Mobile Lending in Africa – Part 1
How telco operators became mobile lenders
Where it started – the introduction of the mobile wallet
On November 8th, 2021, ITWeb published an article detailing the approval of MTN’s licence to operate their Mobile Money platform in Nigeria. With this article fresh in my mind, I thought it worthwhile to take a deeper look at mobile lending and consider what it means for the established financial services status quo in Africa and eventually many other parts of the world.
With the launch of M-PESA back in early 2007, Vodacom and Safaricom changed the way people across Africa interfaced with money. The mobile wallet platform has been phenomenally successful over the last 15 years, starting in Kenya and subsequently spreading across Africa and beyond.
Observing the success of M-PESA, other mobile phone service providers introduced their own mobile wallets into the market, including the likes of Airtel Money, Mobile Pay and T-Kash. We have now come to the point where mobile money transfers have rapidly become the way that most financial transactions are conducted in Africa.
Mobile wallets offer inclusion. To the large unbanked population in Africa, mobile wallets have become a straightforward way to store, send and receive money, and are accessible to anyone who owns a mobile phone, which is almost everyone.
Mobile penetration in Africa
Mobile phone penetration in Africa is high. According to a Business Day Ghana article from April 2017, there were 960 million mobile subscribers across Africa, which represents an 80 percent penetration rate. In the four and a half years since then, the penetration rate can only have increased further.
Unsurprisingly, following the rapid success of mobile money for storing, paying and receiving money, the mobile operators have subsequently moved further up the value chain and now offer a multitude of credit facilities (e.g., loans, overdrafts, and airtime advances) to customers using their mobile wallets.
Mobile lending serving the unbanked and the underbanked
This is ultimately where the mobile operators are threatening the traditional banks. A very high percentage of the African population is either unbanked or under-banked due to the prohibitive costs of traditional banking services, the high rates of extreme poverty, and the informal nature of employment for large sections of the population.
A sizeable portion of people are self-employed entrepreneurs, offering their skill sets and services to the local community. Banks have tried to enter this market, but the high delinquency rates and their large infrastructure costs have prevented them from doing so in a successful manner.
How do mobile providers assess risk for mobile lending?
The above conditions are anathema to the traditional, risk averse mainstream banks. Banks require their customers to show documented, steady, reliable incomes that can be used as the basis for their analytical models and scorecards. The old saying that ‘you can’t get credit until you’ve had credit’ came about because of the conservative nature of banks.
Banks want to be assured, with a great degree of certainty, that their customer will repay a loan before they are prepared to grant it. The primary source of data that banks use to make these credit decisions is the performance of the customer on other credit products (either with the bank, or with other credit providers, i.e. data obtained from the credit bureaux).
Mobile operators have had to look elsewhere for data points that could be used to predict repayment behaviour, as the majority of their customers are unbanked or under-banked. Fortunately for the telcos, they were sitting on a wealth of data within their own customer base that could be used to create new predictors of risk.
This telco data includes the performance of the pre-paid cellular account and the usage of the existing mobile wallet, which have proven time and time again as highly predictive in many developing countries.
Scorecard characteristics that mobile operators use to determine mobile lending?
Scorecard characteristics can be built around the length of time the customer has had a relationship with the service provider, the frequency and value of airtime purchases, phone usage (split between call and data), and other characteristics of the prepaid account.
On the mobile wallet side, the value and frequency of the payment made and received can give a good picture of the customer’s financial position.
A lending model that is fit for purpose
As prepaid phone and wallet data are relatively untested in some countries, mobile lenders still need to offer credit products that minimise the potential credit losses but still provide value to their customers.
Several factors have played into the mobile lenders hands to make this possible:
- They are able to use their own infrastructure for credit applications, disbursements, and payments. Operating costs can therefore be kept to a minimum.
- Due to the lack of operational overhead, they are able to profitably offer extremely low value credit products, which are still useful to their target market.
