How to build a bank peer group

Start with the group the FFIEC assigned

Peer pages rarely fall apart over the arithmetic. They fall apart when a director points at one bank on the list and asks why it's there, and the only answer is that it came through the screen. After that, every figure on the page is suspect, the flattering ones most of all.

The safest start is a group nobody at your bank chose. The FFIEC's Uniform Bank Performance Report shows each commercial and savings bank's ratios beside its peer group average and its percentile rank, and the FFIEC's Central Data Repository publishes it free.

The groups changed this year. Since February 26, 2026, the FFIEC has grouped commercial banks with average assets up to $300 million by asset size alone, without regard to office count or metro location, which cut the commercial bank groups from 17 to 8. It applied the new groups to Q4 2024 through Q4 2025 first, then to older quarters. A bank under $300 million now sits with every commercial bank its size, rural or metro, and a UBPR peer figure in last year's board deck may rest on a grouping that no longer exists.

Here is how all 4,296 institutions that filed a call report for June 30, 2026 fall across the groups.

Source: BankingLens, from the FFIEC's UBPR peer group file and call reports for the quarter ending June 30, 2026. Banks are group members that filed; median ROA is year to date, annualized. Not ranked: fewer than five banks, so our scorecard computes no medians or percentiles. n/a: no ROA median in our data. De novo groups for 2019 to 2021 were empty.

Charter type changes the benchmark

Run down the ROA column. The commercial groups from $100 million up sit between 1.16% and 1.36%. The savings bank groups run 0.87%, 0.64%, 0.59% and 0.38%, largest to smallest. Match the sizes and the gap holds: between $300 million and $1 billion the median commercial bank earned 1.35% and the median savings bank 0.64%, and one band down it was 1.29% against 0.59%.

A gap that wide at every size comes from the business model long before it comes from management. Benchmark a savings bank against commercial banks and it starts every comparison behind. Let a few savings banks into a commercial bank's hand-built list and the bar drops without anyone saying so.

De novo banks sit lower still: a median of 0.34% for banks opened in 2022, -0.26% for 2023 and -0.94% for 2024. A balance sheet built from nothing takes years to pay its way, and on an established bank's list a de novo drags the median down for reasons that say nothing about the established bank.

The specialty groups make the point from the other side. The seven credit card banks over $3 billion posted a median ROA of 2.53%. The non-deposit trust companies posted 10.89%, because a company that takes no deposits and makes almost no loans has very few assets to divide by. A single one of either can wreck the average of a small peer list.

When the asset band isn't enough

For commercial banks, the FFIEC group is asset size and nothing else. That's the right place to start, and for some questions the wrong place to stop.

Subchapter S banks

An S corporation passes its income through to shareholders, who pay the federal income tax on it, so an S corporation bank's after-tax ROA beats a C corporation's running the identical business. In a mixed list, compare ROA before taxes. You don't have to guess who elected: the FFIEC 031 and 041 instructions for Schedule RI include a memoranda item asking whether a Subchapter S election is in effect.

Banks on the community bank leverage ratio

In Q2 2026, 1,798 of the 4,238 insured banks reported a leverage ratio and no risk-based capital ratios, which is what a bank that has elected the community bank leverage ratio (CBLR) framework files. That was 671 of the 1,214 banks between $100 million and $300 million, 36 of 264 between $3 billion and $10 billion, and none above $10 billion. Put total risk-based capital on a community bank peer page and a large share of the rows come back blank. The tier 1 leverage ratio is the one capital ratio all 4,238 reported. These June 30 counts predate a final rule, effective July 1, 2026, that lowered the CBLR requirement from more than 9% to more than 8%, so recheck the split each quarter.

Concentrated CRE and farm lenders

Two banks of the same size can hold entirely different loan books, and Schedule RC-C Part I, Loans and Leases, shows it. The 2006 interagency CRE guidance flags a bank when construction and land loans reach 100% of total capital, or when non-owner-occupied CRE reaches 300% of total capital and the CRE book has grown 50% or more over 36 months. Those are screens, not limits (our CRE concentration explainer covers them), but a bank past them belongs beside other concentrated lenders. Use one denominator for every peer: since 2020 the agencies have accepted tier 1 capital plus the allowance as that denominator, because CBLR banks no longer report tier 2 capital. Farm lenders are the other obvious split, and Schedule RC-C Part II, Loans to Small Businesses and Small Farms, helps find them.

