Smartling CEO Jack Welde on Embracing MT and Scaling with Customers – via slator.com


Smartling CEO Jack Welde on Embracing MT and Scaling with Customers

Smartling CEO Jack Welde joins us for SlatorPod #58 to talk about technology-enabled language services, scaling with fast-growing customers, the M&A and funding wave sweeping the translation management landscape, and much more.

Jack shares why Smartling’s attitude toward machine translation shifted from “MT bad” in 2010 to fully embracing the technology in creating their own NMT engines in 2020. He also talks about his experience of running a 200-person company through a laptop in the era of Covid-19, and steering Smartling through its best-ever year in 2020. 

First up, Florian and Esther discuss the news of the week, leading with the breaking news that UK-based TMS provider XTM has sold a majority stake to California-based investors K1, after nearly two decades of bootstrapped growth.

Advertisement

Florian talks about another language industry investment story from the past week, as Germany-based Lengoo announced they raised USD 20m in Series B funding based on an AI-agency investment thesis. 

The two discuss the market for MT data services, as machine translation provider SYSTRAN launched its MT Model Studio, which allows people to enhance existing MT models with their own bilingual or monolingual data, using SYSTRAN’s tech to clean and prep the data. Users can then choose to publish their models for sale on SYSTRAN’s marketplace, or reserve the models for their own purposes. One of the data sellers on the SYSTRAN marketplace is TAUS, which recently launched its own data marketplace.

Esther shares an update from Sweden-based media localization company Plint, which appointed a new CEO in February 2021. Thomas Roberts comes from a product background and joins Plint as the company continues to develop its cloud-based Plint Core platform, both for internal use and as a product licensed to customers.

Subscribe to SlatorPod on YouTubeApple PodcastsSpotifyGoogle Podcasts.

Stream Slator webinars, workshops, and conferences on the Slator Video-on-Demand channel.

Transcript

Florian: Smartling is one of the core companies at the heart of the localization industry. First off, what is Smartling in a nutshell, a bit of the history and the milestones? 

Jack: In a nutshell, Smartling is a translation company. It is just that simple. We provide a variety of technologies and services to help our customers solve whatever problems. It is a real end to end solution of translation technology of language services and professional services. 

We started the company a little over a decade ago. I am an engineer, I have been developing software since I was about eight years old. I have a computer engineering undergrad and a minor in linguistics and was always interested in how software and language were comprised. It sort of lights up that part of my brain. I have been involved in a lot of startups over the years and at some point in every part of the startup somebody says, we have got to localize our software, we are selling to people around the world. That just was always a surprisingly daunting problem, a really big challenge. 

Back in 2010, we saw the world changing pretty rapidly. The iPhone was three years old at that point. The cloud was becoming really important with every platform moving to the cloud. Every platform was creating APIs and an expectation of interoperability. As mobile expanded, more and more people had computers in their pockets rather than on their desktop. All of these things just led to a huge demand in language. As the world moves to the cloud, more mobile, more languages, more agile content production, people producing content to be translated all the time on an hourly basis, not on a monthly or quarterly basis, we thought a company that produces the right technology platform combined with the right services could be an important and valuable company. That is what we did. 

Florian: We met a couple of times before you were also an air force pilot? 

Jack: I was an evaluator pilot for about 10 years. I still fly today. It was a lot of fun and I think part of that was I lived and worked around the world. I spent a lot of time outside of the US and I that was a pretty good lesson in thinking about people and culture and language. Frankly, we bring a lot of that approach to how we run and how we build smart Smartling.

Florian: Is there anything from that time that is helpful for running a company immediately? Or is it more the general mindset when you are building a company? 

Jack: I have literally stolen all of our core values from the military, just thinking about things like how do you execute with imperfect information? How constraint drives more creativity and how people solve problems? How do you take care of your people? These are all core values of great organizations so we have brought all of those into Smartling and it has helped us to grow incredibly fast. It has helped us to better serve our customers and to balance the needs of the customers, the company, and our individual employees and partners. 

