Test an outlet closure on pension day, not on an average Tuesday

Every bank with a large rural network will close outlets in the next decade, and it should. Transactions are moving to phones, cash use is falling and a merged bank carries duplicate counters in the same towns. I have closed thirty-eight outlets since our merger and expect to close more. My argument is about how the decision is tested. A closure justified by average footfall and branch profit is tested against the customers who need the outlet least. It should be tested against the day the outlet is needed most, by the customers with the fewest alternatives. Those are different questions, and in a rural prefecture they produce different closure lists.

What the average day hides

When Xuantang Rural Commercial Bank was formed in November 2023 from eleven predecessor institutions, it inherited 356 outlets across a prefecture of 4.7 million people. In forty-one towns two predecessors had built outlets within two kilometres of each other, often on the same street, because each had grown up serving its own county and had never needed to think about the other. Twenty-three townships had a single outlet that lost money every year. The merger plan proposed closing sixty-one outlets for an annual saving of RMB 74 million, and on the figures it presented, every one of them looked reasonable.

The figures were averages. A township outlet might see eighty customers on an ordinary weekday, most of them younger people who could have used the app. On the days when pensions and farm subsidies arrived, the same outlet served four or five hundred, many of them over seventy, who withdrew cash, updated passbooks and asked a clerk they had known for years whether a text message they had received was genuine. That last question matters more than it sounds, and it is rarely asked of a stranger behind a glass screen. Our counter staff intervened in 1,240 attempted scam transfers by older customers in 2025, and most of those conversations happened on busy days between people who recognised each other.

An average also hides the journey. Two kilometres is nothing in a county town with taxis. In a mountain township it can mean an hour on a bus that runs twice a morning, and for a seventy-eight-year-old carrying her passbook and identity card it can mean not going at all. When a bank closes such an outlet on the strength of average footfall, the cost does not vanish. It moves to the people least able to object.

Three tests before a closure paper reaches the board

I proposed, and our president accepted, that no closure would go to the board unless the outlet passed three tests. The first is proximity: another of our outlets within two kilometres, or a mobile service vehicle on a published weekly route. The second is capacity: a timed trial at the receiving outlet on a real pension day, with the closing outlet's customers simulated by staff, and a median queue below forty-five minutes. The third is access: nine in ten customers over sixty-five able to reach the receiving outlet within forty minutes by public bus, checked against actual timetables rather than a map.

The pension-day trial is the test that changed most decisions. Several receiving outlets that looked spacious on an ordinary day could not absorb a neighbour's busiest morning without queues of two hours or more. Some of those closures went ahead after we added a counter or changed staff rosters. Others did not go ahead at all.

Our chief financial officer objected, and his objection was serious. The merger's financial case had been approved on the full saving. Loss-making outlets that survived the merger would, he argued, become harder to close each year, because every year added to the case for keeping them. His memo went to the board with my paper. The board accepted the tests with a condition I agreed with: every retained loss-making outlet is reviewed annually against the same three tests, so that keeping it is a decision taken again each year rather than a habit.

What we closed, what we kept and what it cost

Thirty-eight outlets passed the tests and closed between mid-2024 and mid-2026. We kept twelve loss-making single outlets and moved eleven others to three days a week. We gave up about RMB 31 million of the annual saving in the plan. We bought nine mobile service vehicles, held meetings with village committees and wrote to every affected account holder ninety days ahead. Staff from each closing outlet moved with their customers to the receiving outlet for six months, so that the face at the counter stayed familiar through the change.

A year after closure, deposits held by the closed outlets' customers stood at 96% of their earlier level, and the median pension-day queue at receiving outlets was twenty-six minutes. Complaints about closures fell from 214 in the first year to 61 in the second.

The tests did not prevent every mistake. We closed a mountain-town outlet in January 2025, two weeks before the Spring Festival, when pensions, wages from returning migrant workers and festival withdrawals all arrived together. On the first pension day the queue at the receiving outlet passed two hours, in the cold. We added a temporary desk within the week. The tests had been run in autumn. We now run them in the season of the planned closure and do not close outlets between December and the end of the festival.

I do not think a bank owes every township a permanent counter. I think it owes its oldest customers a decision tested on their busiest day rather than its quietest one, and a replacement service that is already running when the doors close. A board reviewing a closure programme can ask for exactly that: the pension-day queue at each receiving outlet, the bus time for customers over sixty-five and the date the replacement service began. If those figures are missing, the paper describes the bank's saving but not its customers.