Find & Compare Banks is one page with four readings of one set of banks: draw a market on the map (or load a saved group), then compare those banks, screen inside them, or combine two of them. The set is the noun; Compare, Screen and M&A are what you do with it, and the map never leaves the screen.
A group is the set of institutions your readings compare against. You build it, you name what it is for, and its definition travels with every figure it produces — so a peer line can always be traced back to the rule that chose the peers.
Three ways to populate a group
You choose the population method when you create the group, and it is fixed afterwards. Changing how a group is populated means building a new one, which is deliberate: a group whose method changed underneath you would silently invalidate everything you compared with it.
- Manual — you name the members. The list is the definition.
- Criteria — states, charter classes and an asset range. Membership stays live: institutions move in and out as they file.
- Geographic — counties, states, a radius around a ZIP-code area, or a polygon you supply as GeoJSON. Membership is resolved from branch locations in the FDIC Summary of Deposits.
Geographic groups, and what "in the area" means
Geographic membership is measured in deposits, not branch headcount: an institution is in the group when the deposits it holds at branches inside your shape meet the floor you set. The floor is required and you set it — the form offers a suggestion with the evidence for that suggestion stated, but no engine default is applied on your behalf.
Two facts about geography that the surface states and the guide will repeat, because they change who is in the group: counties and states are read from where the branches sit, not where the charter is; and a radius is measured from the centroid of a ZIP code area, not from an address.
Summary of Deposits is annual, as of 30 June, so a geographic group carries a vintage. The surface names the vintage and how stale it is rather than presenting it as current.
The map, and drawing a market
Groups has two views, List and Map, and the choice is in the address so a link opens on the view you sent. List stays the default because it is the complete one: the map can only plot members whose branches carry coordinates in the Summary of Deposits vintage, and it always states how many members it could not plot rather than quietly showing fewer banks than the group has.
On the map you can draw the market instead of describing it:
- Draw a circle — click the centre of the market, then click again to set the radius. The instruction line tells you which click comes next.
- Draw a shape — click each corner of a polygon and finish it when you have at least three. Cancel abandons it.
A drawn circle is stored as a 72-segment ring rather than as a true circle, and the confirmation says so, with how far inside the true circle that ring sits. A shape you draw does not become a group on its own either: the name, the purpose and the deposit floor are still yours to supply, exactly as they are for a typed geographic definition.
The basemap is OpenStreetMap raster tiles, and the surface says so — the branch locations plotted on it are FDIC Summary of Deposits data, not OpenStreetMap data. The map itself is only downloaded when you open it, so the list view never pays for it.
Membership is live, and its changes are recorded
Criteria and geographic groups can be re-evaluated. The result is a diff, not a silent replacement: entries, exits, retained members, and two categories that are the point of the design —
- Excluded, preserved — institutions that still match the definition but that you hand-pruned. Exclusions are sticky until you restore them, and you can record why.
- Pinned, preserved — institutions that no longer match the definition but that you pinned in. They never drop out on re-evaluation.
Entries and exits are logged as quarterly events with the report period and the date observed, so a comparison you ran last quarter can be explained this quarter.
Position, not percentile
A custom group reports your ordered position among observed values— "3 of 12, ranked highest first" — and an untrimmed median computed from the member values themselves. It never reports an interpolated percentile. A percentile implies a distribution somebody published; for a group you assembled this morning, nobody did.
Percentiles appear in exactly one place: the FFIEC regulatory peer line, where FFIEC publishes them. That figure is a trimmed average — FFIEC drops observations beyond the 5th and 95th percentiles before averaging — and it is labelled as theirs wherever it renders. It is never called a median.
Small groups are suppressed, not shown thin
Below five members on a given measure, no statistics are computed at all. The surface prints the reason instead of a median and a rank, because for a group that size both are noise. You still see the member values — those are reads, not statistics.
Building your first group
- Open Find & Compare Banks and choose a population method. If you are unsure, criteria is the most forgiving: it stays live and you can pin or exclude afterwards.
- Say what the group is for — market, direct competitors, board reporting, aspirational. It is free text and it is the thing future-you will need.
- Save, then read the definition card. That JSON is what travels with every export; if it does not describe the group you meant, fix it now.
- Mark one group as your default. That is the one the institution view's peer lines and the cohort bands will use.
- Open Comparison to see the whole group as a table.
A group you built yourself can be made to say almost anything. That is not a flaw — it is what makes a custom group useful. The safeguard is that its definition is never separable from its output, and FFIEC's published peer figure stays in view beside yours. Nobody reading a figure you produced has to take the peer set on trust.