Two years ago, Louis Carbonnier and Christophe Spoerry founded the Euler Hermes Digital Agency, the innovation arm of Euler Hermes, a global provider of trade-related insurance solutions. Besides working on ways to monetize Euler Hermes assets, EHDA is trying to reinvent trade credit insurance, initiate plug and play partnerships, and create products rooted in Big Data and artificial intelligence.
What’s going on in the industry that led you to found Euler Hermes Digital Agency?
Three big trends that provide both threats and opportunities: Platformization, data proliferation, and substitutions.
Platformization describes the growing proportion of B2B trade that now goes through platforms like Amazon and Alibaba and others. Thirty years ago, if you’d traded with someone you didn’t know in Brazil, you’d need to get credit insurance to protect yourself in case of non-payment from your client. Now marketplaces provide you with a rating on your counterpart as well as a standardised way of transacting, which decreases the perception of risk and thus the need for insurance. But there’s a silver lining: platforms also provide a single point of access to thousands of clients in an unprecedented way. What happened to travel insurance is a good example. Instead of searching for an agent or a broker, you now get it on the airline website, as a simple box ticked at the end of your purchasing journey. I think the same thing will happen in corporate finance, and such platforms will become the gateway for the distribution of financial services.
Data proliferation is the increased availability of data on the web, on social media, and on those large platforms. On the downside, pervasive data lowers the barriers to entry into credit insurance, because everyone can now find information on their trading counterpart. Again, this decreases the perception of risk and increases the temptation for clients to self-insure. On the upside, insurers also have the opportunity to collect more data than in the past and they tend to do so in a more systematic way than clients because this is their core activity. In addition, machine learning makes it possible to mine this data much more effectively than we did only ten years ago, which pushes our risk algorithms into the next generation.
The third trend is that the trade finance value chain is undergoing profound changes as good alternatives to traditional products are starting to blur the frontiers between what used to be very distinct lines of business (e.g. factoring, credit insurance, letters of credit, supply chain finance). On every step of the value chain you see some very exciting startups offering substitution products that support an improved customer experience. FundBox and BlueVine, in the U.S., MartketInvoice and URICA in the UK, or Finexkap in France provide good alternatives to invoice financing. Companies like EzyCollect, CollectNow or Debito are reinventing good old collection agencies. Argo Trade Solutions allows clients to process import-export transactions online in a few minutes and at a lower cost than with letters of credit. And that’s not even mentioning blockchain, where we see a large number of startups tackling trade finance products and processes.
The upside for us is that trade finance products haven’t progressed in decades. But with new technology available—cloud, AI, APIs, blockchain—you have the luxury to restart from a blank piece of paper, at a fraction of the cost. Letters of credit have existed for centuries and credit insurance for more than one century, all living on very old legacy systems. So my challenge is to reinvent credit insurance, right from scratch, for the benefit of our customers.
So how does this reinvention – of new value propositions and new products – take place?
In the old days, a traditional big player would have said, ‘Holy cow, we need to reinvent ourselves. We are going to put our best brains behind it, launch a big initiative and by end of the year we want to have a new product.’ But innovation is not really a top-down process and it has to be much faster and more iterative. To find a great new product, you’ll have to try 10 times, 10 proofs of concept, each generating some learning that you can use. Two to three will be a bit more successful than others, and you’ll learn a little bit from each, and maybe on seven or eight of them, for instance, we’ll learn the client doesn’t want to land on a website, but would rather want to access a well-documented API with an option to speak to a human being.
So we learn from these proofs of concept by putting something in front of the end customer – and by that I mean after just a few weeks. If the proof of concept is successful, we move quickly to the next phase, where the projects can scale up but still without freezing the product – contrary to the traditional process. As we keep getting customer feedback, we go back and adjust the offering. This iteration phase means A/B testing and changing features all the time until you have reached a clear market fit. It’s only then that you can say: ‘Okay, let’s get serious and industrialize this thing.’ This whole approach is counterintuitive for a big corporation, and really for all of us – because nobody wants to fail. But failure is very important to the innovation process, even in the B2B world.
