DBS Bank

Objective

Increase the economic value drawn from each retail customer by pushing, in the app, hyperpersonalized nudges at the right moment: product recommendation, duplicate-payment alert, favorable exchange rate for a transfer, bill reminder. The goal is to get customers to save, invest, and insure more while generating incremental revenue, with value measured by comparing exposed customers against a control group.

The deployment

DBS pushes personalized messages to its retail customers directly inside its banking app, what it calls nudges. Each nudge is triggered by machine learning models that read the customer's behavior: they spot an upcoming duplicate payment, an advantageous exchange rate for a transfer, a bill nearing its due date, or suggest a product suited to the profile. The engine relies on more than 100 AI/ML algorithms that analyze an internal data mart of 15,000 data points per customer to produce seven types of nudges. At group scale, DBS reports more than 600 AI/ML models and about 300 use cases in production. The volumes are public: the retail bank (Consumer Banking Group) sends 45 million hyperpersonalized nudges per month to over 5 million customers, and the group as a whole exceeds 1.2 billion nudges per year to over 13 million customers in the region. DBS does not simply measure usage: the economic value of each case is calculated by comparing exposed customers against a control group. In 2022, these AI/ML use cases produced SGD 180 million in economic value, including SGD 150 million in revenue uplift and SGD 30 million in cost savings and productivity gains. Cumulative value reached SGD 370 million at the end of 2023, then over SGD 750 million for 2024 alone, and the bank claims a record SGD 1 billion in 2025. In Singapore, customers who use the nudges and the AI financial planning tool save twice as much, invest five times as much, and are nearly three times as insured as non-users.

Results Proof B

Volumes and economic value are documented by DBS primary sources (2024 annual report for the 1.2 billion nudges, 13 million customers, and SGD 750 million in 2024; official AI page for the 600 models, 45 million monthly nudges, and the SGD 180 million from 2022 with breakdown) and are consistent with a Forrester analysis (record of SGD 1 billion in 2025, SGD 370 million cumulative at the end of 2023). These are not audited financial result lines but an economic-value metric calculated by the bank, measured in A/B testing against a control group. The caveat that keeps it at B rather than A: the SGD 1 billion aggregate blends revenue uplift, cost savings, and risk avoided, so the strictly marketing share cannot be isolated; only 2022 gives the breakdown (SGD 150 million in revenue uplift out of SGD 180 million). Multiple, consistent official primary sources justify the top of the B range.

1. 1
   Building the data mart data team

   Customer data is aggregated in an internal data mart of 15,000 data points, the basis of the models.

2. 2
   Nudge detection and generation AI

   More than 100 AI/ML algorithms analyze the data mart and generate seven types of nudges (product recommendation, duplicate alert, favorable exchange rate, bill reminder, life stage).

3. 3
   Delivery in the app AI

   The relevant nudge is served to the customer in the banking app, at the moment it is actionable.

4. 4
   Incremental measurement data team

   The outcome is compared against a control group to quantify the economic value truly attributable to AI (revenue, cost avoided, risk avoided).

5. 5
   Consolidation and steering marketing

   Economic values are aggregated and tracked over time (SGD 180M in 2022, SGD 370M cumulative at end of 2023, SGD 750M+ in 2024, SGD 1B in 2025) and feed the prioritization of use cases.