BRIEF Putting Gig Data to Work: Innovations in Expanding Credit Access Malika Anand and Gayatri Murthy Summary Gig platforms can use work and earnings data to provide meaningful financial services to underserved gig workers. This brief highlights the experience of two industry pioneers, Moove and Karmalife, that have painstakingly built algorithms, designed products, and implemented pilots to prove the value of work data to extend transformative credit to workers. Their efforts are still in the early stages, and there is a need for more experimentation and innovative problem-solving in this space. Putting Gig Data to Work: Innovations in Expanding Credit Access CGAP BRIEF 1 I nformal workers are excluded from formal financial that psychological data can improve credit scoring, while services because of the lack of financial data about many others use asset ownership, education levels, bill their earnings and transactions. Without such data, payments, and more to augment earnings data. Although banks cannot adequately assess the creditworthiness of this additional data may help, concerns about consumer such workers, leaving them without access to affordable, protection norms and user experience concerns may appropriate credit. The financial inclusion community posits prevent platforms from collecting it in some markets.  that such work data will likely reveal many informal workers Another barrier is that work data from any single platform to be reliable, consistent, high-quality potential users of is likely only to reflect a portion of that worker’s work and financial services, increasing their attractiveness to lenders. earnings. In most urban centers, workers take jobs on Unlike informal work, platform work creates a rich data multiple platforms, so the true picture of their work and trail about the habits and incomes of workers. Since work earning is spread across platforms. Currently, platforms is tracked and paid digitally, platforms can offer financial tend to guard that data closely, and consumer protection service providers more information about their workers. norms prevent providers from aggregating across sources. Together, habits and earning data make platform workers A few innovators, however, are finding ways to secure worker an excellent opportunity to explore the value of work data consent to access data on their phones to assess earnings in extending financial services. across platforms. As these players scale or as open data norms take form, they may solve the completeness problem CGAP’s work over the past year suggests that a few with regard to the quality of work data.  organizations are finding ways to leverage platform work data to unlock financial services. In this brief, we highlight Still, even when earnings data is complete, it may not be the experiences of two of these pioneers: Moove and very valuable in revealing creditworthiness. Instead of Karmalife. These organizations have painstakingly built reflecting the worker’s quality or commitment, earnings algorithms, innovated data streams, and implemented volatility may be driven by seasonality and price changes pilots to prove the value of work data. Though their efforts on the platform. Moreover, earnings data alone might be are still in early stages, their progress is promising and misleading. For example, higher earnings may not always points to the underlying value of platform work data. As be better if workers burn out or deal with too much debt. In this work matures, we hope that more and better data will these instances, rating data such as star ratings from riders continue to push the boundaries of inclusion and expand may be more indicative of worker quality, but those ratings the range of financial products available to workers.  are often very compressed since platforms tend to boot low-performing workers off the platform or allocate them a few rides. Quantity and quality of work data Finally, past earnings may not be predictive of future Leveraging platform work data to expand access to credit earnings for gig workers. A key advantage of platform and is not straightforward. To start, platforms collect very few gig work is that it is flexible; it is designed for workers to data fields because they want to keep the onboarding be able to work as and when they like, so it should not be process fast and lean and because their systems were not a surprise if workers exercise that flexibility. Churn is more designed for machine-learning or credit-scoring exercises, or less baked into the design of platform work, eroding the just as operational management tools. Platforms collect value of that work data for assessing creditworthiness.  minimal information when onboarding riders or drivers: usually just name, date of birth, contact information, and Still, despite these challenges, both Karmalife and Moove driver’s license. have found innovative ways to leverage platform work data to unlock increased access to credit for asset building, a Moreover, platforms have yet to see the value of collecting long and unwinnable battle in increasing access to credit more data, creating a chicken-and-egg problem for for this segment. collecting additional data points that might aid in credit scoring. For example, innovators like Gigmile have found Putting Gig Data to Work: Innovations in Expanding Credit Access CGAP BRIEF 2 Karmalife: Leveraging Work Data Through Platforms We have seen a positive impact of “ Karmalife is an inclusive fintech startup in India that serves using work data as a primary source gig platform workers and broader pools of blue-collar of truth vis a vis other, underwriting workers with liquidity and savings solutions. Their engagement and research with workers over the years models that are there. And I think have confirmed that while workers benefit from early- that works at two levels. On one wage access, Karmalife’s flagship solution, those workers hand, we can be far more inclusive, also need a broader range of financial