Esther: You have grown, you have scaled. How would you continue to size the overall market opportunity for Smartling and also in the context of the wider language?

Jack: How big is the market? CSA will say 50 billion. Nimdzi says 57 billion. Slator’s latest estimates are 25 billion. That is a lot of money that is being spent every single year on translation. What is important though, is that we think about Smartling’s role in there. We have grown incredibly fast, I would argue that we are one of the fastest, if not the fastest-growing technology and service-oriented company in the industry. It is a big market and there can be a lot of players. Customers are probably looking for that true overall suite of technology and services that can help them to solve their problems. Frankly, there is plenty of room for everybody and I am pretty excited about the overall market opportunity.

Esther: Tell us about the mix between services and technology then and the move to add services? When did that come about and how? 

Jack: At Smartling we have been a technology and services company from the very beginning. I hire world-class engineers that build incredible, scalable, secure, reliable platforms but I hired my first two translators within a month of starting the business, they both still work with us. Sometimes people do not realize that because we are building what I think are the market’s leading technology platforms. We talk a lot about our TMS technology, our proxy-based technology, our connectors and so on. Ultimately, I do not know any large freestanding tech only businesses in this industry. Every one of them is small. The market for people that are looking for only pure-play technology is incredibly small in this industry. Most companies will figure that out pretty quickly. We figured it out within about 30 days that it was something that we had to do. 

Most companies think about looking for a great service provider that has the ability and the knowledge and the understanding to help them to solve their problem. Sometimes that is with language services, sometimes it is with professional services, sometimes it is with technology, sometimes it is a little bit of everything. The vast majority of Smartling’s revenue comes from technology. As soon as you begin to combine these things together, as soon as you begin to combine a platform with the ability to customize with language services, you get a lot of benefits from this. 

Whether you use an iPhone in your everyday life if you are part of the Apple ecosystem or the Google ecosystem. I would say that people always said, Apple cannot do hardware, they can only do software. Then they rolled out the iPhone and the benefits of having hardware and software operating incredibly well together. I think the same exact thing happens with technology and services. Now people say Apple cannot do services. I would say they are doing a pretty good job with their services. They are the biggest company in the world. We are taking a similar approach that companies are looking for a real end to end solution that involves all of these different aspects. 

Esther: As someone who has been in the industry for at least for 10 plus years, what are some of those differences that you observe in localization, particularly viewed through the lens of technology between 2010 localization and 2021localization?

Jack: Our thesis has remained largely unchanged in the last 10 years. Our thesis was that everything moves to the cloud. Every piece of content in every cloud-based content platform, content management systems, marketing platforms, websites, mobile apps, source code repositories. Every piece of content companies will want to translate and localize into multiple languages. The most important change that we have seen is probably the adoption of MT. We were very bare-ish on MT back in 2010. Our message was super simple back then. It was machine translation – bad, human translation – good. Clearly, there has been a profound step function improvement, massive improvement, largely driven by neural MT. Neural and AI and everything else that has certainly changed our point of view on machine translation.

I would argue that Covid-19 has been the great accelerator. Covid-19 has pushed most businesses forward 10 years. Great businesses are thriving, bad businesses, unfortunately, are going out of business. We are fortunate that we had an exceptional year last year. We had the best year in the company history by a long shot. A big part of this is that Covid-19 has accelerated everything. It is going to accelerate machine translation as well. It is also going to encourage more and more companies to put more content into more languages. I think that thesis has not changed. I would say that machine translation probably is the biggest shift that we have seen since 2010 to 2021.

Florian: We covered a company called lengoo out in Germany. They got $20 million in funding and they were funded also in the thesis of the AI agency. What do you think about that? This thesis that a custom-tailored MT is the single most important differentiator in this space and the most important thing that is driving your competitive edge? Would you agree that in the long term, your MT custom setup is the single most important differentiator? 

