Preferred Infrastructure (Preferred) is storming Japan’s big data landscape. The thirty-strong company was founded by graduate students, overcame Japan’s own busted tech bubble, and now sells its natural language processing software-as-a-service to some of the country’s top firms.
The list is media-heavy, with the Asahi Shimbun, Nikkei Business, and NHK (the national public broadcaster) leading the pack. However, Preferred’s services are varied enough to also entice the likes of NTT, Japan’s largest telco. Preferred’s promise of providing installation and maintenance on revenue-generating software has proved to be appealing so far. In fact, the co-founders of Preferred would say that their company could have an appeal worth approximately US$200 million.
When Toru Nishikawa (President) and Daisuke Okanohara (Vice-President) created Preferred in 2006, they accurately predicted the current boom in big data analysis. “Big data analytics” usually conjures up images of an all-seeing corporate eye but Preferred’s services are more varied. Though it does offer technical solutions for companies wanting to get inside the minds of their customers, its initial ascent was fueled by a different sort of analytics.
NLP goes to grammar school to get rich
Preferred’s first big hit was Sedue, a natural language processing (NLP) search engine. NLP technology is what allows a machine to understand language in the same way as humans. This is an enormous challenge, as Google Translate’s noble attempts at multi-lingual interpretation show. Without NLP, a machine would be light years away from matching a human’s grasp of language. It would be forced to rely on basic grammatical heuristics like “a pronoun’s antecedent is the immediately preceding noun.” That approach is wholly insufficient as seen by the following sentences (h/t to Wikipedia Japan for the example):
*We gave the monkeys the bananas because they were hungry
*We gave the monkeys the bananas because they were over-ripe
Detecting these nuances is how Preferred became a trusted client. The technology is let loose inside a media company’s servers, allowing that company to better tailor reading suggestions to users. Better search results means more time spent on the website and more exposure to advertisements, an essential combination for any media firm wishing to survive in the digital age where pageviews matter more than circulation numbers.
Two doctors and a baby
Nishikawa started working on Preferred when he was still a doctoral candidate at Tokyo University. The initial team was himself and five friends who had met at a programming contest. In recent years, doctoral students face an uncertain future, with 47.5 percent unemployed or underemployed. While those students are at the mercy of the job market, Nishikawa was always more interested in creating his own opportunities.
Preferred was born during 2006, the final year of their university enrollment. The company found early traction working on government projects. With sleepless nights and the sacrifice of weekends, it stayed afloat for the next couple of years but might have been felled by the 2008 global financial crisis if not for a fortunate deal. They managed to get a meeting with Asahi Shimbun and wowed their potential clients. That deal gave Preferred legitimacy and financial breathing room. The company still has not taken any outside funding.
Though the company became known for Sedue, its future may lie with Jubatus, an open-source data analytics tool. The technology was developed in association with the NTT development team during 2011 after they became dissatisfied with using Hadoop. As Nishikawa explains, “We were not trying to satisfy a need. We were drawn to the idea of creating something that could do scalable, real-time analytics.” (translation ours) That interest lead to a new vertical and attracted the National Transportation Authority as well as a firms involved in healthcare and the so-called internet of things.
Pressing for more
For their obvious business applications, big data/NLP firms have become hot property recently. In 2013, Intel acquired two firms for approximately US$66 million. Yahoo also made a splashy acquisition of its own last December though terms of the deal were not disclosed. Preferred might be in line for a similar windfall but its founders are expecting even more favorable terms.
When pressed on the possibility of Preferred entering a funding round for the first time, Okanohara cites the recent interest in the internet of things and says that funding might be useful for expanding the company’s operations. But Nishikawa estimates that funding at this time would be against a US$150-200 million valuation. By his reckoning, Preferred is better compared the artificial intelligence startup Deepmind, recently acquired by Google for US$500 million.
Nishikawa’s confidence stems from his team, thirty strong and almost entirely made of engineers. He prefers to let the technology speak for itself, employing only four non-engineers. He can see a future where the company scales up to 100 employees and enters the global market in earnest. Already, Preferred has placed two employees in Silicon Valley, with two more to follow. If things progress smoothly, Singapore could be in play as well.
Despite the growing interest from investors, Nishikawa is straightforward when he talks about the company’s future. “We’re not really thinking about acquisition,” he says. “It’s a possibility but we would probably focus on an IPO first and focus on keeping our product strong.” Coming from Nishikawa, this sort of thinking is not surprising.
When asked why he made a company instead of getting a cushy job at a big firm, he talked about how he recognized that it was better to work together with his programming friends rather than they all do their own projects. “A lot of my friends went to Google but I was more drawn to the chance to make my own venture and start a new company. We could realize our vision of creating something new with computer science.”