What’s the biggest surprise thus far for you and EHDA?
There’s a lot more we can do with existing data – to an extent that came a bit as a surprise to us. We knew there was a lot of new data out there that we wanted to test, but a key learning was that we could already gain huge insights from data we already have by working better on it and by using machine learning and upgrading today’s algorithms. This has been a key focus area for EHDA and led to the creation of a dedicated team: the Euler Hermes Data Lab. Eighty percent of what we do is improving algorithms on existing data and making sure the data is siloed in a format so we can mine it quickly. The other 20% is finding more fancy sources of data on social media or stuff like that – but it’s a more complex journey because you need to weed out a lot of blank noise and cat pictures before finding the diamonds in the rough.
We were also surprised how fragmented the data is that banks and corporates have. They have all the data, and, incredibly, they don’t – or rather, they can’t – make full use of it. It’s hard to run simulations on legacy systems, for example: you can produce P&Ls and balance sheets in an ERP, but these systems are not geared to perform advanced analytics on the source data.
However, once you manage to pour your data into a data lake, you can start running analyses that yield conclusions that will directly translate into the bottom-line. For instance, we anonymised the ledger of a factoring client, dumped it into a data lake, and ran a simulation on millions of invoices. We found that one of the key predictors of credit risk was the length of the written description of the goods being traded on the invoice. The longer the description, the higher the risk of default. It’s super counterintuitive. This was a risk factor never taken into account before. Overall, we’re only at the very beginning of the Artificial Intelligence journey. There are hundreds of insights like this that will come from applying machine learning to existing data.
What’s next in the world of credit risk insurance?
Nobody has a crystal ball, but if you push me to make predictions, I’d mention three trends: finance autopilot, “Wiki Data”, and a world where insurance and financial products are done over blockchain. Note that those three trends are not mutually exclusive – in fact quite the contrary: they reinforce each other.
You’ve heard of roboadvisers? I think the same will happen to the finances of small- to mid-sized corporations before moving to larger corporates. They’ll all be automated and look more and more like a “Finance Autopilot”. Research shows that small and medium-sized enterprises (SMEs) have an increasing appetite for “peace of mind” products providing cloud-based outsourcing of admin tasks, allowing business owners to focus on their core business. We’ve reached the stage where most of the “product bricks” have been reinvented as software as a service (SaaS) covering the full AP/AR cycle (e-invoicing, information, receivables finance, credit insurance, finance analytics, collections, and etc.), and the industry is ready to move to the “Autopilot” phase.
The second trend is what I call “Wiki Data,” which is a play on Wikipedia, and reflects the power of the crowd over the power of experts. The Encyclopedia Britannica got destroyed by Wikipedia in a few years; the same happened to the Michelin guides with the rise of Tripadvisor and Yelp. Now extend this to the world of credit risk. All companies are working on small islands of data, but now, more of them are willing to share that data in order to benefit from services – e.g. through QuickBooks, which provides a SaaS accounting package for millions of small- and medium-sized enterprises in the U.S. and Commonwealth countries. The richness, the granularity of this data is unprecedented and we’ve barely scrapped the surface of the new applications that will be enabled by such technology. And all these services will be 10 times more powerful than in the past because of their access to data. So what used to be small islands of data are starting to be connected. That means there’s a massive revolution in how credit risk is analysed, and the sequel is a step change in how financial products will be priced and sold.
The third prediction is whether blockchain will be the equivalent of the Internet in financial services area. Whilst blockchain still suffers from the limitations of a nascent technology (uncertain regulatory environment, transaction throughput and latency issues, competing protocols), it has the potential to become the backbone of future trade infrastructure. As a corporate innovator, it’s difficult to bet the house on blockchain because it relies on network effects which make the outcomes very binary (all or nothing), but the transformational power is too large to be ignored. As a result, we’ve gathered a small blockchain team which explores trade-related use cases and connects to other industry initiatives.