Trust-heavy banks

A big trust department carries payroll with no loans attached, so the bank's efficiency ratio looks heavy beside spread lenders, while fee income that needs little balance sheet can flatter its ROA. Schedule RC-T, Fiduciary and Related Services, shows how big the business is.

Banks close to a band line

A bank just under $3 billion is measured against banks as small as $1 billion. Once over the line it's measured against banks up to $10 billion, and its percentiles move though its ratios didn't. Near a boundary in either direction, add the banks just across it or show both groups.

Recent mergers

An acquisition steps up the balance sheet in one quarter, while ratios built on average assets and year-to-date income catch up over the quarters that follow, so for a while the acquirer's ROA and efficiency ratio describe neither bank. Footnote those quarters if your bank did the deal. If a peer did, leave it off the list until it has a clean year of combined results.

Building a list you can defend

When the question needs more than the FFIEC group, build the list in this order.

1. Write the board's question down first. "Are we earning enough for our size?" is answered by the FFIEC group. "Are we overpaying for deposits in our markets?" isn't, because none of the eight commercial groups looks at where a bank operates. Pick the banks before the question and the page ends up flattering whoever built it.
2. Use a few hard filters and write each one down: charter and asset range, then one or two that fit the question, such as capital framework, Subchapter S status, a CRE concentration screen or a farm loan share. Every filter shrinks the list, and the smaller it gets, the more one bad quarter moves the median. Our scorecard won't compute a percentile for a group of fewer than five banks, and a list that small is already fragile. For a list you pick yourself, the FFIEC's custom peer group report is free.
3. Read the list by name and cut what doesn't belong: a lender that's mostly a card issuer, a bank run for its trust department, a de novo, a bank that closed an acquisition last quarter. Write a one-line reason for each removal. When a director asks why a bank isn't on the page, the answer is already in the file.
4. Freeze the list for the year. If peers come and go, the median moves when nobody's ratios did. When a peer is acquired, remove it from every quarter shown, not just the latest, so the trend still compares the same banks.
5. Report medians and quartiles, not averages. Even the FFIEC trims its UBPR peer averages, dropping banks above the 95th and below the 5th percentile for each ratio, which also means averages for different ratios come from different banks and can't be combined. A median needs no trimming. Quartiles show how wide normal is: the middle half of group 5 earned an ROA between 0.98% and 1.75%, while group 1's middle half sat between 1.02% and 1.34%.
6. Show the percentile beside the ratio and the group median. A percentile alone hides scale: climbing a quartile in group 5 takes a far bigger move in ROA than in group 1. Check the direction, too. Our scorecard orients every percentile so higher is better, which makes a low efficiency ratio percentile a sign of heavy costs; confirm how any other report runs before its percentiles reach the board.
7. Pair ratios that explain each other. ROA is the result; net interest margin and the efficiency ratio show whether it came from revenue or cost. The nonperforming loan ratio needs reserve coverage beside it, since a high NPL ratio fully reserved is a different conversation from a low one thinly reserved. The loan-to-deposit ratio belongs next to cost of funds, because a bank that has lent out most of its deposits may be paying up for the next dollar.

A worked example with Frost Bank

Frost Bank, the bank on our public sample scorecard, sits in peer group 2, commercial banks between $10 billion and $100 billion. The scorecard counts 116 banks in the group; 114 of them filed a call report for June 30, 2026, the number in the table above. Its Peer group tab ranks each ratio within the group; here are three.

Source: BankingLens sample scorecard, quarter ending June 30, 2026. Percentiles are computed against the 114 group members that filed, and higher is better on every ratio.

Read the rows together, as a director will. An ROA of 1.31% against a group median of 1.25% is a touch above the middle, and a margin at the 51st percentile is the middle. The efficiency ratio at the 26th percentile means higher costs relative to revenue than about three quarters of the group.

Those don't reconcile on their face. A middling margin and a cost ratio most of the group beats would usually mean a below-median ROA, so something outside the three ratios is making up the difference: fee income relative to assets, the provision, taxes, or the share of assets earning interest. That's the question for the next page of the board package, and the scorecard's Call report tab, with the income statement line by line and a year-ago column, is where to answer it.

Frequently asked questions

Related reading

• Call report vs UBPR: what's the difference?
• What is a good ROA for a bank?
• What is a good tier 1 leverage ratio for a bank?
• Bank peer benchmarking tools compared
• How FFIEC peer groups work