‘Big data’ from emerging-market customers might not instantly seem rich with potential—aren’t lower-income customers poor sources of data? However, in an excellent presentation from Maha Khan and Annabel Schiff during the 2018 #DFS4What event, participants learned of the many ways in which fintech firms are starting to use big data in exciting ways.

As mobile phones become increasingly endemic, with movement towards smartphone and Internet-enabled feature phones, the information available on users is only growing. Primarily, these big data are used for segmentation; if nothing else, having more data facilitates better design of financial products based on better segmentation of customer bases. Once a financial product is digital, what was often only achievable via expensive on-the-ground surveys can now be discovered via product usage.

Data sources have chiefly been mobile network operators, with data mainly being used to drive credit scoring, although innovative insurance, credit and advice products from companies such as Apollo and Pula have used a combination of data from mobile phones and satellite earth observation to create new insurance risk models as well. In these more innovative use cases, the challenges are often around sourcing the data—particularly from new satellite providers, where the solutions are not always ready for business-to-business customers. The use of credit scoring data from mobile network operators for loan services is more established, though still not always a perfect process. Mobile network operators effectively act as a partner between fintechs and customers, but data protection legislation is starting to make such relationships problematic. Regulation varies wildly from country to country, as companies and governments across the globe try to navigate what is permissible and how to protect consumer privacy in a digital age.

Looking to China, Maha and Annabel shared insights from a recent trip organized by the Partnership for Finance in a Digital Africa. There are fantastic glimpses of what is possible from the Chinese experience for big data and financial products, as well as some concerns about potential problems. The wealth of data that Chinese providers such as WeBank and JD Finance collect from consumers gives them fantastic abilities. A mixture of artificial intelligence and these vast data sets allows loan providers to approve applications within seconds, based on very small amounts of input from users. Such services are also emerging in India, where providers such as Cashe use big data to approve credit scoring. However, there have been recent reports on the social credit scoring system in China, which uses big data for more government surveillance and control. Advances in the area of big data within China have not occurred without negatives emerging, and the challenge in learning from this market is going to be how to extract the positive lessons from their phenomenal experience.

Maha and Annabel’s presentation then turned to the types of data that can be available to fintechs. Earth observation data have huge potential, particularly in reaching rural populations and understanding conditions on the ground, and can track dynamic data such as weather and soil conditions in near-live time. As more and more providers enter the booming earth observation data market, and as providers such as Radiant emerge to specifically meet the needs of developing markets, getting the data is often not the problem but analysing it still is. The eternal problem of big data, it seems, is not accessing data sets but establishing how to pull insights from them.

Zooming down from a 40,000-foot view to the personal, the presentation next examined psychometric analysis and conversational interfaces. Both of these systems aim to use large data sets to better understand the human. For psychometric analysis, that means using large bodies of data to better assess customers and establish credit scoring or segmentation for them. For conversational interfaces, the challenge is to be able to understand natural language communication, from voice or messaging/chat services, and then respond as if the systems were a person. In both of these cases, cultural bias and sensitivity are paramount, as understanding what the data can show depends on a deep understanding of the local culture and user behaviour. For natural language processing, as an example, ‘code-switching’ between English and Swahili mid-sentence is common in spoken and written online conversations. Understanding how systems can cope with these particular problems will be the next challenge to implementing them.

Much is promised from big data, and this session provided an excellent opportunity to introduce many of the key themes. If data is the new oil, it is clear that turning on the tap is not the problem. From users becoming ever more digital in their everyday lives, to satellites and sensors providing ever more detail on the world, it is unlikely that ‘peak data’ will ever be reached. But the issues around successfully analysing the data, navigating unstable regulation around consumer privacy, and understanding how these data help build new and efficient fintech products, will need to be overcome.

December 2018. Copyright © UN Capital Development Fund. All rights reserved.

The views expressed in this publication are those of the author(s) and do not necessarily represent the views of UNCDF, the United Nations or any of its affiliated organizations or its Member States.

Mention of any firm or licensed process does not imply endorsement by the United Nations.