products to attain by a factor of four or five compared financial resilience. The mobility segment workers that Karmalife serves most commonly cite the need for higher- to, let’s say, a bureau score driven ticket, installment-linked loans, often to finance a vehicle or eligibility model. The second is the fuel purchases, both of which enable them to earn more ability to deduct our repayments at from ride-hailing work. source when there is a payout from Under the assumption that platform work data can be the platform, with the user’s consent, used to assess a worker’s creditworthiness, Karmalife analyzes platform data on rides, earnings, and other work making the repayment seamless and performance metrics to determine eligibility and then effective for our business model. And offers workers loans of between one and six months. Their together, there is a seamless and risk- experience shows that greater earnings, longer working informed solution for the worker that hours, and higher driver ratings are all associated with lower repayment risk. Though their pilot with CGAP is still is personalized to their earnings and underway as this is being written, early results suggest that work patterns.” platform work data could be as good as credit bureau data —Badal Malick, in predicting the creditworthiness of drivers who apply for Co-founder Karmalife, a fintech providing such loans. Furthermore, early results show that platform credit and other financial services to gig data can materially improve the predictive accuracy of workers in India customized credit models that also use bureau data. Initial results suggest longer-tenure loans improve driver engagement in the immediate weeks after getting the loan. be creditworthy. To date, 90% of loans to such workers Based on a cohort of 1500 platform workers that were extended by KarmaLife were repaid on time. eligible to borrow, 93% of workers who took out a loan were available to work the next week in comparison to the 85% of Scaling lending to this otherwise excluded segment holds workers who were eligible but did not take out a loan. The great promise, but fintechs and FSPs must pay attention analogous figures six weeks after loan eligibility were 95% to and mitigate the risks of losing money on bad loans or for borrowers and 89% for non-borrowers, suggesting that losing drivers’ trust in platforms. While fintech startups these loans are having the desired retention effect. might have higher risk tolerance, their debt partners may not. Or they could be constrained by limitations in Although experience to date is limited, most loans risk-sharing mechanisms. Building a robust business model disbursed by Karmalife using its platform-data-driven requires time to understand customer behavior, tweak scoring engine has benefited “thin-file” (or no-file) workers. risk models, and test out alternative pricing structures – Bringing loans to the previously excluded demonstrates requirements that can create tension between a platform’s the possibility for platform work data to drive inclusion. need for quick results and a fintech or FSPs’ need to learn It also adds some proof to the hypothesis long held in iteratively over time. the financial inclusion community that such workers can Putting Gig Data to Work: Innovations in Expanding Credit Access CGAP BRIEF 3 Platforms that can show patience and commitment stand automated communications as well as support to help to gain from such partnerships. While more research is them meet their repayment obligations, all while securing needed, tailored financial services may offer platforms a way the vehicle asset. For example, drivers who are unwell or to serve, engage and retain their best workers. Karmalife’s traveling away from their work area are invited to park their experience shows that early-wage access can lead to vehicles at Moove parking lots during those periods, during greater driver engagement, productivity, and retention. which time their repayment obligations are adjusted. Moove: Managing risk The different types of data we have “ with additional data access to – whether real time rides or Moove is an African-born mobility fintech, launched long-term productivity through the in Nigeria with a mission to unlock access to financial platform partners, or telematics on services for mobility gig workers globally. The team has developed a proprietary data collection, credit-scoring, and the vehicle and location data – helps risk management methodology to get ride-hailing drivers us drive much better understanding on a path to owning a critical work asset: their vehicles.  of customer productivity, Moove partners with ride-hailing and mobility platforms affordability and our revenue-based like Uber, Glovo, and Careem to identify drivers that financing model is really predicated meet performance criteria like trips taken, ratings and cancellation rates, much like those designed by Karmalife. on this.” Drivers who meet those criteria are invited to apply for —Tingting Peng, vehicle financing, including a driving test and background Chief Strategy Officer of Moove, a fintech checks. With these pieces in place, approved drivers are and logistics company, operational in several emerging markets focused on auto-financing. issued a brand-new vehicle, sourced by Moove itself. The driver then continues work on the platform, in a work-to-own arrangement. Like Karmalife, Moove deducts repayments at source directly from the platform, recovering Finally, Moove’s drive-to-own product also includes vehicle, the loan over the course of two to four years. During health and life insurance, repairs, service and maintenance. that time, Moove tracks and analyzes work data from This ensures that vehicles stay on the road, drivers the platform to set revenue and trip targets for drivers, continue earning, and the asset retains its value. Moove helping them to meet their repayment obligations while still makes this work by partnering with local service providers, maintaining a decent take-home income. These targets including vehicle manufacturers, at bulk rates and by only also provide value to platform partners, as they can better buying new cars in selected models that are most suitable meet revenue and supply driver availability goals.  