Jack: I would not call it the single most important differentiator, but I do agree with the thesis in general. First of all, if you cannot get your content out of different platforms and into some platform where content can be bundled and organized and put through an appropriate workflow to be translated, then it does not matter how good your MT engine is. I would argue that the platform, the connectors, the proxy, whatever you are using to get content out and to deliver that content back is critically important. I would argue that your platform itself has to have the appropriate feature set, it has got to be scalable, it has got to be reliable, it has got to be secure. To do that at an enterprise level, it takes a lot of money. It takes a lot of discipline. It takes a lot of audits. I would say that I agree with the thesis. 

Andreessen famously said an original variant on that, which was “Software is eating the world”. I have got software delivering my television and movies. Software is running my nest thermostat. Tesla is a computer with wheels basically and it gets updates overnight on it. Even the airplane that I still fly for fun now is basically a metal shell with wings that has a huge computer dashboard in front of it. Software is eating the world. I would agree with that thesis, particularly in an agency context, the things where traditionally humans did all the heavy lifting, whether it is old travel agencies, any kind of PR agency, digital marketing agency and translation agency. Definitely, AI or the ability to use a machine learning approach with all of the data that is available that you have, that your customer has, is all going to be critically important. 

Smartling has been quietly developing its own neural machine translation engine for about a year and a half now. We have made incredible progress on this, we are up to maybe a dozen languages right now. A customer will come to us and say, we have this massive corpus of content and data, and we would like you to be able to use your machine engine that is customized to our data to help us to translate more content into more languages, across more different platforms. That is an AI approach to solving this particular problem. Of course, there are still humans involved. We have human translators that are helping with transcreation or whether the MT is not particularly good or supervising the output of the MT and that is an important part of what we are doing. It is an important shift that is happening and the companies that do not embrace it probably are going to get left behind. 

I will say that we are not fans of raw MT. We think it is not a super appropriate use for most enterprise customers and there is not a big market for selling MT as a raw source. We think it is technology, humans, services, platforms, all combined together to solve particular problems. It is one more tool in the tool shed. Do I choose a connector or do I choose a proxy? It depends on whether it is built on top of a commercial, off the shelf, third party platform like Adobe Experience Manager, Sitecore, Contentful, Sanity, Contentstack, whatever it may be. If it is, you have a connector that is ready built, then plug it in. If you do not, how are you going to get the content in and out? A proxy is an incredible tool and we still generate a huge amount of revenue from our proxy, which we think is the best in the industry by a long shot. It is a great way to solve problems for our customers that have a cornucopia of content in a lot of different places. I would say, MT falls in the same category. 

Florian: Are you seeing any demand for MT for website content at all? Generally, on websites do you see any demand for raw MT in consumer-facing content?

Jack: We do, and I would say that our customers that are experimenting with raw MT in the marketplace, they are experimenting with languages that they are thinking about, going into that market or they are doing it for user-generated content. User-generated content is probably the most common place where we see our clients using raw MT. Our customers are large enterprise customers that either have tremendous B2B presence or B to C presence. We are in six of the top 10 software companies in the world. We are in three major pharmaceutical companies, one of which is curing Covid-10. We are in major fashion and luxury brands, two of the top three hotel chains. 90% of the apps on your home screen of your iPhone are  probably Smartling customers. They have a connection with their customer. They are trying to create an emotional appeal. They are trying to sell more things or to give more information. I generally do not think MT is the right way to create that sort of emotional connection, appeal to a customer. If you are translating user-generated content, you are translating comments about a review or you are doing internal training or you are doing a rough first draft, so people can generally understand it. We think it is more about testing and a way of reaching more people, given available budgets.

Esther: With these fast-growing companies, fast-growing customers that you are talking about, say you start working with a promising start-up, then it scales, gets traction super fast. How does Smartling scale along with the customer and what are some of the challenges and opportunities there?

