Centralise the loan paperwork and leave the judgement in the county

After a merger, a rural bank finds itself with dozens of county lending teams doing the same work in slightly different ways, and the obvious response is to centralise. Much of that response is right. My argument is that it should be applied to the paperwork around a loan and not to the judgement inside it. For small businesses and farm households without a long credit history, the decision belongs with the county loan officer who knows the borrower, while the documents, contracts, registrations and disbursements belong in a central unit that can do them faster and more accurately than any county office.

Twenty-one days, of which four were judgement

In 2024 Xuantang Rural Commercial Bank had RMB 46 billion of inclusive loans to small businesses and farm households, about 186,000 borrowers and 1,150 loan officers across the prefecture's counties. The median loan took twenty-one working days from application to disbursement. Large banks had begun lending in our counties at lower rates, and a farmer who needed money for seed and fertiliser before spring planting would not wait three weeks for us.

My team traced 400 recent files from application to disbursement, recording where each one waited. About four of the twenty-one days were spent on the credit judgement itself: visiting the borrower, reading the household's or the business's accounts, talking to the village committee and writing the recommendation. The other seventeen went on asking for missing documents, re-keying the same data into several forms, routing approvals and, above all, registering collateral. Post-loan reviews found documentation defects in 31% of files. Loan officers told us they spent close to half their week on paperwork, which meant less time on the visits that are the real work of rural lending.

The diagnosis made the centralisation case for us, but it also showed where centralisation would stop helping. The four days of judgement were not the problem.

What moved to the centre and what stayed

In May 2024 we opened a central credit operations unit of 140 people, staffed largely by colleagues from the eleven back offices we were merging. It took over document checking, contract preparation, collateral registration and disbursement, and it now carries out the post-loan file review. The county loan officer remains the borrower's contact, makes the credit recommendation and stays responsible for the relationship. Credit authority still runs through an approval chain that ends with our chief risk officer. My team owns the paperwork and the clock; it does not own the decision.

Our head of inclusive finance wanted to go further. He proposed that loans of up to RMB 300,000 be approved centrally and automatically, using tax, utility and payment data, with county officers involved only by exception. For some borrowers that works well, and larger banks do it at scale. My objection was specific. Many of our borrowers are first-time or thin-file households: a family that grows tea and keeps pigs, a couple running a village shop, a man who repairs tractors for cash. Their tax records are slight, their utility bills say little and their creditworthiness lives in things a loan officer learns from visiting: the state of the orchard, whether the family pays the village shop on time, what the neighbours say. Village-wide credit assessment, which many rural banks use, depends on exactly that knowledge.

Our chief risk officer agreed with keeping the judgement local. We settled on a compromise I still think is right: automatic renewal for repeat borrowers with two clean years of history, and local judgement for everyone else.

By the end of 2025 the median turnaround was eight working days and the defect rate was 9%. Loan officers' paperwork time fell to about a quarter of their week. Inclusive loans grew to RMB 52 billion. The non-performing ratio on loans made in 2024 and 2025 was 1.6%, against 2.4% for the pre-merger book. I mention that figure with care. The new loans are young, rural credit losses arrive with bad harvests rather than on a schedule, and I would not claim the process caused the difference.

Where the line is hardest to hold

The line between paperwork and judgement is not as clean as an organisation chart makes it. Three cases come up repeatedly.

The first is a document that changes the credit view: a land contract that does not match what the borrower described, or a guarantor whose own loan is overdue. A central checker who finds it must send it back to the loan officer as a credit question rather than reject the file as incomplete. We wrote that rule down after several files bounced between the county and the centre for a week.

The second is pressure for speed. Once the central unit was fast, county teams began asking it to disburse before the credit recommendation was fully written, on the promise that it would follow. We refuse. A fast process that lets judgement trail behind the money has centralised the wrong thing.

The third is the loan officers' own standing. Three county branch presidents objected to the unit because it took staff and control over files away from them, and one told me that his officers now felt like salesmen. That concern is fair. We answered it partly by measuring officers on visit quality and early arrears rather than on volume, and partly by asking experienced officers to train the central checkers, so that knowledge flowed both ways.

We also made mistakes of sequence. We centralised contracts first because they were the easiest, when collateral registration was the longest wait and should have come first. And we staffed the unit for an average month, then met the March planting peak with a backlog of nine days.

None of this is an argument against data or automation in rural lending. It is an argument for being precise about what is being centralised. A merged bank should take every repetitive task away from its county lenders and do it well in one place. It should leave them the part of lending that depends on knowing the borrower, and it should defend that part when speed and scale make it look old-fashioned.