for ride-hailing. This standardization allows them to manage Move goes further than most loan providers in managing costs both for the platform and for drivers themselves.   risk by collecting data about driver behavior via remote Moove has found that mobility marketplace platforms sensors such as vehicle telematics, allowing the company value having a supply-side partner that ensures a fleet of to observe driving behavior, speed or brake use, vehicle well-maintained vehicles and responsible drivers that are usage, geographic distances covered and other variables. available to meet the demand on their apps. Furthermore, Moove has developed proprietary data algorithms that Moove’s ability to provide additional financial services, allow them to use that telematic data to predict risk of including insurance, debit cards, and more in the future, repayment. Those predictions secure their assets by empowers drivers to start building their own businesses, allowing them to intervene when drivers might be at risk which has a powerful flywheel effect on driver supply. of non-payment. Those drivers receive personalized, Putting Gig Data to Work: Innovations in Expanding Credit Access CGAP BRIEF 4 Leveraging work data at scale vehicles so the potential returns are clear and may even be integrated into the structure of the loan, such as when While Karmalife and Moove are proving that platform data credit is provided via fuel vouchers. This may be less can unlock financial services for workers, their work is new, clear for platform workers who do not always have clear and the value of platform work data has yet to become an investment opportunities. Moove solves this by providing accepted belief in the sector. There are several reasons that financing only for vehicles and only supplies the vehicle, acceptance may be slow in coming unless the community not the cash, but intermediate ticket sizes may be difficult accelerates testing efforts. They include: to invest profitably. Solving for this “opportunity side” may • C  hicken-and-egg problem: Until platforms believe be important to unlocking larger sums.  that work and demographic data has value in driving To solve each of these challenges, regulatory authorities loyalty or generating revenue, they do not have any may need to intervene to improve the quality and access to incentive to collect it at the volume and quality needed data and accelerate greater access for platform workers. to test and prove that value. This creates a chicken-and- Our experience suggests that such progress may not egg problem that can potentially be solved via pilots happen without broader support from global stakeholders or learnings from adjacent product experiences. Even since neither governments nor platforms understand when platforms believe in the need to collect more data, opportunities latent in platform work. It may take they need guidance and assurance on how and what philanthropic or public intervention to bring together various to collect responsibly, so that data cannot be misused. platforms, fintech, and bank players in a way that builds, Finally, platforms need to better understand what data it rather than endangers, the financial health of workers. makes sense to share, and what constitutes value that More work can also be done to highlight and facilitate the they should keep internal to their enterprise.  work of innovators like Moove, Karmalife, and others that • L  ow risk tolerance: Understanding whether work data are finding new ways to provide these services.  can predict repayment behavior means offering loans to In the long-run, it may be useful to think about work data those who might otherwise be denied access based on much like we think about financial data and consolidate it bureau data. This would require offering loans to “higher alongside bureau data. Doing so would allow workers to risk” workers, a risk that most platforms and lenders are assemble their data in reliable, transparent, and coherent not willing to take. This low risk tolerance may also extend ways to unlock access to new and bigger products. to investing resources in understanding the connection Currently, that information is owned by platforms capturing between work data and repayment behavior – a learning a small sliver of earnings and financial data. More will need investment that may not come naturally to platforms.  to be done to make such data usable and reliable, a gap • C  omplexity of loan usage: For many lenders, especially that tech innovators may well fill. Open data regimes could as ticket size grows, what the money is used for becomes facilitate access to earnings information and would create a critical question. In the case of businesses or logistics mechanisms for people to control the use of their data. firms, that credit might go towards buying inventory or Acknowledgments The authors would like to thank Leena Datwani, Rani Deshpande, Xavier Faz and Elizabeth Kiamba for their research, analysis, and engagement, which contributed to the insights presented in these briefs. These briefs draw on five pilots supported by CGAP and a learning community of over 50 fintechs, platforms, support organizations, and financial service providers. We would like to recognize their generous engagement with us. We thank Dean Caire for lending his expertise on alternate credit scoring models to support the CGAP pilot with Karmalife. We are also grateful to CGAP colleagues Antonique Koning and Claudia McKay, for their substantial review and guidance on finalizing this work. Thanks also to Feven Getachew Asfaw for her editorial and production support. Putting Gig Data to Work: Innovations in Expanding Credit Access CGAP BRIEF 5