He tells us about XTM’s decision and search to bring on financial backers in 2021, which culminated in XTM’s majority sale to US-based investment firm K1 Investment Management in January 2021.
Bob shares his views on the TMS funding and investment boom in 2020, which he says had little to do with Covid (for the record), and unpacks the landmark shifts in translation management technology over the past two decades — from the advent of the cloud to the integration of neural machine translation (NMT) and AI more broadly.
Bob also talks about milestone developments in XTM’s product, including totally rewriting their translation editor at one stage, and discusses how the company balances out feature requests and customization for enterprise clients against general product enhancements.
First up, Florian and Esther run through the week’s language industry news, kicking off with some key stats from the Slator 2021 Language Service Provider Index (LSPI), which features more than 175 companies on its launch in early March 2021.
The two also talk about the Language Industry Job Index (LIJI), which climbed nearly 10 points in March 2021 to match pre-Covid levels, while Florian discusses the underwhelming consumer reaction to the Apple Translate app.
Stream Slator webinars, workshops, and conferences on the Slator Video-on-Demand channel.
Florian: First, tell us a bit more about XTM, in a nutshell, origin story, trajectory, key client segments.
Bob: Andy Zydroń and I set up XTM in 2002. We started out very small and took a while to ramp up. We quickly set up a development center based in Poznań, Poland and today we have still got over 150 people there. Our development testing, support, marketing, systems admin team are all based out of Poznań. Then we have the rest of our sales team and solution architects dotted around the world in North America, South America, Europe, and Japan. XTM is pure tech. We do not provide any services and we focus on the large enterprise marketplace. XTM is highly scalable. It has been built with the enterprise in mind and that is our target market. As a result, we find that we sell to some very big enterprises, and then they specify XTM for their supply chain, so the LSPs use XTM as well.
SlatorCon Remote December 2022 | Super Early Bird Now $98
A rich online conference which brings together our research and network of industry leaders.
Florian: Before XTM, what was your background? What was your career before?
Bob: I started out in South America working for a textile company as a factory manager in my late 20s and that is a pretty good experience learning about cultures and learning about leadership. From there, I went to Chile and Brazil and came back and set up an Apple dealership. I went down the entrepreneurial route. I did very well in that for a while and also set up an internet service provider called Redneck. We were one of the first in the UK. We sold that business at the top of the Dot-com bubble in 2000.
I then worked for SDL for a few years on one of the early TMS projects called SDL Webflow. Then Andy and I are who I had known for many years, we started talking and he had been working for Xerox and also Ford of Europe and said, I have got this great idea about creating a TMS, a great engine. I said, okay, show it to me, so he explained the idea and then we decided to set up a business to basically write a web-based TMS based on XML from scratch. We have been working hard at it ever since.
Esther: You have come a long way then since setting up the company in 2002. Let us talk about you teaming up with the investment firm K1. It seems like this is the first time you are taking in or bringing in outside capital. Why now?
Bob: We had a couple of very early investors, LSPs who were keen to support us in our early days but this is the first real serious investment. Why now? We have taken the company from zero to a USD 10m turnover. We can see the whole marketplace is really hotting up. There has been a lot of M&A activity. The competition is becoming increasingly strong. We needed to project where we would be in the next five years. We are very ambitious in terms of growth, in terms of our product. We thought it was the right time to bring in some external expertise, bolster our existing management team.
We worked with EY for about two years on their fast growth platform and they helped us enormously in terms of preparing for the investment and getting just the KPIs that you need for a SaaS business. We learned a lot from that, in terms of, MRR, ARR, churn rate, customer lifetime value, all these kinds of things, which are important. Then we approached a number of different private equity firms. We chose K1 because they were able to add the most value to us. They have an operations team based in Los Angeles who have a wide range of expertise. K1 specialize in fast growth, SaaS, businesses, B2B and so their operations team have a huge amount of knowledge there that they can share with us to help us on our next level of growth.
Esther: It sounds like quite a long process, potentially lots of things involved. How distracting was it to go through that process while running the business?
Bob: We tried not to distract the people doing the hard work, the developers and the salespeople and the solution architects and the support team. They were not that involved but for the senior management team, yes, absolutely, we had to, but it is part of the process of growing the business. For me, it was educational and interesting and it has prepared us for the next phase of our growth.
Florian: Why do you think now, especially 2020, was the year for TMS funding. Why now? Why 2019, 2020?
Bob: I do not think it has anything to do with Covid-19, to put it that way, but the whole translation business previously was very people-centric, very reliant on people. It is now becoming far more tech-centric so that in order to provide the service, you have got to have the right technology. As a result, the technology sector is growing rapidly. You have got NMT on one hand, you have got TMS on the other hand, you have got lots of other technologies coming in there. Where there is growth you get companies that are interested in investing, whether through acquisition or as a private equity firm or VC. There is a land grab going on at the moment, there is such fast growth that companies realize that they have to win clients as quickly as possible and to do that, you need the resources to be able to go out there and make an impact.
