Join Paulo on call with Dino Setiawan, CEO and Co-founder of AwanTunai, Indonesia’s leading supply chain financing and ERP platform. AwanTunai is at the intersection of fintech and software, enabling MSMEs and unlocking growth in the massive more than US$100B FMCG industry in Indonesia. 

Dino returns to the show after two years to talk about how AwanTunai has evolved over that time. He shares about the value of optimizing for a sustainable business model versus top line growth, his app-less approach to building tech for the tech illiterate, scaling risk management with the right data sets and engines, and what it takes to win over world-class technical talent, among other experiences and learnings. Don’t miss out on this insightful conversation with a seasoned entrepreneur and banking professional who has been in the industry for quite some time.

Check out our past conversations with Dino on the origins of AwanTunai (in 2021 and in 2020)

Timestamps and Highlights

(00:37) Dino’s back for the third time and gives us an update in three numbers;

“For the longest time, VCs dismissed the idea that lending could be profitable. Back in the easy money days, we were like the ugly [duckling]. But now we’re getting close to profitability, and hopefully, this is our moment in the spotlight.”

(04:15) Evolution of Dino as a leader and MSME digitalization in Indonesia;

“From our side, as a lender, we were making money from the start. But risk management and scaling relied on utilizing a mobile app, which we found to be expensive and challenging in this micro tech-illiterate space. Over the past two years, we’ve grown our ERP infrastructure and adoption at the middle layer of the supply chain. We’ve replaced incumbent systems and captured high-quality transaction data. We realized that serving micro-merchants through an app-less solution may be the way to go.”

(09:36) Serendipity of Growing a Financing Business on Top of App-less Supply Chain ERP Infrastructure;

“A lot of the e-commerce marketplace folks have scaled up their systems to optimize for GTV capture, and sometimes there’s a bit of a perverse incentive where if there’s some fraudulent GTV, it’s still captured because it’s driving up the valuation. But for us, since we started life off as a lender and risk management is part of our DNA, we custom built the system to detect fraud.”

(18:38) Scaling Risk Management with World-Class Talent;

“My chief science officer once said that artificial intelligence can’t beat a trained human. At first, I was surprised by this statement since we are always sold on the idea that AI models can process 50,000 variables, whereas humans cannot. However, she made an excellent point regarding structured and unstructured data. Deep learning models need tens of millions of data points and users, and only a few companies have access to these data sets…​​understanding risk management is critical before building science models that can surpass human performance. And to get the right set of variables and proxies, experienced humans need to guide AI engine development.”

(29:28) #MinuteMasterclass: Winning Over World-Class Talent;

“One of the interesting realizations is that it’s not about the money. I mean, these are extremely expensive talents, right? But certainly, talents that we probably wouldn’t have been able to afford unless there was that X factor. And the X factor, with a lot of technical talent that we’ve been able to attract, is that this is something interesting that they want to work on…”Building Technology for the Tech Illiterate,” which I believe has a very interesting hook point for many people…getting user adoption, especially when the user is tech-illiterate or doesn’t even have a phone, is a challenging problem that is exciting to solve.”

(34:59) #RapidFireRound;

About our guest

Dino Setiawan is the CEO and co-founder of AwanTunai. He is a finance veteran with more than a decade of experience spanning several countries which includes a tenure as VP for Investments at Morgan Stanley among other roles at major financial institutions. After finishing his Master’s at the Stanford Business School in 2011, he started off his entrepreneurial journey running his own fintech venture in Silicon Valley, SimpleFi, before returning to Indonesia as a regional fintech consultant. Together with ex-Gojek execs Rama Notowidigdo and Windy Natriavi, he later co-founded AwanTunai, an Indonesian fintech startup providing accessible financing to micro-merchants by digitizing the FMCG supply chain.

Transcript

Transcripts are edited for clarity.

Paulo J: We would love to hear your perspective on the past two years. Recently, we have been asking our guests to share updates about their role in terms of numbers. It seems like everyone in this industry loves numbers, right? So could you share with us three numbers that best describe how AwanTunai has grown or evolved since our last fall?