Jack: That is actually where we started out. We started the company by saying we were going to explicitly target fast-growing, venture-backed, potentially soon to be unicorn status companies. Why did we do that? That is my network. I have founded a number of companies. I sold my first company to Apple. I have a good network of other fellow entrepreneurs and CEOs and engineers. We are venture-backed. There are many sister portfolio companies. Frankly, 10 years ago, there just were not that many LSPs that were actively targeting that part of the market. They probably did not have the technology chops to be able to service those kinds of customers. They looked around and said, if I was going to build a translation platform as a Bay Area or New York-based or Berlin-based or wherever company, how would I do that? They would look around and they would find Smartling and say that sounds like us, it has got an API available, they focus on four nines uptime, they focus on security audits, they focus on these kinds of rigorous things in the feature set that they are looking for.

At some point, these companies scale really fast, so we have had to put things in place to make sure that we can support them. Do we have time zone support? Do we have the right third-party security audits? If you are not PCI compliant and HIPAA compliant and SOC 2 compliant and all of the other security compliances that these kinds of companies are looking for, you are probably going to be in trouble. We have had customers literally immediately dump billions of words into our platform. If your platform immediately chokes, that is a big problem. Do we do it perfectly all the time? No, I am sure we have made a million mistakes and I am sure that we could have done things better. I would say that our customer base is filled with companies that have started out as six people in a garage and have moved very quickly into a public, unicorn type of company. Actually, the more interesting challenge for us was when we began moving into hotel groups, airlines, pharmaceutical, life science companies, car companies, and shipping companies. Those companies had a different set of demands and a different set of requirements than probably the startups did. I will admit that was probably a bit of a bigger learning curve for us because they did not look like Smartling, they looked like 50 or 100-year-old companies that were on the stock market and worth $20 billion. 

Florian: How did you go to them? How was the lead generation process? You have got your Global Ready conference coming up. What is the sales approach there and the marketing approach? 

Jack: We have a very deliberate sales process. We do a lot of marketing. We have a terrific marketing team that is very focused on telling our story and telling the story of Smartling and telling the story of our partners. They have done a great job with our website and our branding. If you saw our campaign last year about Move the World with Words and the book that we put out that featured translators, it was beautifully shot by an incredible photographer and she took amazing pictures of translators working and in their home and with their families and where they travel. It turned out lovely. 

We have a fairly large sales team that is focused on helping companies to figure out what they are trying to do. We are trying to pull people forward into the modern way to do localization as a combination of technology and services. We have got a sales team that helps understand particular customer’s needs and then how to help to solve them. You mentioned our Global Ready conference. That is coming up in April. We have been doing it every year, but we did not do it last year for pretty obvious reasons. We had some real doubts or concerns that a virtual conference could be done very well. We attended a lot of conferences in April and May timeframe that I would say were not great. I would say they felt like a day-long webinar. People figured it out and I was going to say that I thought the last SlatorCon conference was terrific. I thought Drift did an amazing conference. As people began to think more about the production value and keeping content shorter and lighter and more entertaining and with great information behind it, these things got better. 

We are going to take a page from your book and a page from the book of the conferences that we like, and as platforms have gotten better, and we are going to do a full virtual one. We are already up to 500 attendees that have registered, which is great. We think we are going to get about 100, maybe even 1500, by the time this thing goes off in April. I have got to tell you the team is putting together some incredible stuff. It will be a combination of live and prerecorded, and so the production value is just going to be through the roof. It is free, register today.

Esther: You also mentioned this campaign, Move the World with Words from 2019. Let us talk a little bit about the translators and perhaps what is the modern way of dealing with translators? Is the trend to pay linguists by the hour more advanced or where are we at with that? 

Jack: It is interesting. Honestly, I do not get too wrapped up in this aspect of it. I know people are always looking for different ways to pay translators and somebody will come up with something. My point of view on this is translators need to be paid fairly and properly for their expertise, for the time and work that they are putting into this. One of the reasons that people try to maybe shift to things like hourly, which at the end of the day paying professionals by the hour is not a terrible idea, is that in an age of transcreation, translation, using translation memory, using machine translation, sometimes it is harder to figure out what is the appropriate level of effort that is being put into this and what is the appropriate way to pay people fairly and professionally.

