When Small Startups Take on Big Data
Big data analytics startups combine deep domain knowledge with cutting-edge artificial intelligence to help companies as diverse as banks and shipping firms solve complex problems.
AsianScientist (May 23, 2018) – By Sim Shuzhen – When Mr Martin Markiewicz decided to go into the enterprise search business in late 2013, he and his co-founders went all in, relocating from Poland to Singapore to start their company. Staying put was not an option as it would have limited their access to large multinational customers, he recalls.
“We wanted to go to a place where we could be selling on a regional or global level. We listed all the potential places, and I applied our own scoring system to them. We were looking at everything from access to customers to access to talent and money, and at the end of the day, Singapore won.”
The decision was “purely data driven,” says Mr Markiewicz, a mathematician by training who is now co-founder and CEO of Silent Eight, which in 2014 received seed investment from SGInnovate. Aptly enough, Silent Eight also runs on data. Shortly after it began life as a next-generation enterprise search engine, its founders pivoted to financial compliance and risk; Silent Eight now uses big data and artificial intelligence (AI) to help banks and financial institutions screen customers and detect suspicious transactions.
Today, the collection and application of customer-related data is so ubiquitous that it is often said that every company is a data company. But specialised data analytics firms like Silent Eight are a different breed, zeroing in on unique, potentially highly profitable niche markets and developing deep tech-based solutions for them.
“Instead of supporting multiple customers with multiple use cases, we just focussed on one big use case. In terms of product roadmap, we develop products that fit the use case better and better all the time,”
explains Mr Markiewicz.
No dictators or drug lords allowed
Multinational banks have a vested interest in making sure that they are not unwittingly helping dictators, drug lords, spies, terrorists and other unsavoury types launder money or finance illegal activities. But picking out these individuals from among millions of customers is like looking for a needle in a haystack, and is further complicated by the need to do this screening on a regular basis.
“You have to check customers all the time, because someone who was not a drug lord when you onboarded him could be a drug lord now,”
explains Mr Markiewicz.
Even if a customer passes the screen, he or she could still be up to some criminal activity; banks thus also need to monitor transactions to understand what customers are doing with their money,
says Mr Markiewicz.
The way the financial industry screens customers and transactions today is far from ideal.
“[Banks] employ rule-based screening systems, which are designed to flag out everything that is even remotely suspicious for investigation,”
says Mr Markiewicz.
Analysts, therefore, often find themselves inundated with alerts that need to be investigated.
Silent Eight considerably relieves this burden. Its engine digests all the data an analyst would typically look at: customers’ accounts, transactions and financial products, global news updates from multiple sources, and other types of structured and unstructured information. After analysing it with machine learning algorithms, the engine outputs a report on each alert that analysts can directly use to help them make decisions. Silent Eight now counts major banks around the world, including OCBC Bank in Singapore, among its customers.
“We train our engine to work like an analyst solving alerts. So analysts, instead of starting every investigation from scratch, already have a suggested decision and all the supporting evidence,”
says Mr Markiewicz.
“This saves a lot of time within the investigation process.”
The goal, he adds, is for the engine’s output to look exactly like something an analyst would produce without Silent Eight’s system.
“[The analyst] should be able to just look at it and say yes, I can sign my name under this.”
Data on the high seas
To solve difficult problems, data analytics companies must couple the ability to handle a wide range of data types together with formidable domain knowledge. Another startup doing this—but for a very different application than Silent Eight—is Portcast, founded in 2017 through the Entrepreneur First (EF) Singapore programme.
The company is targeting maritime logistics, which handles the vast majority of the world’s container traffic.
“We make shipping routes profitable by predicting demand, using AI and machine learning that we have customised for the logistics industry,”
says co-founder and CEO Ms Nidhi Gupta, a logistics industry veteran who spent a decade in various leadership roles at DHL.
The maritime logistics industry is hugely inefficient, says Ms Gupta.
“Every decision in the industry is made based on a prediction. But because predictions are primarily made based on intuitive sense, experience, market averages and historical data, they are not accurate,”
Inaccurate demand prediction in fact results in a S$30 billion loss every year for the industry, she adds.
Developed with co-founder and CTO Dr Xia Lingxiao, a data analytics and AI expert, Portcast’s engine takes into account several baskets of data on a real-time basis: economic indices, such as currency rates and commodity and fuel prices; climate data at different locations in the ocean; and satellite location data, which pinpoints where vessels are. To this, it adds customer data, such as trade flows along various shipping routes, for each particular client.
Based on the engine’s predictions, Portcast provides shipping companies with daily information, up to eight weeks in advance, about how much cargo will need to be shipped, between which ports, and when, says Ms Gupta. This allows companies to tune their operations accordingly, redirecting resources to ports where demand is high, adjusting pricing and targeting the appropriate customers—actions that will ultimately improve profitability.
Portcast is now running two proof-of-concept trials with major shipping players, and is in advanced discussions with five more.
“We believe that demand prediction is not a shipping problem, it’s a data science problem,” says Ms Gupta.
Getting off the ground
Compared to those selling to consumers, companies that sell deep tech products to enterprise often have higher starting costs, as they do not have the luxury of starting with a beta version to be improved on over time, says Mr Markiewicz.
“The first version of your software already has to be enterprise-grade. If customers are going to pass real data through this product, they have to be sure that it’s stable and secure,”
SGInnovate’s participation in Silent Eight’s seed round (along with other investors) was thus a huge help in getting the company off the ground, he adds.
In addition to connecting her with Dr Xia, Ms Gupta also credits EF Singapore for helping to push Portcast from ideation to proof-of-concept trials.
“The programme was really intense, which helps because you set up deadlines for yourself as a company, and you want to hit certain targets and milestones. The environment is very positive.”
While Portcast’s goal over the next few years is to capture as many maritime logistics customers as possible, its engine can also be applied across the broader logistics industry, says Ms Gupta.
“We believe that our technology is scalable across the entire supply chain, not just maritime, but also air cargo, manufacturers and ground logistics,” she says.
“What we want to do is make it more efficient and transparent for cargo movement throughout supply chains. We believe that will have knock-on effects for shippers, manufacturers and for consumers like us, so that we get our product on time and at the cheapest cost. That is the ultimate goal.”