Dino S: Speaking of numbers, as an ex-banker with 12 years of experience in banking and almost 12 years in FinTech, it’s almost a one-to-one ratio here. As for the three numbers that describe AwanTunai’s growth, we’ve had an amazing last two years with an eightfold increase in the number of suppliers, which has also boosted our embedded financing by a similar growth of eight times, now amounting to around 800 million annualized.

The driving force behind this lending growth is the valuable transaction data captured by our ERP systems, which has resulted in a whopping 14-fold growth from 2021 in our traditional supply chain space, consisting of wholesalers and distributors. By coincidence, we also experienced a 14-fold growth in contribution margin, as lending has always been a profitable venture.

Paulo J: I think that has been the whole purpose of AwanTunai, to prove that lending can make money in this industry.

Dino S: Yeah, definitely. For the longest time, VCs dismissed the idea that lending could be profitable. Back in the easy money days, we were like the ugly [duckling]. But now we’re getting close to profitability, and hopefully, this is our moment in the spotlight.

“For the longest time, VCs dismissed the idea that lending could be profitable. Back in the easy money days, we were like the ugly [duckling]. But now we’re getting close to profitability, and hopefully, this is our moment in the spotlight.”

Evolution of Dino as a leader and MSME digitalization in Indonesia

Paulo J: And you did talk about that flywheel you have of the embedded financing plus the ERP, which we’ll dive into a little bit later. But first, I also wanted to catch up on you as a leader, and as a CEO. How have you evolved with yourself as a leader? How has your role or leadership approach changed since we last talked?

Dino S: Yeah, well look, you know, with eight times growth in volumes and 14 times growth in data capture, I have to admit that the complexity of the organization has significantly increased. And that’s just running with a single financing product. It’s been a continuous, steep learning curve understanding key issues from every department and how best to align them and organize the team.

It has forced me to better understand the whole organization as a whole, both to go wider to really understand the interdepartmental dependencies and at times to go deep when you really need to troubleshoot a particular issue and go all the way down to the field to uncover what’s going on. It really felt like maturing in being able to handle a much larger and more complex organization.

Paulo J: Would you mind sharing an example of when to go wide versus when to go deep?

Dino S: Initially, when we started the product in 2019, we only had one small capital lender, which was just one of the banks. But now we’re running with multiple lending facilities, all with differing covenants, risk acceptance criteria, and even geographies. The money comes from different places, and this is just in one department, treasury. 

The complexity of asset allocation before was much simpler when we only had one lender, but now we need to allocate the assets in a way that’s fair and non-discriminatory while managing multiple pockets of liquidity. And that’s just one department where the complexity has increased almost five-fold from when we first started.

Paulo J: Yeah, but I would say it’s a good problem to have, right?

Dino S: Indeed, it’s a good problem to have. We have to graduate.

Paulo J: And we’ve talked about how the business and you have evolved. Now, I also want to ask about the market you’ve been targeting, SMEs specifically in the FMCG industry. How has their response and engagement to this digitization wave evolved since we last spoke?

Dino S: That’s an interesting trend that we’ve seen both within our company and the market. Genuine app adoption is expensive, and heavy promotional budgets are required to drive app downloads. However, there’s a struggle to monetize this particular user base. From our side, as a lender, we were making money from the start. But risk management and scaling relied on utilizing a mobile app, which we found to be expensive and challenging in this micro tech-illiterate space.

Over the past two years, we’ve grown our ERP infrastructure and adoption at the middle layer of the supply chain. We’ve replaced incumbent systems and captured high-quality transaction data. We realized that serving micro-merchants through an app-less solution may be the way to go. We don’t aim to change user behavior or force individuals without smartphones to borrow their children’s devices. We go where the customer is and build technology for the tech-illiterate. It’s almost like going back to basics.

Paulo J: So how does that conversation differ from a conversation where you’re trying to get them to use an app versus this app-less servicing that you’re trying to implement?

Dino S: It’s been night and day. We’ve been experimenting with having our suppliers run with the financing for their customers. When it’s a human approach, when it’s messaged by a trusted person for that micro-merchant, we’ve seen over 10 times the conversion rate compared to when we utilize our app-driven channels. 

