Unleashing Five Times the Expected Results with Machine Learning Model and a User-First Approach: The Story Behind “PayPay Funding”

Welcome to the Project Story series, where we delve into the behind-the-scenes of PayPay group’s major projects. This series spotlights key projects, uncovering the challenges faced and the dedication of each team member through in-depth interviews.

On March 26, 2024, we launched a new service for merchants called “PayPay Funding” (PayPay Shikin Chotatsu in Japanese) which is PayPay’s first product leveraging a machine learning. The service has received an overwhelmingly positive response, with applications coming in at five times the anticipated rate. In this article, we spoke with eight key members who spearheaded this launch!

Driving New Business Releases with a Cross-Functional Team of Data Scientists and Business Members

PayPay Funding was launched for merchants on March 26th, 2024. Could you give an overview of the new service?

Watanabe:
PayPay Funding allows PayPay merchants to receive sales that they expect to make in the future, when they are in need of operating capital. The service is designed to be hassle-free, requiring no collateral or guarantees, and merchants can receive funds within minutes of applying, without the need for extensive paperwork. The repayments are automatically deducted from future sales, making it a seamless process.

Reference:『No Collateral or Guarantee Required! PayPay Launches PayPay Shikin Chotatsu, Enabling Invited Merchants to Receive Prospective Sales in Advance ~ AI determines possible funding amount from “PayPay” transaction data, provides funds to merchants in as little as a few minutes any time of the year ~ 』（Press Release, Mar 26, 2024）

Can you tell us about your roles and responsibilities in this project?

Watanabe:
Our Data Credit team, which led this project, is composed of business-side members who define service requirements, and data scientists who supply machine learning models. As the team and project manager, I oversee schedules, milestones, reporting, approvals, and problem-solving within the team.

Nishiyama:
Being on the business side of things, my focus was on legal requirements, developing and enhancing the framework for ongoing management and receivables management. The key was designing a user-friendly service while adhering to the future receivables transfer scheme (factoring). Despite numerous challenges at the outset, we successfully launched the service with support from the legal and compliance departments.

Sekine:
I handle the planning and executing of the go-to-market strategy, and formulate revenue and profit plans to manage metrics, aimed at expanding the use of PayPay Funding.

Tsugueda:
I joined this project from the Strategic Planning team, which is responsible for new business and service planning. My main tasks included defining business requirements and considering the UI/UX.

Nakao:
Out of the engineers in the team, I joined the company the earliest, at which time, not much progress had been made with the prediction model. I began with the basics; figuring out what to predict, what data was available, and how to handle it within our internal systems.

Qian:
When I joined, the basic design of the predictive models had been completed, and the next step was to improve their accuracy for implementation. My main focus was to enhance the models’ predictive capabilities and work on data integration and infrastructure specification.

Oba:
I primarily handled performance validation of the predictive models and also worked on monitoring processes post-deployment.

Minami:
I joined just before the launch of the service, focusing on interacting with the business team to ensure we could complete features without which the release wouldn’t be possible. Moving forward, I aim to continually improve customer experience through machine learning and data science.

Achieving Five Times the Expected Applications by Prioritizing User-First

Reflecting on the project, how do you feel?

Watanabe:
We kicked off the project on June 1, 2023, with nearly 100 participants, but at that time, the Data Credit team didn’t exist, and there were no dedicated members. The machine learning engineers who are here today hadn’t joined yet. On the other hand, the timeline had already been decided, making it a nerve-racking process to navigate.

Minami:
Traditionally, the funding process involves a manual review, but with our machine learning models, we can complete the review in seconds and immediately execute the funding. This is the key feature of the service. This also means the accuracy of the models is critical, and that’s where PayPay’s extensive data played a critical role.

Oba:
PayPay’s strength lies not only in its extensive merchant data but also in its user data. This allows us to leverage information that is not limited just to monthly GMV, but, for example, insights like stores with frequent customers or popularity among younger demographics. Having access to information from these two sources, helps in enhancing our predictive models.

What were the most challenging aspects?

Sekine:
My role was to create the service concept and business plan. Initially, I struggled to concretely envision how this service would benefit users. This left me feeling uneasy, as I was concerned that it might just end up being a pie-in-the-sky idea or fictional numbers on paper. However, in November, I had the opportunity to visit some merchants and hear directly from business owners about their challenges with existing funding options. This clarified the issues we wanted to solve and helped define our target persona. As a result, our discussions became much more meaningful, focusing on which parts of the product to enhance and how to sell it. I feel that this was a pivotal moment for the success of the project, as the whole team became aligned on who the service was intended for.

Watanabe:
Right from the beginning, I felt that it was a difficult service to explain. After Nakayama-san, the CEO, asked whether we had truly considered the needs of business owners and heard from them in person, we interviewed about 30 merchants in less than three weeks, even though the schedule was already extremely tight. In hind sight, this allowed us to maintain a merchant-centric approach throughout the development of the product.

Tsugueda:
We continued to polish the UI/UX until the very end, with the concept of enabling a ramen shop owner in the local shopping arcade to secure funding during a lunch break. Another memorable aspect was that we ultimately decided on the name of the service based on user interviews, instead of using the results of the internal poll, which included over a 100 options.

