How banks decide which branches to keep, merge or close

A bank decides where to keep branches by scoring every branch on the same few questions: how many people live around it, how busy the area is during the day, how close the bank's next branch is, and how many rival branches and cash machines serve the same streets. A branch with many residents, a busy area and no other branch of the bank nearby is hard to close without losing customers. One with a sister branch a few minutes away and rivals all around is a candidate to merge. The bank's own figures on transactions and accounts then confirm or overturn the ranking.

This article is for branch network, retail distribution and property teams in banks and credit unions, and for the strategy teams that brief the board. It covers the measures, how to read them, a worked example, the mistakes that turn a network review into a public complaint, and an AI employee that keeps every branch's score current.

The four measures for each branch

• Residents nearby. Demographics counts the people living within a chosen distance, with the share aged 60 and over, who rely on branches more. Population figures exist only in some countries, among them France, Canada, India, Thailand, Egypt and Australia; not the United Kingdom or the United States.
• Daytime activity. Mobility Footfall gives how busy the area is per day and at its busiest hour. A branch in an office district can serve few residents and many workers.
• The bank's own next branch. Find nearest gives, for each branch, the nearest other branch from the bank's own list and its distance.
• Rivals and cash. Nearby places counts other banks and cash machines around each branch.

How to read a branch's score

Look at each measure as a reason to keep or a reason to merge, not as points to add up blindly:

How each measure points. The bank's own transaction and account figures decide in the end.

Distance alone misleads. Two branches 1 km apart on either side of a railway line can serve different neighbourhoods. A Travel area of a 10-minute walk or drive around each branch shows what each one really reaches, and Demographics can count the residents inside it.

A worked example

Here is a made-up example. A bank has 28 branches in Paris and wants to merge four. The network team uploads the branches as a table and runs the four measures. Three pairs stand out: in each, two branches are within a 10-minute walk of each other, both areas are busy in the day, and five or more other banks sit within 500 m. A fourth candidate is a branch in a residential district whose walking area holds many residents and a high over-60 share, with the nearest sister branch twenty minutes away. The data argues against closing that one, even though its transactions are low.

The review then turns to the bank's own figures: for each pair, which branch carries more accounts and which building costs less. The location score has narrowed 28 branches to six for that discussion, and given the board a reason for the one it should not close.

Mistakes to avoid

• Closing by transaction count alone. A quiet branch may be the only one older residents can reach. The residents measure is there to catch it.
• Using straight-line distance between branches. Use travel time where rivers, railways and main roads cut areas apart.
• Ignoring cash. Count cash machines as well as branches. An area losing its last branch and its last cash machine at once becomes a complaint and a news story.
• Freezing the review. Rivals open and close branches. A score from last year can point the wrong way today.

Doing it with an AI employee

Keep the score running and the next review starts from current facts. The blocks in order: Schedule (monthly), My data (the branches, with an upload of the bank's own list for Find nearest), Find nearest, Travel area, Demographics, Mobility Footfall, Nearby places (banks and cash machines), a Calculated field for the score, an xMap LLM block that explains each branch's score in two lines, and Make a report. Each month the network team's inbox gets the ranked branches, the ones that moved, and a PDF for the committee. Add What changed to be told when a rival branch near one of yours closes.

Start from Branch network review: which branches the area needs, Rival bank branches that closed or moved or Busy places with the fewest ATMs nearby.