It was sunny skies and unseasonably warm on September 12, 2019 as over 120 language industry leaders gathered at the W Hotel in the heart of the city to take part in a sold-out SlatorCon San Francisco.
Participants came to mingle and network with industry decision-makers, which included CEOs, owners, and senior executives of leading language service and technology providers, private equity and venture capital investors, technologists, and service and tech buyers.
The optimism was palpable among participants, as speakers painted a picture of a healthy and growing industry, which has become the target of renewed investor interest over the past couple of years. The program ran the gamut — from language data annotation at scale, enterprise localization, NMT, M&A, corporate development, healthcare interpreting to venture capital and private equity funding — shining a spotlight on just how many different stakeholders contribute to this vibrant industry. Presentations followed SlatorCon’s trademark style of back-to-back 25-minute talks.
Slator Co-founder and Commercial Director Andrew Smart welcomed participants and reminded them to take full advantage of the day’s program and networking opportunities, which, for the first time, included an event app, allowing participants to connect more easily with their peers.
Then Slator Co-Founder and Managing Director Florian Faes kicked off the day’s packed agenda by giving an overview of the USD 23bn language industry’s main markets, players, tech developments, and growth opportunities.
He outlined how the industry’s five Super Agencies outperformed their smaller rivals in 2018, surveyed the current M&A environment, and touched on recent deals — such as Big Language Solutions’ acquisition of ProTranslating and the BTI Studios and IYUNO merger.
Faes explained how, in his view, the industry as a whole is undergoing a major operational retooling as (neural) machine translation is becoming an integrated part of the supply chain; a process driven by both the availability of the technology and clients’ increasing desire to reap NMT-related efficiency gains and pricing advantages.
He closed by highlighting media, gaming, life sciences, and the broader SaaS and tech space as growth areas for providers in the next couple of years.
Next up was Claire Tsai, Head of Globalization at Cloudflare, which went IPO the day after SlatorCon. According to Tsai, the advent of social media has shifted the way customers around the world interact with websites and other digital content. “It’s actually something really dynamic, mostly user-generated,” Tsai said, adding that Cloudflare focuses on specific data when analyzing customer intent and buying behavior.
For the greatest impact on the bottom line, Tsai explained, Cloudflare decided to concentrate its efforts on localizing the user experience. The fast-growing cloud platform and cybersecurity company built a flexible framework that promotes collaboration with industry specialists and local experts, and allows content to be tailored to the market.
Phil Shawe was next. The TransPerfect CEO walked participants through the company’s journey from its humble beginnings in the early 1990s to what is now the language industry’s largest provider by revenue. Shawe said he is looking for people with an entrepreneurial mindset to build and run two dozen smaller divisions that, together, form TransPerfect.
He said the company gives individual sales managers control over where in the organization a particular project is processed, thereby fostering healthy internal competition. Shawe explained how they have created a culture where “no organizational problem is too small to care about” and where they make financial deviations (e.g., a gross margin miss) as burdensome and painful as possible. Shawe said they try to maintain the energy of a startup at the 27-year-old industry leader, not least by spending big on international travel in order to align the company culture.
Next on the agenda was the panel discussion on institutional investment in the language industry. Hosted by Slator’s Faes, the panel was joined by Tomasz Tunguz, Partner and Managing Director at Lilt investor Redpoint Ventures; Sri Chandrasekar, Partner at Point72 Ventures, an investor in NLP startup PolyAI; and Charles Stubbs, Partner at MSouth Equity Partners, which teamed up with language industry veteran Jeff Brink to form Big Language Solutions and, in July 2019, acquired Florida-based ProTranslating.
The discussion revealed the interesting differences between the traditional world of private equity investing, which looks for moderate returns by partnering with proven businesses in growing markets, and the high-risk, high-reward approach of venture capital.
Redpoint’s Tunguz predicted how a new kind of AI agency will transform the language industry. “They’re going to look just like the existing agencies, but they’re going to have a different engine; a fundamentally different engine that is far more efficient.”