Minami:
As PayPay’s first machine learning product for external users, there were many pains and insights gained in the development process, as the engineer creating it. Unlike traditional waterfall development, machine learning model development involves iterative experimentation by data scientists. This creates a quality gap between experimental programs, which are usually disposed of once the experiment is done, and the production code that is ultimately provided for the customer. We aimed to smoothly connect R&D with production code deployment, but there were moments of confusion as we navigated this new territory.

Qian:
In my previous career, my role in projects often ended after building and analyzing models. However, this project required implementation for actual service operation, necessitating collaboration with engineers outside the Finance Business Strategy Department. For example, I worked closely with the infrastructure team to ensure stable and efficient model operation.

Nishiyama:
As a new business, there was a significant amount of back-office work, requiring collaboration with the System Division, the team responsible for developing internal systems, in addition to the Product Division. Managing development across three departments was challenging, but we managed to launch within the fiscal year by sorting out issues on the Finance Business Strategy side and setting clear directions.

How was the service received after its launch, and what do you plan to do next?

Watanabe:
We’ve received applications at nearly five times the pace we anticipated in our business plan, and we already have repeat customers. More than anything, this is evidence that our product has been well-received. I believe it’s the result of addressing the pain-points of our merchants that we identified through user interviews.

Sekine:
Hearing directly from our merchants, I realized how overwhelmed business owners are with their daily tasks. They want to spend more time thinking about how to delight their customers and turn them into fans, but they are bogged down by day-to-day operations. Through PayPay Funding, we aim to reduce the psychological and operational burdens associated with fundraising, allowing merchants to focus on their core activities. We want PayPay Funding to be a tool that helps merchants achieve what they truly aspire to.

Minami:
From the perspective of leveraging machine learning, we plan to expand the use of PayPay’s data into broader fields.
At the same time, we will continue to tackle challenges in improving the PayPay Funding product from both user and business perspectives.

Passion Stems from Trust in the Team and Product

Which of the PayPay 5 senses do you value in your daily work?

Watanabe:
Believes in our PRODUCT & TEAM
We emphasize thinking from the user’s perspective rather than as suppliers. This approach has proven successful in this project.

Oba:
I chose the same one. Many people at PayPay are genuinely driven by the desire to provide value to users. This conviction stems from our belief in the value of PayPay’s products and our team’s ability to deliver that value.

Nishiyama:
Ego is not welcome, Communication is necessary
Rigid or self-centered thinking distances us from providing the services users need. It’s crucial to quickly discuss ideas online or gather in the office for focused discussions to reach the best solutions while reflecting stakeholder’ aspirations.

Nakao:
Be Sincere To be Professional
PayPay is a group of professionals with various specializations. Respecting differing opinions while clearly expressing our own leads to the best decisions for PayPay.

Sekine:
Work for LIFE, or Work for Rice
Our goal is to create a positive cycle where merchants thrive and focus on their core activities, making their customers happy. We aim to create truly valuable products and take on new challenges to achieve this.

What is the culture like in the Data Credit team?

Watanabe:
Our team members come from diverse backgrounds. It’s a high-level environment where individuals from various fields collaborate to achieve results. Launching a new business in just nine months under remote conditions is almost unheard of (laughs), but it’s a testament to our team’s ability to harness diversity and individual strengths, something only possible at PayPay.

Nishiyama:
I agree. The Data Credit team excels at maximizing individual strengths while coming together to tackle challenges, producing outstanding results. Each member is committed to delivering exceptional performance as a pro in their field and approaches their work with a sense of urgency.

Qian:
Working alongside business-side members to develop services is stimulating and educational. It’s rewarding to consider how feedback from market, customer, and competitive perspectives can be translated into technical solutions, broadening my horizons as an engineer.

Pioneering New Forms of Finance in a Rapidly Evolving Field

What are your future goals and missions?

Nishiyama:
Being in a position to create new businesses from scratch, we must listen closely to user feedback, visualize their needs, and collaborate closely with internal departments. We will continue to prioritize user-first services, promote the use of PayPay Funding, and develop new services.
I also believe that the application of data based credit has the potential to extend beyond finance. With my background in insurance, I aim to explore new business opportunities in that sector as well.

Qian:
There’s still room for improvement in PayPay Funding, especially in the offer conditions derived from predictive models. As a data scientist, I want to refine these models to offer more attractive terms to merchants. Additionally, I’m eager to take on the challenge of planning and developing new services.

Nakao:
I want to stay involved in the data credit field. By enhancing our predictive models, we can potentially increase funding limits. My goal is to develop models that make faster and more accurate credit decisions than human judgment.

Finally, do you have any messages for our readers?

Oba:
The appeal of this job lies in engaging with data from a wide range of perspectives, from micro to macro, encompassing users, society, and finance. It’s the perfect workplace for those who want to help shape the future of finance.

Qian:
Collaboration is key at PayPay, and those who excel at bringing people together will thrive here. There are still many challenges left in the field of machine learning that I want to take-up. We welcome anyone interested in data science in this area.

Nakao:
The successful release of this service was made possible by the immense support from team members and related departments. The collaborative environment at PayPay is one of its greatest strengths. I would love to work with those who enjoy creating services through teamwork.

Tsugueda:
With PayPay Funding, we were able to provide truly useful services by deeply understanding our users. We welcome those who want to engage in such meaningful work and services!

*Job openings and employee affiliations are current as of the time of the interview.