Access to capital in Indonesia for MSMEs is notoriously poor, as banks continue to use outdated collateral schemes such as fixed assets, which most MSMEs do not have. However MSMEs often have strong revenue streams, and would be able to pay back small loans.
What it does
Moksha is a marketplace that pairs MSMEs with retail investors in order to provide flexible micro-financing Income Share Agreements. Investors benefit from metrics provided to them through the platform, including formulas for assessing potential profitability as well as risk probabilities. These metrics were developed by training machine learning algorithms, and are aimed at helping investors make better-informed decisions on their investments.
How we built it
Machine learning with PyTorch alongside a frontend consisting of React, NextJS, and Tailwind CSS.
We used a NextJS app with TailwindCSS for the first time and that was super fun! We loved that Next had native server side rendering and Tailwind gave us the flexibility to code out our prototype faster than we'd be able to otherwise.
Challenges we ran into
Defining success and profitability metrics. This was the most challenging part. We researched consolidated metrics that we could use to evaluate business profitability. The ones we settled on were free cash flow, customer churn and capital efficiency. These metrics were cited in several texts in the theory of value investing, which we thought was the best school of thought to model our risk metrics off of. Most of our businesses and investors would be risk averse and are not looking to 10x or 100x their investments, so we wanted to give them metrics that would allow them to assess whether the business they were going to invest in was risky business or not.
Finding datasets pertaining to loans in Indonesia and Southeast Asia was difficult, which hindered the development of our risk model. This was super challenging because Indonesia is still a developing economy and does not publish data at the same scale as that of developed economies. We instead used an American loaning dataset, but we're confident that we can replace this with a more relevant dataset in the future.
Accomplishments that we're proud of
Building a full stack web app, learning tailwind css, and building a platform that will help Indonesian small businesses get the funding they currently lack.
What's next for Moksha
Finishing the web app, Improving ML algorithms, onboarding initial investors, and finally expanding beyond Indonesia.