Africa’s largest financial institution, Standard Bank, is on a mission: to find ways to set itself apart so that the bank can meet its customer needs in today’s complex, ever-changing world. Operating in more than 20 countries in sub-Saharan Africa and globally, the Bank aims to show customers that It Can Be.
This means that it needs the right tools, resources and platforms to enable the innovation, resilience and efficiencies required to empower its customers. For Standard Bank, that means using Microsoft’s Power BI platform to harness data insights and predictive and prescriptive analytics through AI and machine learning. This allows the bank to better serve customers through more accurate targeting and service delivery.
In 2017, the organisation moved from a traditional centralised business intelligence team to a self-service BI model because its value-added capabilities allowed more accurate and effective targeting and serving of customers. “Power BI helps make citizen data scientists out of anyone who wishes to be one,” says Ziyaad Valli, Lead Analytics Engineer on the BI Data Visualization team at Standard Bank.
By 2020, adoption of the company’s self-service BI was such a success that Standard Bank’s Microsoft Teams channel for Power BI included more than 3,000 members.
The bank’s data visualization team was using a descriptive analytics model to evaluate what had already happened. But Valli wanted to explore the predictive (what might happen) and prescriptive (what are the possibilities) paths to add to the toolsets for internal teams. As a result, the data visualization team collaborated with the bank’s Insurance Business group to use AI and automated machine learning to predict and accurately target customers.
Using data to tell a more accurate story
In July 2020, Standard Bank Insurance launched a funeral cover product – insurance to cover funeral expenses – that sold well: customers were queuing at branches and phoning the bank’s call centres. Mohammed Tootla, Manager of Data Visualization for the Insurance Business group at Standard Bank, decided they needed more from their data.
“We started looking at alternatives to tell a better story with our data,” says Tootla. “We wanted to explore machine learning, but we had no data scientists on our team.”
Tootla then contacted Valli, explaining that his team wanted to move forward with predictive and prescriptive analytics. Valli’s and Tootla’s teams began collaborating in Power BI. They aggregated and analysed sales for the funeral cover across multiple dimensions, discovering a surprising insight that aligned with feedback from business leads: millennial customers were creating higher demand than any other generation.
The other key learning was that sales were happening on a particular day of the week. According to marketing and industry standards, most sales are expected to happen on Wednesdays and Thursdays. However, data showed that funeral cover sales were happening on Fridays.
“We were able to uncover these patterns and key insights solely because of the AI analytics capabilities within Power BI,” says Tootla.
Understanding these trends in the data allowed Standard Bank to make proactive adjustments to target the right audience and prepare for sales accordingly.
Better serving customers and mitigating customer cancellations
Next, Valli and Tootla delved into the forecasting models to compare what sales might look like if factors remained unchanged, with “what-if” parameters – adjusted policies forecast – such as: what happens if South Africa returns to a hard lockdown because of COVID-19 and movement is restricted?
They created hindcast data models for four time periods to compare predictions to actual results and presented them to business leads, who determined the hindcasts accurately stayed within the upper and lower thresholds of the actual numbers.
Because of the forecasting model, the leads now use predictive and prescriptive insights to more properly staff employees in branches and call centres to better serve customers.
Tootla’s team also had another objective: to analyse the high churn rate in his group’s short-term insurance book. The previous report already utilised descriptive analytics models, but these only offered insights as to why a customer had cancelled insurance in the past. They didn’t indicate what might happen in the future.
Valli and Tootla built a binary classification model using the automated machine learning (AutoML) feature in Power BI, which quickly yielded value in affecting the bank’s bottom line.
Using the churn model, 88,000 customers were evaluated—40,000 active and 48,000 previously cancelled. The churn model predicted the risk of losing approximately 8,000 active customers and broke down the 8,000 customers by segment, policy persistency, age, province, and income, so the insurance group could target them when running retention campaigns.
“We were highly impressed with the fast turnaround time with which the Power BI AutoML led to outputs,” says Tootla. “A data scientist would come in and might not produce a report for months. We produced the first binary classification model within three weeks.”
Uniting datasets and adding value
Valli and Tootla agree that the project is still in its infancy. Tootla envisions expanding the forecasting and churn models to the full Standard Bank insurance product line. Beyond that, he foresees uniting the sales, cancellations, and retentions datasets.
“We’d like to bring in customer demographic information and build an all-in-one dataset model we can use as an input into our customer lifetime value model,” he says.
The collaboration and outcomes have altered how Valli contemplates data by assessing any new features broadly and contemplating how they can add value to any business unit his team supports. He says AI analytics is about helping the bank extract value and demonstrating the robust capabilities of the Power BI tools.
“This is only the start. We’re going to be doing a lot more going forward and we’re all really excited about that,” says Valli.