Tunguz explained, “They’re going to go and sell exactly the way an agency would sell, but under the hood they’re going to be much more efficient; which means they’ll have higher gross margins. So they’ll either have one of two passes: They can either run the business at a higher net income margin or they can underprice the market.”
Sri Chandrasekar of Point72 echoed Tunguz, adding, “There’s also an opportunity for efficacy. You make people actually better and deliver a higher quality of service in some way, shape, or form, whether it’s faster or better. But perhaps just as importantly, technology is enabling a whole new class of language translation and actually expanding the size of the market.”
MSouth’s Charles Stubbs said that he might have a slightly more simplistic strategy than his fellow panel members. Stubbs explained how they like the language industry because of its fragmentation, huge customer demand, and high single-digit annual growth. Although he acknowledges that technology disruption is coming, Stubbs thinks they have a sufficient moat around the business over their firm’s typical investment time horizon.
Following an extended networking break, Chip Huyen, Senior Deep Learning Engineer at NVIDIA, took to the stage. Huyen highlighted the ways academia differs from “the real world” when it comes to neural machine translation: “Real-world applications are more complex than academic datasets,” she said.
Academia has more control over factors that can impact the quality of machine translation, such as sentence length and subject matter, she said, while industry is largely governed by the content that clients provide.
Huyen outlined some of the current research directions in NMT, such as document length, translation, and unsupervised training, as well as quality estimation, which she described as “underexplored” in academia and mainly “driven from industry.”
Much of machine translation is powered by data and that is the world of Appen, the world’s leading provider of data for numerous AI and machine learning applications, such as speech and image recognition. “The more data the better,” Appen CEO Mark Brayan said, adding that one needs to keep in mind that “data is expensive, complicated to collect, and complicated to work with.”
One such complexity is the sheer number of participants required to create the underlying data. According to Brayan, Appen’s recruitment team works around the clock, processing 100,000 job applications each month in order to staff their crowd operations. It is not just a numbers game, however. For speech data collection, customers may specify not just the languages required but the accents, ages, settings, and even excitement levels required for their particular use case, Brayan said.
Next was managing healthcare interpreting at scale. A 900-bed safety net hospital was the ideal setting for piloting an in-house interpretation call center, according to Meredith Stegall, Director of Language Services at Parkland Memorial Hospital in Texas.
Challenged by the 1.3 million interpretations required yearly, Parkland Memorial Hospital outsourced 70% of interpretations to third-party vendors in 2017. The move cost Dallas County taxpayers USD 9m and prompted Stegall to explore a “better, smarter, faster, cheaper” way to meet patient needs.
Since launching in November 2018, the call center has employed over 80 interpreters, achieved significant savings, and sparked new opportunities for collaboration using Parkland’s language data.
Closing the day was Smartling’s Jack Welde, CEO of the New York-based language services and technology provider. Welde made the case for a much more data-driven approach to localization, where clients can track content and understand how local versions perform. He presented a number of use cases, including highly-branded content where quality was key, e-commerce content where automation and technology made the difference, and a big manufacturer, which increased the number of languages while attempting to maintain the same cost.
Welde concluded by emphasizing the crucial role of individual linguists to localization success. Smartling recently published a book called Move the World with Words, showcasing a number of their partner translators.
Slator’s Andrew Smart ended the conference by inviting participants to join the evening reception out on the hotel’s fantastic social terrace. Next stop, SlatorCon Amsterdam on November 28!
SCSF19 Presentation Meredith (Parkland Hospital)
SCSF19 Presentation Mark (Appen)
SCSF19 Presentation Phil (TransPerfect)
SCSF19 Presentation Chip (NVIDIA)
SCSF19 Presentation Claire (Cloudflare)
SCSF19 Presentation Jack (Smartling)
SCSF19 Presentation Florian (Slator)