It’s super exciting, although it comes with a whole host of tech challenges because essentially we are outsourcing our sales function. With that comes a lot of potential fraud risk. Risk management and fraud management have always been in our DNA, and a lot of the technologies that we invest in are geared toward figuring out how we can pass on this sales function to our network while still being able to control fraud.

“From our side, as a lender, we were making money from the start. But risk management and scaling relied on utilizing a mobile app, which we found to be expensive and challenging in this micro tech-illiterate space. Over the past two years, we’ve grown our ERP infrastructure and adoption at the middle layer of the supply chain. We’ve replaced incumbent systems and captured high-quality transaction data. We realized that serving micro-merchants through an app-less solution may be the way to go.”

The Serendipity of Growing a Financing Business on Top of App-less Supply Chain ERP Infrastructure

Paulo J: So outsourcing that sales function so that you could just invest more in the risk management side. For our listeners, stay tuned and we’ll definitely dive into that aspect as well. But first, you mentioned that you’ve been targeting this middle layer in this value chain.

And since we last spoke in 2021, I think at that time you had initially started with merchants, and then through the pandemic discovered that you could also do financing for the wholesalers. And then more recently you’ve been also doing that with higher up that value chain, I guess the distributors and principles as well.

So what have been the implications of that? And I guess you touched on it a little bit with the wholesale discussion but maybe other implications operationally with this expansion.

Dino S: Back in 2019 when we first launched the supply chain financing product, it was a very manual affair, where we had our risk teams gathering paper sales invoices and purchase orders, entering that into the database. It worked from a risk perspective, but it’s not scalable, right? It was just, much too expensive, much too high-touch to really manage. 

And that’s when we came to the realization that we had to build an ERP system here to really capture all this data. And sometimes in business, luck plays a factor. And with the pandemic, the luck factor for us was all of a sudden all of our suppliers that we were trying to push these pod systems, and inventory management systems into were suddenly starved of capital. 

The banks literally ceased lending, and some, well, in fact, most tried to pull their credit lines within this SME segment. So a lot of our suppliers came to us for financing, and that really drove the initial opportunity to bundle the supply chain financing with a supply chain operating system. And from there, we really took off. 

Right now, we’ve got over 500 suppliers with combined sales of over 3 billion USD per annum. And I believe that’s almost, if not on par or exceeding the big unicorns out there who tried to go into this mitra type of environment. The great thing is that at this middle layer here, the data capture is very efficient, right? There’s enough concentration of data for us to capture it economically and also validate it economically. 

And that validation is really the key differentiation for our company here. We’ve seen a lot of folks with mountains of GTV really struggle to try and monetize that through financial services, namely lending because it’s such a high-margin business. But for us, validating is really the key to enabling this financing to work, right? We have 3 billion USD in sales, and we’re able to lend well over 800 million USD on an annualized basis. So that’s a really high ratio.

Paulo J: Do you see the current market driving the business and continuing to drive it, similar to how the pandemic did in terms of the wholesalers needing capital?

Dino S: Yes, absolutely. Initially, in 2019, we thought that our supply chain financing would be more for the micro SME market. However, due to the unexpected pandemic effect, customers came to us wanting larger-scale SME financing, which led us to design a product specifically for the supply segment. 

We probably wouldn’t have done this without the customers coming to us, but we saw the gap in the market as suppliers came to us for financing. Now, our product is highly sought after by our banking partners. So, we definitely see the current market driving and continuing to drive our business forward.

Paulo J: This market has certainly been different from 2019, especially with the pandemic, and there are now many fintech platforms offering services to SMEs. Where does AwanTunai stand in this landscape?

Dino S: I would say we were the pioneers in transaction-based financing, where no loan asset gets generated unless there’s an inventory purchase transaction. While it’s flattering to see that many competitors are offering similar lending programs, we’ve noted that they are doing the lending without investing in a proper supply chain ERP system like we have, which optimizes for fraud detection. 

As far as I hear in the market, we are still the best performing in terms of risk management. It’s a bit annoying but also a bit proud that it’s known in the market that if you’re an AwanTunai customer, there’s this perception that that’s an automatic approval. It annoys my team because they do all the hard work of risk assessment.

Paulo J: I wanted to ask, what do you think is holding the rest of the market back from building out their own supply chain ERP systems?

