[Product Roadmap] Started with 100 loans a month, how StashFin grew to process 150 loans an hour

A product roadmap clarifies the why, what, and how behind what a tech startup is building. This week, we feature the fintech platform StashFin and how it achieved 800 percent growth in users in four years.
153 CLAPS
0

In 2016, Stashfin was started by Shruti Aggarwal to make credit lending easy, transparent, safe, and secure for borrowers across India. The startup is servicing a segment of customers largely ignored by the banks by offering them a product that would befit premium customers of a bank. These could be students, unbanked, and first time credit consumers. Parikshit Chitalkar and Tushar Aggarwal are co-founders in StashFin.

“We are targetting a wide range of customers for which we need to build underwriting models to support them, and have operations that can service these customers very efficiently. This needs a lot of rapid evolutions as we don't really have another competitive product from which we can learn,” Shruti tells YourStory

On StashFin, personal loans are delivered either through a bank account or through an innovative visa prepaid card, with the interest charged only on the amount withdrawn, making it suitable for borrowers who wish to have financial flexibility at a lower cost.

“The nature of our business is very similar to what a bank would do. So imagine every single application or solution that a bank runs with thousands of people in tech is built and run by us with less than 40 people,” Shruti explains. 

The technology covers the entire lifecycle of the customers – starting from origination, onboarding, decision-making, loan management, repayments, dashboards, reconciliation, loan operations, and risk analytics. 

Shruti Aggarwal

Complete technical control 

“We are heavily dependent on Machine Learning and use it to optimise various aspects of the business – ranging from risk assessment and decision-making, to optimising operating efficiencies. The stack has helped us to scale from processing approximately 100 loans a month initially to now over 150 loans an hour,” she adds.

AWS recently featured StashFin as one of the best of breed fintech stacks in India after thoroughly reviewing the platform. In these four years, StashFin has shown close to 800 percent growth in users

Speaking about the importance of complete technical control, Shruti says, “For example, we must have server-side content delivery in our app so that if we decide to open up a low-income segment of customers in Odisha, we can quickly deploy a targeted page with different imagery and a different language localisation, all the while having an affluent customer from Gurgugram log in the same app.” 

Thus, Android/iOS releases have to be carefully planned wherein the updates have to be minimised by doing configuration downloads run-time as opposed to releasing new assets. 

One of StashFin’s flagship products was conceptualised and launched on the app in under five days. That product today does one loan every minute and brings in over 40,000 new customers each month. 

“Another example is that during the lockdown, our infrastructure team was able to set up the work from home (WFH) infrastructure in less than eight hours, so that every application that was available in the office was operating on home networks, including VOIP calling for our call centre executives. Our professional call centre partners took weeks to get online, having our in-house team ready overnight helped us a lot in collections as we were able to call customers as we normally did from the office,” explains Shruti. 

Building the MVP 

For the minimum viable product (MVP), the team wanted to focus on its time-to-market strategy get the product out fast, learn from feedback, and improve constantly. 

“We sometimes have done over 10 production releases in a day. The core aim of our engineering team is to develop a keen understanding of the business. All our developers understand the business in and out. They are co-located with business teams on the same floor, which automatically leads to the evolution of the product from the bottom-up,” says Shruti. 

The focus is on training the engineering team to be ‘Full Cycle Developers' – right from design, finding vendors, to sitting with the calling team and listening to customer phone calls.

Shruti explains that if the engineers hear the customer feedback first-hand, the product turns out better. She says the team has removed the layers in between so that reaction times are faster and the end result is a better product for the customer. 

The tech is built in-house. Shruti says that this made the hiring process hard, as they were looking for engineers who had good development skills but were also keen learners at the same time. 

“We are very big on monitoring; we want to know what is failing, where, and why. That is the main source of our learnings, then we train very heavily in-house, do lots of POC’s on new technologies and when something sticks, we acquire that skill and put it in production,” she adds. 

Keeping it simple 

StashFin is built on cloud stack, thus, she says scaling the infrastructure was less challenging. The DevOps and infrastructure team built tons of automation to enable that so engineers could focus on the product. Everything from CI-CD Pipelines, to testing, to quality reviews and code merges is fully automated.

“We are in a continuous improvement mode – whether it is product features, infrastructure improvements or better engineering, we are always doing new things. We are very conservative about putting experimental things into production but we experiment a lot,” says Shruti. 

The consumer apps were launched within three months of their inception. Around 7,000-10,000 users log into it at any time of the day, and the apps have amassed more than 20 million downloads. On Google Play, StashFin has a rating of 4.9 stars. 

“Each engineer has alerts set on their code in their modules that let them know how things are working. This way, we get to know how people are using the product. We have found many instances where users were doing something entirely different than what we had envisioned and we fixed it, then re-measured to see the gains,” she explains.

Following user feedback 

Adding a vernacular language feature is another example. The team saw that users from Tier III cities were taking more time to make payments than Tier I cities, so they added a language option only on that one page and saw a huge uptick. 

“We put the onus on the team. We have two sessions a week for learning and development where people showcase what they have learnt and tried over the week. We use these sessions to jointly review the code and improve it, make the data model more efficient, optimise queries etc,” adds Shruti. 

She adds that this way, they have over 30 people thinking about the same problem in different ways. They also invite business users to these sessions and who often give them great feedback. 

“We need to innovate to remain market leaders and at the same time, we are in the business of making money, so running a stable system is absolutely essential. This presents a unique challenge for us. It's harder for us to make large-scale changes very rapidly, which can risk stability. The way we work around this challenge is to automate, monitor, and alert heavily. That way, we increase the team's trust in the system so they can focus on innovating while the assurance of stability is managed by the machine,” says Shruti. 

Future plans 

The legacy belief that code can fix everything does not apply to StashFin. The team now believes that less code is better. Shruti adds they use managed services and built integrations around them to do the job. This significantly increases the stability and the team has an assurance of performance at scale.

Speaking of the future roadmap for the product, Shruti says, 

“We are working on multiple feature additions to our product. We are working on a lot of integrations to support new partnerships. We have a strong roadmap of what needs to be developed by our Machine Learning team, where all of our models will become fully self-reinforcing. We also constantly improve our fraud detection, where we are starting to experiment with NLP to go through customer interactions with our executives in voice calls/chat, and figure out patterns to improve customer service and weed out the bad actors.”

Edited by Kanishk Singh

Latest

Updates from around the world