Esther: Expanding on that, how do you broadly think that the TMS landscape has evolved and changed over the past two decades? What are some of the key shifts?
Bob: Well, obviously when we first started out, the predominant TMS’s we are not web-based and we set out right from day one to ensure XTM was going to be web-based so we wanted a self-provisioning SaaS model. We wanted XTM to be entirely web-based, not something that you have to download files and work offline but we wanted the functionality to be as good as, or better than anything else on the market. That has been a big step forward, obviously, other companies have subsequently come into that space and are doing very well.
Other areas where I have seen big changes are, obviously, machine translation which has come on enormously and the way we look at machine translation is as it is another tool for the translator. Obviously, that changes the work that the translator does. It is not just a question of seeing a translation memory match and accepting it or modifying a fuzzy match. Now it is more post-editing and I think that trend will continue. There will be more and more use of machine translation and AI for automating other things apart from the actual machine translation. We have done a huge amount of work on our proprietary technology that we call Inter-Language Vector Space. That is a cornerstone for our AI strategy so we are using that for alignment of parallel texts, we are using it for Bilingual terminology extraction and we have lots of plans for enhancing our translation environment with it as well, like predictive typing. That is all coming in and we are going to see more and more automation of the processes, so that project managers, for example, only have to get involved for exceptions rather than doing the routine stuff day in, day out.
Esther: You mentioned a couple of things that XTM has been working on. Is there anything that you think looking back in the past four to five years has helped you or has served a particular purpose for clients?
Bob: Maybe four years ago we decided that what we call XTM Editor, our translation environment, which we had started building way back in the early 2000s had reached the end of life so we took the conscious decision to rewrite it from scratch, using the latest technology. We had a lot of input from translators to make sure we got the user experience right and so that took us a long time. It is very complex, with a lot of functionality but we have completed that, it is launched. We have built into what we call the visual mode that allows users to be able to visualize the source and target while they are translating in real-time.
We also have what we call CAR, computer-aided review, which is a translation environment or a review environment, which can be customized to the specific requirements. You can hide certain features because the problem with a fully functional workbench is that it gets very complex and you have got so much going on there that a reviewer thinks, I do not want to use this because it is too complicated. If you can strip it down and simplify it then it makes their whole work experience much better and we have found that is a very key thing for us.
The other thing that we have been working on very hard is our connectors and that is an ongoing project for us. We are building connectors to all the major content management systems, to repositories and it is not just a question of simple integration. We want to have a deep integration with lots of functionality where the user within the content management system can decide which content they want to send, how to send it, when to send it, track the content as it is going through the translations, set, for example, a workflow template within XTM. All from within the content management system and then obviously receive the translated content back automatically to the right place. In some cases, we can do the preview of the web page within XTM so that you can reduce the requirement to do a review cycle within the CMS.
Florian: On the connector side as a non-developer is it primarily a development problem or is it also a contractual problem? Where you need to go through all these AP servers and ask for their permission and then it is a two-year approval process or is it just for people that are not familiar with that process? How hard is it and where is the problem?
Bob: The problems are multiple. First of all, we are not experts in many of the CMS’s out there so we need to find people who are experienced in them and we have done that for many of them. Then you have to start building the connector and whenever you are connecting to systems, there are always issues that you come across in terms of connectivity. They have to be sorted and then you have to bear in mind that the CMS will change versions, it will update, and that may break your connector so you have got to be one step ahead of that. Then of course, while we are selling a lot of XTM, only a percentage of those users will want a connecter so the volume of sales, relatively speaking of a connector, is not as great as XTM itself. That is another factor to bear in mind, but in terms of the improvement in functionality, it is enormous. It just cuts out so much routine work.
Florian: For the cloud, when you started in 2002, there was no AWS. You had to basically just ditch the entire previous CAT component and recode it from scratch. On the cloud and just the infrastructure layer, do you reach a point where you are like, now there is AWS, let us just re-engineer the entire thing, or is it much more of a gradual transition?
Bob: We have the flexibility to deploy in a number of different ways. We have our multi-tenant cloud which is the self-provisioning one that people know, but we also have private cloud and that is a single-tenant solution for the larger enterprise. Then having done that, we can also do single-tenant private cloud on AWS as well. We use a combination. We also use a company called OVH who is one of Europe’s largest hosting companies. We use them for some solutions and AWS for others.
Florian: Tech only versus tech and services, let us go back to this age-old question. You are very firmly on the tech only side. Makes sense, makes a lot of conversations with investors also easier when you do not mix the two, but would you see any kind of scenario where it would make sense for a service provider to pursue a path of developing all of the technology in-house?
Bob: Transperfect has done that, haven’t they? They have done that very successfully. However, from our point of view and what we hear very loud and very clear from our customers is they would prefer to have a vendor-neutral technology so that they can then pick and choose which LSP they use and they can evaluate one against the other. The point is also that the enterprise controls its own assets. They are reducing or mitigating the risk to their assets, to their IP, because essentially the translation memory, all the content is their IP. We feel that a lot of companies want to control that very closely.