Dino S: It’s an interesting thing because when we did the build versus buy analysis back in 2019, we realized that the third-party ERP systems in the market were not designed to detect fraud that we knew existed. We couldn’t see that in the general market ERP systems out there. So that’s how we ended up building the ERP system from scratch, specifically focused on fraud detection. Maybe that’s been the differentiating factor. 

A lot of the e-commerce marketplace folks have scaled up their systems to optimize for GTV capture, and sometimes there’s a bit of a perverse incentive where if there’s some fraudulent GTV, it’s still captured because it’s driving up the valuation. But for us, since we started life off as a lender and risk management is part of our DNA, we custom-built the system to detect fraud.

Paulo J: Yeah, that’s definitely an important point to emphasize. Let’s dive into the differentiators – ERP and fraud detection and risk management. Can you illustrate how the ERP has been changing the way they operate and what that means for long-term benefits for business owners?

Dino S: Most of the largest suppliers have developed some kind of custom system because it’s a fragmented software market. Our operating system and financing have played a strong role in enabling the best-performing suppliers to grow significantly – we’re talking 4x-5x or even more. 

For example, one of our best-performing wholesalers initially had just two warehouses. With the control systems we put in place using our operating system, they were able to manage more remote sites and grow to 11 warehouses within one and a half years. No bank would lend at a 500% increase as it’s too risky. Our model not only lends, but our operating system helps manage businesses and provides deep visibility into operations.

“A lot of the e-commerce marketplace folks have scaled up their systems to optimize for GTV capture, and sometimes there’s a bit of a perverse incentive where if there’s some fraudulent GTV, it’s still captured because it’s driving up the valuation. But for us, since we started life off as a lender and risk management is part of our DNA, we custom built the system to detect fraud.”

Scaling Risk Management with World Class Talent

Paulo J: The ERP has enabled businesses to expand and grow, but another crucial aspect we’ve been discussing is risk management. This is vital for protecting customers and maintaining the ecosystem you’ve built. Could you share with us how your approach to risk management has evolved over the past few years?

Dino S: My chief science officer once said that artificial intelligence can’t beat a trained human. At first, I was surprised by this statement since we are always sold on the idea that AI models can process 50,000 variables, whereas humans cannot. However, she made an excellent point regarding structured and unstructured data. Deep learning models need tens of millions of data points and users, and only a few companies have access to these data sets. Hence, deep learning is not feasible for startups like ours.

This is why we believe that humans need to train AI engines. For instance, when a borrower is verified before loan disbursement, the credit officer will validate their name, address, and borrowed amount. These are three variables in a structured data model, and it’s almost like a binary input. However, the credit officer’s call with the borrower generates unstructured data that is crucial to risk management.

An experienced underwriter listening to the call can determine whether the borrower’s responses are unusual or suspicious. If the underwriter detects anything suspicious, they ask more questions. This unstructured data doesn’t end up in a 50,000-variable model, and that’s where the human element comes in. We always validate the risk process or outcome, whether it’s possible with humans.

Many AI credit-scoring vendors sell their products by claiming that they can separate good users from bad ones. However, fraudsters have become sophisticated and can even game the system. For instance, they can make three successful payments, which increases their limits. Once the limits go higher, they hit the lender harder. 

This is why understanding risk management is critical before building science models that can surpass human performance. And to get the right set of variables and proxies, experienced humans need to guide AI engine development.

Paulo J: Ultimately, AI is based on human activities. So it’s important that the output is well-founded with well-defined ingredients. This is part of the theme of our discussion today, which started from the app-less discussion and now goes back to the basics of how humans analyze risk. It’s essential to get the best-in-class expertise to train the models that you have. 

Speaking of ensuring that the humans that are training the AI are actually good at their job and scaling that, what have been the challenges so far in ensuring that you’re able to do this across different categories, scenarios, and use cases?

Dino S: The science behind automating manual field risk management operations is not trivial. Back in the early days, we were thinking about how to automate a field survey. A human goes down to the shop and checks whether the borrower is the owner of the shop. We can’t send a drone down there to do that.

But we’ve patented a lot of the science work of finding the right proxies. A strong baseline of good decision outcomes from our humans allows the science team to start looking at a range of proxies to mirror human decision outcomes. A field survey agent will say X. What are all the proxies that get close to X? 

