The standard economic case against branch networks relies on cost-per-transaction, a metric that treats every branch visit as interchangeable. My research into what customers are actually visiting branches for in 2026 suggests this metric is measuring the wrong unit of value entirely.
The transactions that remain in branches, now that routine banking has largely migrated to digital and self-service channels, are disproportionately the ones customers couldn't resolve elsewhere — disputes, complex product decisions, situations requiring judgment rather than a standard workflow. Treating these visits as equivalent to a routine deposit, cost-wise, dramatically understates their actual value to the customer relationship.
A more accurate economic model would weight branch visits by the complexity of the issue resolved, not just the cost of resolving it — recognising that a branch visit which prevents a customer from leaving the institution entirely has a value that a simple transaction-cost calculation will never capture.
Institutions applying this weighted model to their closure decisions have, in cases I've reviewed, found individual branches that looked expensive under standard cost-per-transaction and looked clearly worth retaining once weighted for the genuine complexity of issues they were resolving — issues that would otherwise migrate to call centres poorly equipped to handle them with the same authority.
The economics of a branch visit in 2026 aren't what they were a decade ago, and institutions still evaluating branches with a decade-old metric are making capital allocation decisions based on a measurement that no longer reflects what branches are actually for.