My view is, you have got to pay translators fairly for the work that they are doing. They are putting food on the table for their families. I do think that most enterprise buyers still want to buy by the word. It is a known quantity. It is something they understand very well. It is what they have done traditionally. I am not super fired up about paying by word, by the hour. I am fired up about paying people fairly for the good work they are doing as professionals. 

Florian: We have had this almost wave in 2020 now with funding going into the TMS space and AI agency space. Why do you think now, why 2020, 2021, what is going on? 

Jack: I would argue that there has been funding in this space for about 10 years. Smartling has raised probably more money than any other company in the industry, for any of the other sort of tech-focused, platform-oriented TMS type of company. Great companies that are growing, that are scaling, that have a unique value proposition, will always be able to attract funding. I am happy for the other entrepreneurs that are able to attract funding and/or sell a portion of their company. It is a hot market right now and a lot of people are intrigued by whether it is a $50 billion industry or a $25 billion industry. That is still a lot of billions and people are excited about it. I think that more technology, more AI, more software probably is going to be important to our industry. It is not super surprising to me that companies are attracting these kinds of funding opportunities and are able to raise money. I am happy for my fellow entrepreneurs. 

Florian: Generally, do you just think there is a macro opponent to this that is totally industry-independent, just a flood of money, low-interest rates, literally hard to get any return on any capital at the moment unless you go a little bit further out the risk curve?

Jack: The US government has been pumping a lot of money into the economy as part of Covid-19 relief. It has been important and had they not done that there probably would have been some significant challenges in the US and abroad. I know every company is doing things like that. Interest rates are literally at a historic low, governments are funding or pumping money into the economy. I think your thesis is right, people are looking for different alternative places to make investments. Venture investments and private equity investments are probably places where people are looking to be able to do this. These are professional investors. They are not foolish, they are not making willy-nilly investments. They are looking for high-quality companies that are growing, that are valuable, that have high margins, that have incredible customer basis. That is why, companies like Smartling and others, have been and continue to get funding going forward. It is a very attractive industry. 

Florian: Machine translations coming around and maybe everything is getting automated, but actually the more tech that is coming in, in a sense, the faster it seems to be growing. There are so many new companies coming up, so it is quite interesting. 

Jack: Covid-19 has probably accelerated this in a massive way. MT is not suddenly making the industry go away. It means new products. Look at how many streaming services, thank goodness they were around since Covid-19. Look at how many new products were launched last year in 2020. Look at the success of Peloton and Apple and other brands that have done a great job in a very challenging year. They are not resting on their laurels. They are saying, how do I spread my product and service to the entire world. That just means more demand. MT does not mean it is going away, and as I have said in this podcast and many others, I am a very big believer in the combination of human and technology. As we talked about, I am a former military pilot. I always use the example of a pilot and autopilot on a typical commercial flight. A pilot has his or her hand on the stick for about six minutes on takeoff and six minutes on landing and the autopilot is flying the rest of the time. That kind of supervision between human and machine is important. 

Florian: One thing that I am not sure what to do with it these days, is all of this NLP stuff that is coming out, GPT3, speech to speech, media, all of these announcements. Are you following this at all? Do you feel that there may be a next breakthrough in natural language generation or is this just a couple of degrees too far removed from day-to-day localization work? 

Jack: There is a huge amount of R&D and commercial applicability to all of those things. I would say that several major sort of megatrends are leading to it. It is the cloud, it is ready to access computer platforms. It is machine learning platforms that are publicly available. At Smartling, we have built on top of TensorFlow and PyTorch and other actively developed machine learning platforms. It is a huge amount of data that is going into this. All of this GPT3 is super interesting. Right now, it is more, look how cool this is, but maybe it has not found a home in sort of commercial applicability yet. 

