Digitizing microbusinesses and informal work can help low-income workers and businesses access more and cheaper financial services, but only if the data created by digitization can improve risk evaluation by lenders. However, even as informal work and microenterprises are increasingly digitized, the value of this alternative data is still unclear as it has yet to radically transform access to larger-ticket loans for these segments. In general, the data currently being produced by platform models is not yet complete enough to unlock transformative credit for historically excluded small businesses and informal workers.
A happy ending to the digitization story?
The classic digitization story goes something like this: lenders are unable to accurately assess the credit worthiness of microenterprises and informal laborers since they lack assets and earn and save in cash without corresponding transaction records. As a result, these segments only qualify for small loans and remain underserved. However, digitizing their income flows creates data that lenders can use to disburse larger amounts at lower interest rates, as lenders’ risk would be lower.
The impact promised by this story is tantalizing, as the credit gap for small businesses is enormous: “The International Finance Corporation (IFC) estimates that 65 million firms, or 40% of formal micro, small and medium enterprises (MSMEs) in developing countries, have an unmet financing need of $5.2 trillion every year.”
However compelling the promise of digitization to unlock credit may be, its transformative impact largely remains a hypothesis as the value of the “alternative data” generated by platforms is not always clear. Although digitization has increasingly provided transaction records for businesses and workers in mature digital payments markets like India and Kenya, those records have yet to systematically unlock dramatically larger flows of credit for users.
However compelling the promise of digitization to unlock credit may be, its transformative impact largely remains a hypothesis as the value of the “alternative data” generated by platforms is not always clear.
Leveraging platform data to offer new credit models
Innovative fintech startups are starting to leverage platform data to offer credit to both microenterprises and informal workers. However, the amounts and tenure of those loans remain low. They are a percentage, rather than a multiple, of earnings and are generally designed to be repaid in their entirety in the next earnings cycle, which is often weekly or biweekly.
In the case of startups serving microenterprises, there are two dominant models – one that leverages digitized inventory orders and one that leverages digitized sales records. In the former, startups interface between small retailers and distributors/producers to take and facilitate inventory orders. They then use this order data to provide a rotating line of credit to allow merchants to stock more inventory. In these cases, a shopkeeper may be able to stock 20-30% more than the cash she has on-hand, an amount she repays in the next order cycle. In Africa, startups offering these models include Wasoko, Boost, Zanifu, Shopa, Copia, and others, and similar businesses exist in both Asia and Latin America.
In the case of models leveraging digitized sales records, e-commerce platforms track sales made by businesses online and offer them credit based on those income streams. As with models leveraging inventory orders, this model offers microenterprises a percentage of invoice value to facilitate stocking. Jumia and Alibaba, among others highlighted in a 2019 CGAP publication, have pioneered such models. Data on the size of such loans is hard to come by, but it may be that this model is able to unlock somewhat larger loans.
Among informal workers, startups serving the segment have similar mechanics but leverage data created by gig work platforms. In these models, startups track earnings and work routines of gig platform workers to offer credit-based advanced wage access. For example, in India, KarmaLife partners with gig work platforms to gain visibility into workers’ earnings and uses those records to allow workers to withdraw earnings before payday to purchase fuel, airtime and repairs so they can keep working even when they lack liquidity. The amounts are responsibly capped at a small percentage of earnings to ensure that workers retain substantial payouts at the end of the month.
In all three models, initial evidence suggests that the credit provided as a result of leveraging platform data is valuable. Zanifu reports that retailers using their service have been able to increase product turnover by 40%, which means more and more regular income for small merchants. 60 Decibels found that 82% of KarmaLife users reported quality of life improvements thanks to income digitization services.
In all three models, initial evidence suggests that the credit provided as a result of leveraging platform data is valuable.
However, in both cases the data created by digital transactions has yet to unlock substantially larger amounts or longer tenure credit. As noted, the credit amounts are typically a percentage, rather than a multiple, of recent earnings and are deducted from the next round of earnings. Our experience suggests that amounts are typically 20-50% of earnings for that period, meaning that it can help with cashflow and day-to-day management, but is unlikely to unlock substantial investment or meet lumpier expense needs.
