Vietnam lenders shift to behavioural data to judge loan applicants

They measure the sales, staff numbers, and inventory turnover in SMEs, for example.

A customer’s loan application can be influenced by the number of lattes you buy in a week, according to banks and credit institutions in Vietnam.

Financial institutions (FI) and credit modelling firms have pivoted to studying not just financial statements and receipts of their customers, but also considering financial behavior on a daily basis, Leos Gregor, Chief Risk Officer for Home Credit Vietnam, said in a panel discussion during the Asian Banking & Finance and Insurance Asia 2026 Summit - Vietnam.

Home Credit Vietnam processes about 800,000 applications a month, with an approval rate of 7 out of 10 customers.

For the new customers that go through the door, 30% to 40% are new to a bank—that is, they don’t have a prior record, Gregor said. This is where behavior becomes a good source of data for credit and approval decisions.

For small and medium enterprises (SMEs), behavior is based on the platform or means that the payment is made, and how the money from that transaction is used.

“So how is the SME collecting the sale, whether the sales really correspond with their VAT, whether the sales stay in the company expense, or whether the sales just pass through,” Nguyen Minh Anh, General Manager, SME Banking, Standard Chartered Vietnam, told attendees at the World Hotel Saigon, 31 March.

Hong Leong Bank, meanwhile, partnered with business micro platforms to access behavioral data, from sales to staff numbers and inventory turnover.

“Auditing finances are good for a certain purpose, but it's still a bit old, because auditing finances typically lags six months to a year. What we're looking at is real-time data. What happened yesterday? What happened last month? What happened last year?” noted Fred Lim, Chief Digital Commercialisation Officer at Hong Leong Bank.

“Generally, behavioral performance data of the merchants of SMEs are important. Alternate data in terms of spending, even data on telco, are also important variables that we look at,” Lim said.

All firms now say they use automation or artificial intelligence (AI) for underwriting. The challenge here is not on getting data, but the speed of decision, said Andy Nguyen Tran, Chief Financial Officer (CFO) at Zalopay.

“The model that the data scientists build, they may be very, very accurate, but it takes a lot of time to speed up the decision,” Tran said, noting that a model faced with tens and thousands of new users will take a while to give a yes or no answer.

Zalopay is exploring how to make near instant credit decisions. Tran said that it is about striking a balance between the model’s accuracy and the speed that a decision is made. If made to choose, accuracy is the choice to go, Tran said.

“We can be as accurate as possible. If that takes 30 minutes, then it's a trade-off that we're willing to take, or our partners are willing to take,” he said.

“So it's really a trial and error process, where you have to really push the model to thelimit in terms of speed, and then we can see what type of accuracy we're getting, and then feed into the feedback loop again,” Tran added.