How the analysis is built

1. Data source

Every figure on BankingLens originates from the Federal Financial Institutions Examination Council (FFIEC) Call Reports — the quarterly filings every US commercial bank is required to submit. These filings are public domain, downloadable for free from the FFIEC website, and are the same data examiners use.

Our coverage is the universe of US commercial banks — approximately 4,477 institutions as of the latest filing. We don't include credit unions (regulated by the NCUA, different filing schedule), broker-dealers, or non-US banks.

2. Refresh cadence

Banks file call reports 30-60 days after each quarter-end (Q1 by April 30, Q2 by July 30, Q3 by October 30, Q4 by January 30). The FFIEC publishes shortly after the filing window closes. BankingLens ingests within roughly 30 days of the FFIEC's official release, so a quarter's data lands here approximately 60-90 days after the quarter actually ends.

We don't promise real-time data because nothing in this domain is real-time — call reports are quarterly by design, and even regulators see them on the same lag.

3. Peer-group construction

A $400M community bank in rural Texas and JPMorgan don't operate the same business. Comparing them on absolute metrics (ROA, NIM, efficiency ratio) is misleading — the small bank's cost structure, regulatory burden, funding profile, and customer base are radically different. Peer groups are how we make comparisons apples-to-apples.

BankingLens uses two peer-group conventions:

FFIEC asset bands (default)

Six buckets following the FFIEC's published convention for UBPR peer reporting: Under $100M, $100M-$300M, $300M-$1B, $1B-$3B, $3B-$10B, Over $10B. These are the bands examiners use, and they're what shows up in each bank's official UBPR.

Coarse bands (dashboard chips)

For quick filtering, the dashboards offer four coarser buckets: Under $1B, $1B–$10B, $10B–$100B, Over $100B. The percentile rankings themselves still use the fine FFIEC bands — the coarse bands are filter convenience, not comparison cohorts.

A bank's peer group is determined by its total assets at the most recent filing. The dashboards proactively flag banks within 10% of a band boundary because a crossover next quarter would reshuffle that bank's peer cohort and shift its percentile ranks. Peer-group guide →

4. Score derivation

Bank Peer Intel — percentile rankings

For each metric (ROA, NIM, Tier 1 capital ratio, efficiency ratio, NPL ratio), we compute each bank's value, sort the asset-band peer cohort, and assign a percentile (1-99). The dashboards visualize this with quartile-colored chips: green (75th+), slate (50th-74th), amber (25th-49th), red (under 25th). Year-over-year change indicators are computed from the bank's own time series across 5 prior quarters.

Borrower Assist — match composite

For Borrower Assist match scores, we compute five sub-signals per bank-scenario pair, then weight them into a 0-97 composite (intentionally capped under 100 — see why below):

Why 0-97 and not 0-100? The 3-point cap is intentional. We don't claim to predict an approval decision — that involves credit committee judgment, relationship factors, and specifics of the borrower that we can't see. The cap is an honest signal that this is appetite analysis from a balance-sheet position, not a forecast of yes/no.

SBA industry-fit signal (NAICS)

For SBA 7(a) and 504 requests, we add an industry-fit signal — the single highest-weighted input on those scenarios — built from the U.S. Small Business Administration's public 7(a) & 504 FOIA loan-level dataset (published on data.sba.gov). Unlike the call report, the FOIA data names the lender on every approved loan and tags it with the borrower's industry, so we can see which banks actually book loans to a given sector.

We map your stated industry to its 2-digit NAICS sector (for example, restaurants and food service to sector 72, health and dental practices to sector 62, manufacturing to sectors 31–33) and look up each lender's documented SBA volume in that sector over the most recent three completed fiscal years. A lender with a meaningful, repeated track record in your sector scores higher; one with little or no SBA activity in it does not. A lender that has funded your sector specifically in your state earns an additional bump. Lenders are joined to their FFIEC identity by name so the SBA history lines up with the rest of the scorecard.

Two honest caveats. SBA-channel volume is a sample of a bank's lending, not its whole book, so a bank that does heavy conventional commercial lending but little SBA can be understated by this signal alone — which is why it blends with, rather than replaces, the peer-relative signals above. And NAICS is rolled up to the 2-digit sector for ranking, so it reflects sector-level appetite, not a single 6-digit industry code. The signal only applies to SBA loan types; for conventional requests it is inert and scoring falls back to the composite above.

Existing banking relationship

If you tell us the bank you currently use, we factor it in — your existing bank usually has the strongest shot at the loan. The incumbent already holds your deposits and sees your cash flow, can price the relationship, and tends to underwrite faster, so it earns a modest 10% relationship bonus on top of its merit score. We resolve the name you type to a specific institution conservatively: an exact or distinctive core-name match against the FFIEC universe (and the FDIC legal name where we have it), never open-ended fuzzy guessing — a wrong match would be worse than none, so an unrecognized name simply applies no bonus and leaves your ranking unchanged.

The bonus is intentionally small. It can lift your bank into the shortlist when it is otherwise competitive, but it cannot bury a genuinely better-fit lender — you paid for alternatives, not "just call your own bank." A loan-size capacity check still applies: a bank's single-borrower lending limit is roughly 15% of its capital, so if your current bank is too small to legally fund a loan your size, it is not recommended just because you bank there. In that case we don't hide it — the report calls it out and suggests you still ask them, since they may participate alongside another lender or refer you to a correspondent bank. The signal is inert when you leave the field blank.

5. UBPR reconciliation

The FFIEC also publishes derived ratios in the Uniform Bank Performance Report (UBPR) — official, examiner-grade calculations like ROA, ROE, NIM, efficiency ratio. Where the UBPR is available for a quarter, BankingLens reconciles headline ratios against it. The bank scorecard shows the reconciliation status explicitly:

BankingLens is restatement-aware. When a bank refiles a prior quarter, the call-report history is updated in place — meaning a percentile rank from six quarters ago can shift if the underlying numbers change. That's accurate but worth knowing if you reference an old export.

6. Limitations & what we don't do

We try to be useful for the analyst, banker, or borrower making a real decision. Useful means honest about boundaries. Here's the boundary list:

• Not a credit decision. Bank Peer Intel doesn't predict whether a bank will approve a loan. Borrower Assist surfaces likely-fit lenders to call, not pre-approvals.
• Not real-time. Call reports are quarterly, lagged ~30-60 days. Anything more frequent is misleading.
• No qualitative data. We don't see relationship strength, brand reputation, recent leadership changes, or pending M&A. A bank with a perfect balance-sheet profile may still say no for reasons not in the call report.
• No loan-level data. Call reports are aggregates. We don't know which specific loans are in a bank's portfolio.
• No fraud / AML / regulatory action signal. If a bank is under a consent order, you won't see it here — we don't ingest enforcement actions.
• No forecast. We don't predict future rates, future bank performance, or future credit availability. Everything is backward-looking from the most recent filing.
• US-only. FFIEC call reports cover US commercial banks. Credit unions, foreign banks, and broker-dealers are out of scope.

If any of these are dealbreakers for your use case, BankingLens probably isn't the right tool. If they're acceptable trade-offs for fast, transparent, public-data analytics, we'll save you time.