Finding the right proxies has been a challenge. In our early days, we followed a similar methodology. We couldn’t find strong proxies in publicly available data sets, such as purchasing data from telcos or social media scraping. There was too much noise in the data. However, when we went into the supply chain, a lot of the inventory purchase data and sales data was highly predictive of business and credit performance.

Paulo J: AwanTunai has evolved into a more mature operation since its early days in 2017 when it was still figuring things out. For investors out there, hearing about AwanTunai may remind you of other businesses in emerging markets. Are there any learnings from these other businesses that you don’t want to do, or anything that influences or impacts the way you approach things with AwanTunai?

Dino S: There are two critical factors: execution and building the right team. In the early days of 2017-2018, it was diabolically difficult to get data scientists. We had a few data engineers build out an AI engine but were heavily reliant on off-the-shelf machine learning solutions. 

Over time, we have been able to leapfrog to where the majority of our risk assessment is run by the engine. We don’t need humans to go into the field or underwrite, and the majority of our transactions are run by the engine. 

Our science team is amazing, spanning the world with our chief science officer based in Chicago and folks in the UK, China, and more. Even my data scientist in China joined the Stanford Computer Science Ph.D. program, which only admits 50 people worldwide each year. 

The pandemic has enabled us to cherry-pick top talent across the world with the work-from-anywhere setup. It all comes down to execution, and we see many copycats out there, but our risk performance is superior. 

Most risk management practices rely on publicly available data, but we built a full-blown ERP system from scratch to capture and validate transactional data. Traditional suppliers often have double or triple bookkeeping, and we built a system that can capture and validate data to ensure accuracy.

Paulo J: I mentioned earlier that the organization has world-class talent, and that’s definitely not an exaggeration. They’re actually some of the best in the world. It’s great to see the evolution of the organization in that way. 

Speaking of evolution, something I often ask growth stage founders and leaders who come on our show is what are the things that you’ve had to let go of or unlearn as a leader? When we come into these roles, we carry assumptions and baggage, but in order to stay and endure, we have to let go of some things. So for you, what were those things that you’ve had to unlearn as you’ve grown AwanTunai to where it is today, after 12 years of being a fintech founder?

Dino S: Well, I’ve felt genuine growth in my personal managerial capacity and the ability to comprehend an increasingly complex business. With that, I guess I wouldn’t say unlearned, but it comes with a more mature approach to handling a range of issues such as talent management, decision-making under uncertainty, and strategic planning. 

If I were to sum it up, it would be an evolution in how I make decisions. I make fewer knee-jerk reactions and don’t accept things at face value. This comes with experience and being able to make better decisions with limited information, and understanding how to validate internal data points before making a business decision.

“My chief science officer once said that artificial intelligence can’t beat a trained human. At first, I was surprised by this statement since we are always sold on the idea that AI models can process 50,000 variables, whereas humans cannot. However, she made an excellent point regarding structured and unstructured data. Deep learning models need tens of millions of data points and users, and only a few companies have access to these data sets…​​understanding risk management is critical before building science models that can surpass human performance. And to get the right set of variables and proxies, experienced humans need to guide AI engine development.”

#MinuteMasterclass: Winning Over World-Class Talent

Paulo J: You talked about your chief science officer and your chief data scientist being some of the best in the world, and how the pandemic and work-from-anywhere have contributed to your access to this talent.

Could you share with us a key lesson in actually convincing them to join the ship? Why would they choose to focus on a business that’s working in Indonesia, specifically with MSMEs and doing lending specifically?

Dino S: One of the interesting realizations is that it’s not about the money. I mean, these are extremely expensive talents, right? But certainly, talents that we probably wouldn’t have been able to afford unless there was that X factor. And the X factor, with a lot of technical talent that we’ve been able to attract, is that this is something interesting that they want to work on.

My chief science officer is super experienced in data science, and I was actually curious because the competing job offer she had was mind-blowing. I won’t reveal it, but it’s a huge US corporate leadership position. I chased her for three years, so when she finally accepted, I was super grateful. But I also asked her why she joined us, and she said that she really enjoys creating new science, and she felt that our dataset was unique in this regard.