Microsoft just announced that they had a zero-shot new spelling checker in 100 languages that were designed to help them so that when people type in queries in Bing they can spell check on long-tail languages that they do not have a huge amount of content for. I think they have said 20% improvement, even in the two dozen languages that they were specializing in. That is incredible. Smartling and other companies are absolutely going to build on top of these kinds of ideas and these platforms and the ready availability of data and the availability of GPU’s. We are at a bit of a Renaissance on this, but again, people should not be worried for their jobs. Pilots do not walk around an airport, moping around, worried that the autopilot is going to take over for them. At least not yet, maybe 20 years from now. I am not ready to hop onto an airplane that flies from New York to California without a human up in the front of this thing. Airlines are safer because of all of the technology that is gone into it, today. Similar kinds of things are happening. 

Florian: What are the plans for 2021? Are you guys planning on going back to the office at some point or is it completely open in the air at this point?

Jack: It is open in the air at this point. I do not know what to expect. We moved everybody into a work from home posture in very early March last year. We completely changed all of our goals to protect our people, protect our customers, conserve cash, make sure we are still supporting our partners, all the things that we needed to do late last year. We have been in a work from home posture ever since. I am still paying exorbitant fees in New York and London and Dublin and Kyiv for an empty office right now. People are torn and they both miss going to the office and they crave that human connection. 

I will be honest, I had to figure out how to be an okay CEO through a laptop to a hyper distributed team of over 200 people around the world. That has not been easy and I like management by walking around. I like interacting with people. I like saying, hey, what are you working on? Hey, what is that? What are you doing? What happened with that customer? People do crave the human connection and we get a lot of interesting things done in the office, but people have enjoyed more flexibility around working from home, not spending a couple of hours on the train or the subway or whatever your commute time with this. 

We will probably find a way, maybe a hybrid approach, smaller offices, hoteling, people can come in and grab their box of stuff and move it to a dedicated desk. Maybe people are in the office two or three days a week and they are working remotely two or three days a week. Actually, it is a pretty exciting opportunity, but I do not know yet, and a big part of it depends on vaccines and immunity and my top priority is keeping people safe.

Esther: Have you noticed any changes in productivity or productivity improvements while people are working at home? 

Jack: Not to generalize too much, but I think my engineers have never been more productive. They are in their room, they are coding away by themselves without somebody like me tapping on their shoulder and saying, can I ask you a question about something? They have never been more productive. I think my customer-facing sales, marketing folk thrive on the energy of lots of people and noise and hearing somebody else closing a big deal two desks down and rubbing elbows with the people around them. 

Again, and I say this without too much arrogance because I know that there are a lot of people that struggled last year and a lot of companies that have struggled. We had an exceptional year last year. I do not know that I knew that was going to happen. In March I was very concerned about this virus and I was very concerned about keeping our people safe. We figured out how to operate. We have figured out how to use Google and Atlassian products and Smartling and other cloud-based products to help us to operate. We rolled with the punches. We helped our customers where we could do so. Our customers also figured it out. I am pretty thrilled with how things worked out last year. We had an exceptional year and I feel very lucky and very happy that is the case. 

I will tell you one thing we did that I think has been interesting is I started doing a daily standup with my leadership team literally every single day. Just keep it very tight, about 20 minutes that just focus on, what are the two or three significant things that you were working on between yesterday and today that your colleagues need to know about? We go around the room and then we do a metrics review every single day of the two dozen metrics that are the most important drivers of what is going on with the company. Covid-19 forced that behavior and the team is more welcoming of that behavior when maybe that would have been a pain in the butt when they were pre-Covid-19 and maybe they did not want to do that. It has been a game-changer. 

Little things like this that Covid-19 has helped to accelerate have become important. I think that has been very helpful. I look at 2020, not as a lost year, but as a different year, and as an opportunity to think about what has worked and what does not work. Even though it was incredibly hard and a lot of people had very real challenges in their personal lives, with their families, with sick friends and family, with their business. It was a different year and we have tried to do the things that we needed to do to make sure that not only could we survive, but that we would thrive.



Source link

0