- The credit products issued by the mobile lenders do not require credit bureau enquiries in countries that have established credit bureaux, so a significant cost saving is achieved.
- As the mobile wallet is a transactional account, mobile providers are able to sweep funds that come into the account in order to settle past due loans, minimising the amount of reactive collections activity required. There is of course the risk that the customer will open an alternative mobile wallet account in order to avoid account sweeping, but this risk can be factored into the upfront customer credit assessment.
- As the credit products are often used to purchase pre-paid airtime, or are specifically sold as airtime advances, there is even less at risk for the mobile operators because the wholesale cost of servicing the airtime is considerably lower than its retail sales price.
Over time, as mobile operators gather performance information on their mobile product offerings, they are able to improve their predictive models with actual performance information. This results in stronger scoring models and strategies, enabling them to expand the credit offerings to increasingly higher value, longer term credit products.
Challenges faced by the mobile operators
The biggest challenge facing the mobile operators in creating and growing their credit offerings is in the streamlining of the processes to ensure cost effectiveness. Due to the extremely low value of the loans being offered, every step of the process needs to be as low cost as possible, and almost entirely automated. This is a very high volume, low transaction value business model.
Tools are required that can make the most of the data that is available, in order to continually refine and improve the automated decisions taken on every credit application as well as ongoing account management strategies.
Models need to be redeveloped on a regular basis, to take in the latest performance trends and new data that becomes available. Traditional scorecards which are built for stability as much as risk prediction are not the preferred option in this space.
Traditional scorecards vs machine learning models
Traditional scorecards have been replaced with machine learning models that can be developed quickly and with a high degree of automation and limited developer interaction. These machine learning models are replaced quickly as new data sources become available or predictive patterns change.
Although this method of model development has its risks (and would leave traditional credit risk managers like me in a cold sweat), it is well suited to the task at hand.
As I mentioned in my previous article, ‘Why is end-to-end decisioning so important?’ model development is only the first part of the puzzle. In order to successfully use the machine learning models to grant appropriate credit offerings, mobile lenders need a robust decision engine to automate the process.
Once a credit application is received by the decision engine it can calculate (or retrieve) the score from the latest model, and use it in the assessment of the application, and setting of the appropriate credit facility to offer to the customer.
The decision engine needs to have a broad range of functionality. It should provide simulation and champion/challenger functionality in order to rapidly deploy and evaluate new test strategies. This is a rapid learning environment, with customer performance available extremely quickly, as the majority of products are short term.
The ability to implement and track a considerable number of concurrent champion/challenger strategies will enable the mobile lender to refine and grow their understanding of the credit portfolios quickly and consistently. We refer to this process as ‘test and learn’ and it is never ending.
The decision engine must be able to handle an extremely high throughput in mobile lending. Although the facilities offered are low in value, the potential volume of applications is extremely high. This is still a high-risk environment, which means that decline rates will be quite high, and all of the applications still need to be processed, scored, and decisioned in an automated, low cost fashion.
Decision engine automation
Finally, decisioning needs to be entirely automated. The high volume of applications coupled with their low individual value, means that there is no capacity to manually review applications for a credit decision. As such, the decision engine must have access to all of the data that is to be used in the decisioning process. As the pool of data grows over time, the decision engine must be able to grow with it, enhancing the ability of the lender to rapidly react to changes in the highly competitive market, and expand their credit offerings.
With the data and tools at their disposal, and expertise growing in the unbanked and under-banked market, the future looks bright for mobile lenders. But what does all of this mean for the traditional banks? How can mainstream lenders react, and avoid losing entire lending portfolios to mobile platforms?
In the next article, I explore strategies and tactics that are available to the banks in order for them to operate in this unbanked and under-banked market. This will be a straightforward task, as mobile lenders have a significant head start, and have made great strides in building awareness and trust in the market. However, there are always ways to compete, and once again it will be led by data, analytics and decisioning!
About the Author
Jarrod McElhinney is a Client Solutions Manager at ADEPT Decisions (www.adeptdecisions.com).