners went to Mercury at 8.5, then Wise and Payoneer at 8.0. Mercury turns up in all three lists.

I went looking for business bank ratings that show their working. This is the first one where I could redo every number myself. The reading order I would suggest comes from the Eagle cards, where I got it wrong before I got it right.

Open the card, not the ranking row. The line under the heading that starts How this is built tells you whether the scores are backed by filings or set by the kind of institution. A score of the second sort is a placeholder with a number on it, and it deserves the same weight as a blank.

Read the reason sentence on the two or three needs your business really has, which for a company paying staff in one state and a supplier in another is usually fees, onboarding and cross-border payments, and let the rest of the card wait. If it says the terms have not been checked yet, treat that line as empty. If it quotes the bank's own fee page, click through anyway. Fee pages change. The card only knows what the page said on the day it was read.

Follow the source link to the FDIC record and write down the certificate number, which the earlier piece on this site explains. Polson is certificate 58282, Bethesda 34742, Glenwood 8824 and Everett 90191. Check the namesakes too, on the list of US banks under E, now 85 cards long, before you trust a name.

An aside about the name

A short digression, since the search results mix the two. Bank index is also what people call the KBW Nasdaq Bank Index, a stock market measure that follows 24 publicly traded US banks and thrifts. It has nothing to do with this directory. No score here moves with a share price, and a five-office bank like the one in Glenwood is scored on exactly the same lines as JPMorgan.

Back to the point. A share index tells you how investors feel about banks this morning. The directory tries to tell you something duller and more useful, which is whether a given bank will take a company like yours at all, what it will charge once it does, and whether anyone has actually read the page that says so. Those are different questions, and it is odd how often the same two words stand for both.

What I still cannot resolve

The index lists 4,509 US banks. The FDIC's BankFind data I read on 23 September held 4,232 active institutions. I have not reconciled the 277 extra line by line, and I cannot tell you from the outside how many are trust companies, uninsured charters or records that ought to have been merged into another card.

I suspect the Everett score will move once its filings are read, since the other two Eagle Banks earned their 7.5 for strength exactly that way. I would bet on up rather than down. That is a guess, and the card will settle it in public.

I have thought about the Everett card more than is reasonable. The bank has been open for 137 years. It took a certificate number to get it into this directory, and it arrived with a 4.8 that, by the card's own admission, nobody has checked yet.