By Dwijo Goswami
How ‘big’ is big data? This question is often confronted with a flurry of statistics. One out of every 3 people on earth has access to mobile internet. One out of every 5 has an active Facebook account. By 2020, the amount of data generated by the world will be 44 times the data generated in 2009. My personal favorite: every 30 minutes humans generate data equivalent to all written works from the beginning of recorded history. This one definitely feels true every time I plug back into my family WhatsApp groups! Clearly, there is no debate about the volume of data we’re generating. The current debate about how ‘big’ big data can get is centered on how well it can be used to predict future behavior.
The history of big data
Well before today’s internet giants, Walmart was one of the first large corporations to understand the value of data. As early as the 1990s, Walmart began applying algorithms to its large data network of retail outlets, customers, and suppliers. Walmart went several steps ahead of better inventory management – they optimized the economics of their entire supply chain. They began anticipating consumer demand and started beating the competition by getting earlier and better deals from suppliers.
Today, such predictive analytics is par for the course for online retail players like Amazon and Alibaba. Amazon is continuously pushing the boundaries for predicting consumer demand. In 2013, Amazon filed a patent to ship deliveries in anticipation of an order. In other words, Amazon knows someone like me in the Greater Boston area has a regular hankering for muesli. The system ships the muesli out towards Boston, and simply updates the specific delivery address en route, when I actually place my order. This may sound crazy but in a saner world, everybody would be eating bircher muesli all the time.
Big data in fintech
For Accion and the larger world of financial service providers, like banks, credit unions, and microfinance institutions, big data’s predictive power can open doors we have never walked through before. Armed with the right data and algorithms, financial service providers could assess clients in terms of their future potential, rather than their past behavior.
Conceptually, this is a game changer. Traditional credit scores are calculated on the basis of previous credit re-payment behavior. A forward-looking, data-driven approach could be groundbreaking in markets where otherwise credit-worthy individuals do not have any traditional credit history. The Omidyar Network recently estimated that in the six largest emerging economies alone, leveraging big data may provide over half a billion individuals access to formal credit markets for the first time in their lives.
Weighing risk vs return
Despite this potentially large payoff, there are many challenges to scaling predictive power. The biggest current challenge is pairing the strongest predictive algorithms with the right data sets. The most intuitive place for leveraging big data for financial inclusion would be traditional banks with rich historical data. However, banks are not designed to be agile, data-mining institutions. Much of their internal data remains scattered and inaccessible. This is a challenge that must be surmounted, as banks play a critical role in financial inclusion — according to a recent study by CFI and IIF, 90 percent of the accounts opened between 2011 and 2014 were opened at a traditional financial institution.
Fortunately, many players innovating with credit scoring algorithms — including Accion’s partners Aire, Revolution Credit, and First Access — are positioning themselves as partner organizations to traditional financial institutions. However, not all banks have the risk appetite to jump at customers with poor or no traditional credit scores. As a result, many banks are cautiously waiting for newer models to prove their business value.
This conservative approach is not completely unjustified. This brings us to the second big challenge facing the financial inclusion industry — all of the big data players suffer from what is called a ‘recency’ bias. Most big data algorithms have only been in play post-2008. Traditional credit scoring, on the other hand, has withstood the test of time and matured through many credit cycles.
The profitability of big data analytics hinges on its ability to predict consumer behavior. Knowing the likelihood to default, for example, allows financial institutions to correctly price loans for non-traditional clients. In a changing macro environment, such as a credit boom and bust cycle, these predictions may not remain stable. As these models scale and become more prominent, they will also attract a wider range of clientele with a different risk profile than the clients the algorithms were initially tested on. These are the kind of challenges currently being faced by Lending Club and Prosper, who have grown to a certain scale, and are now seeing rising default rates.
These challenges together pose the big risk, big return conundrum holding back the data revolution for financial inclusion. What is at stake is an opportunity to finally build a sustainable financial ecosystem for the 2 billion individuals who are currently unbanked.
For Accion, this return is certainly worth the risk. We expect to see many of the models mature, and yes, there will be some hard lessons learned along the way. That is the kind of journey we have had with several decades of work in traditional microfinance, and believe we will see an equally rewarding journey ahead of us in the world of big data. As leading financial inclusion players have highlighted, adapting to change and technology is one of the biggest risks faced by the microfinance industry. Accion currently occupies a unique position between the worlds of traditional microfinance and innovative fintech. We intend to continue leveraging this position to transfer knowledge within the industry to further the global agenda of financial inclusion.