Translation: more, better, faster – with Safaba
Sometimes I get the pleasure of writing about a startup from the very beginning because I was there. Safaba is one example. The two founders, Alon Lavie and Bob Olszewski, came to me for some advice during my stint as embedded entrepreneur for Project Olympus at CMU. It was late Spring, 2009, and the pair were thinking of creating a startup in the language translation space. I was told that the name “Safaba” came from two words in Hebrew, “safa,” meaning “language,” and “ba,” which means “within,” so “language within.” I loved it.
Alon is a research professor in the machine translation area within the Language Technologies Institute in the School of Computer Science at CMU. His whole career has been at CMU, since he came to Pittsburgh from Israel 23 years ago to get his PhD. Bob also got his PhD from CMU, which is where they met. Bob’s career took him out of academia into an early-stage CMU spinout, where he led software product development for assessing and teaching spoken language skills.
They told me something that I liked even better than the name, which was that they had not come to me looking for advice about how to commercialize something that stemmed from Alon’s research lab. Rather, they had discovered an industry need and believed that they could develop a solution. Alon discovered the problem through his leadership position in the Association for Machine Translation in the Americas (AMTA), an organization that brings together people working in the machine translation (MT) field from both academia and industry. It was through AMTA that Alon learned how difficult it was for companies to actually use MT:
“The technology itself is complex, poorly understood and hard to use. It was also lacking in accuracy. Most companies told me that you had to have a team of experts on staff to deploy MT. So they were living with existing solutions and old paradigms. I thought that I could build something more far more suitable for commercial enterprises which would increase throughput and lower cost, which is the whole point of MT.”
Separately, Alon had been approached in his lab by the CEO of a regional language service provider (LSP), who realized that he was way behind the curve in exploring the use of MT and thought that Alon could help. He was willing to fund some early beta tests. Alon recounts the story:
“He [the LSP CEO] had heard about MT. He knew he should consider using it but did not know how. We started to look at the market broadly and realized that the existing technology was far behind. It was clear that the demand in terms of volume was growing. There was a real opportunity here.”
The newco. Alon and Bob founded the company and raised a bit of seed capital. They created software that makes building custom MT solutions affordable for commercial clients, and a platform to meet the growing demand for fast, accurate translation. They embraced a SaaS (software-as-a-service) model which gives them control over development and implementation, and an easy integration path with their clients’ software systems. Safaba’s secret sauce is in the underlying technology which allows them to specialize the translations they generate for each enterprise client and to produce accurate translations that align with the client’s specific branding language.
Funding. Safaba was supported with seed funding from Idea Foundry, a regional economic development organization, a grant from the Keystone Innovation Zone (KIZ), CMU Gap funds, NSF SBIR Phases I and II awards, and Innovation Works, Pittsburgh’s largest economic development organization (for a total of ~$1M to date).
Safaba has several large revenue-generating clients. The company has developed a partnership with Welocalize, one of the 10 largest commercial language service companies in the world. To date, Safaba has built and deployed systems for 27 languages, including some of the more challenging ones for MT.
Today, Safaba is still small with 10 people: Alon as President and CEO, Bob as VP of engineering, Udi Hershkovich as VP of business development, two full time software engineers, and five part-time developers who are graduate students at CMU.
The company is based in the Squirrel Hill section of Pittsburgh, ironically, in an office leased by one of my former students.
Why the market needs Safaba. Safaba helps companies expand into new, global markets by giving them the tools necessary to transform their large volume of content (which is largely in English) to the language of the new markets – the process of localization. The need for translating web content, user interfaces, instructions and the like has never been greater. Manual translation has not been able to meet demand fast enough and affordably. MT helps handle the ever-growing translation volume, but accuracy has been a challenge and the technology has been notoriously hard to use. What is needed is a solution that generates high-quality, individualized output that is easy to integrate and use – which is exactly what Safaba provides.
Safaba’s market consists of B2B global enterprises in industries ranging from IT to finance, legal, manufacturing and pharmaceuticals. The platform is not oriented towards the consumer market. Google Translate, for example, is a free consumer-grade automation solution, but it’s not adaptable to the language of individual users. “It might be fine for consumers, but not for Fortune 500 companies,” Alon tells me.
Udi has been with Safaba for six months and has thoroughly investigated the market:
“The market for translation is $34B worldwide, growing at 12% per year. Half of that is in translation services. That half is split between translators and middle men. Translators are about a third. So we are now down to ~$6B. Translation automation doesn’t replace translators. But it is disruptive and it will change the market. About half of the existing market can be automated; that’s our market: $3B.”
Udi also tells me that the bigger and faster moving market for MT may be “MT-enabled corporate business processes which can introduce real-time multilingual communications where traditional translation is simply not relevant.”
Founders’ chemistry. Who the founders are is always important in a startup. Alon and Bob have complementary skills: Bob has engineering experience building software and managing large projects; Alon understands the solution and the technology and has experience managing a research group. The two founders brought Udi on board to help cover the business strategy, sales and marketing aspects, areas in which he has significant prior experience. They are also a couple. Bob tells me, “We don’t keep it a secret, but we also don’t make a big deal out of it.”
Alon gives me his take:
“Bob and I formed this company together because we figured that we could do this better together. There are advantages and challenges in doing this. One aspect is that I knew that I couldn’t do it alone. I knew the skills that Bob would bring to the table. I also knew that we already had the necessary level of trust. That’s really important. But that doesn’t mean that there aren’t challenges. Mixing personal and business is stressful.”
I ask how their being a couple is perceived by others; what happens as they hire people? Bob’s answer is crisp:
“We don’t ask them ‘Who did you vote for?’ Or, ‘What’s your religion?’ But I am concerned about appearing that we are really only one person. I am on guard so that that doesn’t happen.”
Challenges: Alon is clear that, like any startup, the company faces mountains of challenges:
“One was why we brought Udi in. We don’t have business backgrounds. We have no experience in scaling up a real business – business development, marketing and sales.”
They also face pressure on pricing, Udi tells me:
“We thought we would be getting more money from each deal. Customers want us to do it cheaper – it’s a commoditized market. The mindset is a price-war market. People are used to talking about price not value. So they don’t do any comparison.”
Alon talks about the difficulty in bringing a new technology solution to market:
“Customers are still at the early adopter stage. That means the end clients are driving the business. They need it. But there are only a few early adopters that get it at this point. They don’t care where they get their solution from. This sets the barriers fairly low. We need these guys to prove the value of our solution so that others in the market will follow.”
I am sure that Safaba has the right stuff. I could tell when I met them, and I have supported the company all along. They discovered a real need, built a solution that worked, developed early customers and traction, and are now reaching for the sky. Isn’t the way it is supposed to work?