Esther: When you are thinking about what is important to clients, how do you prioritize things like feature requests, how do you balance out customization versus the general progress of the product? What takes precedence?
Bob: That is a pretty tough situation I am sure all tech providers face, we have. Our product manager Sarah has something called a product board where she gets all the feedback from our customers and we enter all these things on there. We have internal items that we want to do as well because things need to be updated or we have our own strategic direction that we want to go in. They will get put on the product board and then we have a meeting and we discuss what we can squeeze into because we do four releases a year and how we partition those items into each release.
Right now we are just about to release a new feature on our website based on the product board that will allow our key clients to go in and they will be able to see what we have released, what we have lined up for our next release, and they will be able to see or create requests for new items or vote on existing items that are already there. It is a more interactive way for customers to be able to see what we are doing and participate in the whole roadmap. You have to have a balance but at the same time, we like to stay responsive to our customer’s claims. We would love to be able to sell XTM and show it to the customer and they say, yeah, that is perfect, just what I wanted, thank you. Inevitably they say, yeah, that is great but we wanted to do XYZ and so we are responsive to that and try to help them achieve their goals.
Florian: The dynamics of where they are in the localization maturity level, can you speak a little bit to that? When some clients are far ahead, they have internal localization teams, they have a lot of it figured out as opposed to maybe, companies that are in an earlier stage. Is that a dynamic that is very important to you or it does not really matter in day to day?
Bob: I think there are two different types of customer. The mature customer probably has a TMS already, probably has a translation or localization department with a head. With their existing TMS, they may be coming up against some kind of limitation on that or some issue that they are looking at alternatives. For us, we love those kinds of customers because they know what they want and we can usually help them very well. On the more greenfield site where they do not have a TMS and they are probably less experienced in the whole process, that scenario we are increasingly targeting because we can see that it is a huge area that we have not really focused on a lot in the past, but certainly in our plans for the future. That is an area where we want to tackle more seriously.
Florian: You have some LSP clients, but as you mentioned, the focus really is on the enterprise. What is the dynamics there?
Bob: LSPs know what they want, they know their business so from that respect it is good. There is a bit of a challenge with LSPs in that they have to be able to process anything their customer asks them. Whereas an enterprise generally has a more consistent flow of translation tasks because they have a standard production workflow but having said that, XTM is a very flexible agile tool and so can easily be adapted to all the requirements that an LSP would need.
Esther: What is your view on how machine translation is shaping TMS development generally, and also, what do you perceive to be the natural limits of the integration between MT and TMS if there are any?
Bob: Machine translation is having a profound effect on the whole production cycle. There is far more machine translation going on now than there was last year or the year before so what is happening is that we need to adapt XTM or our TMS to be able to accommodate that. In order to do that we have recently, for example, added the ability to calculate the cost based on edit distance. What has a post editor actually had to do to bring the machine translation up to an acceptable standard? This is the edit distance, hence this is the cost. That change that we have had to make in order to accommodate it. How far will it go? It is a tough question. I am sure we have not reached the limits of machine translation yet. Some of the machine translation suppliers are now using translation memory in real-time to enhance their matches so there are lots of tricks that can still be done to improve it.
Florian: You said the ability to calculate costs based on the edit distance. Is that post or before? How does that work?
Bob: You would have to calculate the cost after the editing has been done, but, historically you would calculate the cost based on the number of words or some kind of algorithm based around the words. Now, when you are principally dealing with post-editing, how many words is not a relevant thing. You can work on time. You can base it on how long the post editor has had to spend on this task, or how many characters have had to be changed within this particular segment of text.
Florian: The industry has gone through this super challenging year. Generally, I think it was a positive year, given the circumstances. What is your outlook for the next three to five years? You have an outside partner now, you can ramp up investing in all kinds of areas so where are you going to deploy some of that capital? What are your priorities in terms of growth?
Bob: From our point of view we see the market continuing to grow. We see XTM in an ideal position, we have got a great product, great team, but at the moment we are just scratching the surface of many markets. We have a very wide cross-section of verticals from tech companies to construction, to retail, through life sciences to LSPs. We have a very wide cross-section of companies, but realistically, there are a hundred more out there where we have one market leader, we could be selling to a hundred. From our point of view, there is great potential. K1 recognized that XTM is a market leader in this respect and they will support us to capitalize on that potential.
Florian: When we talk to people offline, I am sensing a lot of buzz, I am sensing a lot of optimism generally, which is a bit different from a little bit of the MT panic of 2017, 2018.
Bob: There was definite panic thinking this is the end of the translation industry as we know it today. Whereas now you just have to accept that things evolve and MT is another tool and we have got to make the most use of it we can and just provide a better service. The volume of translation just goes up. The better tools you have, the more translation you can do.