She said that our supply chain purchase transaction data, especially in the traditional segment, is not visible at scale like she’s seen in a few data science papers. But with our infrastructure, ERPs in place, and middle-layer suppliers, we’re collecting it at scale. She feels that this is new data where new technologies and science can be developed, and that’s what excites her. It’s not about the money; it has to be something really interesting for top talent to work on.

Paulo J: Right, and you mentioned an interesting detail there that I’d like to explore further. You mentioned that you chased after your chief science officer for three years before she finally joined the team. I’m curious, what changed in your pitch or approach that convinced her to come on board?

Dino S: Well, for her, it really came down to our data set. When we first started chasing her in 2018, we were only sourcing publicly available data. We would purchase telco data or scrape everything we could from smartphones, but it was data that everyone else also had. There was nothing truly proprietary about it.

However, as we transitioned into collecting data from traditional supply chains and building out our own ERP system, that’s when our data collection became very interesting for her. She saw the unique value in the data sets we were collecting and believed that new technologies and science could be developed from them. It wasn’t about the money for her, it was about the opportunity to work on something that truly excited her.

Paulo J: And so if you were to give a class on this to some fellow founders or CEOs trying to hire top-class talent, what would be the key takeaway for this class? Is it to chase them for three years or something else?

Dino S: Well, passion is always talked about in business, but different people have different passions. In my case, I caught my chief science officer at just the right time in her career when she was sick and tired of US corporate politics. She’s been a CTO of large organizations with 200+ engineers. So, sometimes, the stars just need to align. 

But if I were to give a lecture, I would call it “Building Technology for the Tech Illiterate,” which I believe has a very interesting hook point for many people. One real barrier to adoption is the well-beaten path of trying to get people in remote areas to download an app on a smartphone and make it work. It’s great on paper and in pitch decks, but getting user adoption, especially when the user is tech-illiterate or doesn’t even have a phone, is a challenging problem that is exciting to solve.

“One of the interesting realizations is that it’s not about the money. I mean, these are extremely expensive talents, right? But certainly, talents that we probably wouldn’t have been able to afford unless there was that X factor. And the X factor, with a lot of technical talent that we’ve been able to attract, is that this is something interesting that they want to work on…”Building Technology for the Tech Illiterate,” which I believe has a very interesting hook point for many people…getting user adoption, especially when the user is tech-illiterate or doesn’t even have a phone, is a challenging problem that is exciting to solve.”

#RapidFireRound

Q1: What digital technology or innovation excites you the most today?

Dino S: While many are excited about digital technologies, I’m actually intrigued by recent breakthroughs in fusion energy.

Q2: If you were to produce a Netflix or OTT series, what would it be about and what would be the title of your show?

Dino S: My show would be like Silicon Valley, but more like a sitcom. We all need to laugh once in a while, especially at ourselves.

Q3: Looking back now, what skill do you think you should have learned when you were a student?

Dino S: I definitely could have benefited from learning stress management.

Q4: If there is something you could automate in your job, just by wishing for it, what aspect of your role would that be?

Dino S: Performance evaluation is a painful task, so I wish I could automate it.

Q5: What is your favorite go-to destination in Southeast Asia? What trip are you most looking forward to taking in the region?

Dino S: Bali has become my go-to place for business meetings. I think everyone needs a break and some stress management.

Q6: What’s your favorite activity to de-stress?

Dino S: As a PC gamer, playing computer games is my favorite way to de-stress. I’ve been playing World of Warships for quite some time.

Q7: What is something that you’ve read or taken up recently that you’d like to recommend to our listeners?

Dino S: Although it’s not recent, I highly recommend meditation for stress management.

Paulo J: Thank you for joining us today, Dino. It’s been great having you back on the show after two years. Hopefully, we won’t have to wait another two years to hear about your new projects and insights. Thank you for sharing your counterintuitive learnings with us. In this industry, we often rely on certain assumptions or “truths” that may not always hold up, such as the idea that everyone needs a phone or app to be efficient. Similarly, the belief that AI alone can solve all our problems is not always accurate. Thank you for providing clarity on these topics and offering valuable takeaways for our listeners, whether they’re in the industry or not.