The small, short-term amounts of credit workers and microenterprises are currently able to access help make their existing livelihoods more profitable, but it’s not the kind of transformative credit that allows the purchase of productive assets, or for families to avoid expensive credit during emergencies. Ride-hailing workers cannot purchase a car or motorcycle (since many rent), storekeepers cannot open new locations, and most still have to borrow to meet health expenses.. For example, a larger loan to purchase a car could help a ride-hailing driver substantially increase his income, as many drivers pay up to 45% of their daily income in car rent. In all, the livelihood impact of smaller credit is still extremely valuable to meeting day-to-day needs, but somewhat limited in achieving the broader goals of unlocking new livelihood opportunities and enabling social mobility for low-income workers.
Larger ticket loans are needed if credit is to achieve the promised transformation for low-income users
Limitations of platform data
These limitations may persist because the story outlined above overlooks fundamental characteristics of data created by platform models. Until the industry is able to address three core data challenges, digitization models are unlikely to unlock transformative financial services at scale:
Partial picture of work transactions:
Although gig platforms and MSE models may be ascendant, their users are almost always using multiple providers. For example, ride-hailing drivers in Kenya report driving for Bolt, Uber and Little simultaneously, just as drivers in the United States may work on Uber, Instacart, Lyft, Grubhub and Doordash at the same time. That means a fintech provider partnering with a platform is likely only seeing a small portion of workers’ data.
Similarly, a shop ordering via a startup is likely both purchasing from traditional distributors and going to the central market as well. This means that the fintech credit providers looking at online orders are only looking at a partial picture of earnings and transactions.
To be able to use data as hoped – to unlock access to substantially larger amounts of credit – providers need to see a complete picture of earnings and transactions. Among gig work models, some providers are addressing this gap by partnering with multiple platforms to see a more complete picture of work data, but platforms are unlikely to see the benefits of such models as many consider embedding financial services to be a competitive advantage. Similarly, it is hard to see how fintech providers could gain visibility into market purchases or why multiple inventory providers would have incentives to collaborate.
Automatic deductions preclude non-repayment data:
An advantage of both microenterprise and gig worker models is that repayments are automatically deducted. For platform workers, the amount withdrawn is deducted from their next paycheck. Similarly, credit to MSEs is automatically squared against the following week’s order and payment. While this is an advantage in terms of securing repayment, it does not help providers distinguish good credit risks from bad ones. When repayment is assured, providers have no data on who has not repaid, which is the best predictor of who will not repay in the future. Ashirul Amin, Managing Principal Consultant at BFA Global, who has worked on more than a dozen credit scoring algorithms notes that “most models show that previous delinquency levels are highly correlated with future ones.” Unlike other credit models, where small loans are used to assess creditworthiness, the automatic deduction aspect of these platform approaches has meant that past loans lack that value.
The automatic deductions have another downside in the case of gig work platforms, in that users may be less motivated to work on platforms where they know part of their earnings will be automatically deducted. In these instances, they may choose to work on other platforms, especially since they have multiple options to choose from.
Low data diversity:
A fundamental premise of data for credit scoring is that the data in question is useful in sorting good risks from bad. However, platform work data may not have that value because it is largely uniform around the characteristics we have seen to be related to repayment behavior: workers have consistent demographic profiles (younger men with a primary education), financial behaviors (a few loan apps, and a bank account), and similar earnings records (most work long days, seven days a week, at exactly the same rate). With so little diversity, the data may not be rich enough to sort workers into good and bad credit risks. Even where there is diversity in responses – like family size or type of phone – the fields often indicate little about creditworthiness for larger loans.
While there remain challenges with making platform data work for transformational livelihood impacts, we believe there is promise to the story. In particular, offering embedded finance to workers on gig work platforms is a remarkable opportunity to help an underserved population with few livelihood prospects improve their opportunities and resilience. That is why CGAP is partnering with five players in the platform space to test ways to offer transformational financial services for platform workers.
This blog is part of a broader CGAP effort over the coming months to work with platforms and financial services providers to pilot new financial solutions for this small but fast-growing segment of the economy that is transforming livelihoods opportunities for low-income communities. Visit www.cgap.org/platform-workers for more information and see our paper and reading deck for a deeper dive into the ideas introduced here.