1 month ago
August 20, 2021
Smartcat CEO Ivan Smolnikov on Translation SaaS and Marketplaces
Physics and Technology graduate Ivan discusses his not-so-typical professional background and how he got into the language services and tech industry. Since 2016, he has been focused on building Smartcat into an accessible SaaS platform with a marketplace of suppliers.
The CEO talks about the complex ecosystem that comes with servicing three different client segments — enterprises, LSPs and freelancers — and how Smartcat aligns its interests with them.
Ivan also discusses how they use multi-layered technology to find qualified suppliers and match them with buyers by analyzing terminology and collecting customer feedback. Although buyers increasingly directly edit MT output in Smartcat, he still sees a bright future for specialist and tech-enabled LSPs.
The Pod rounds off with Ivan’s views on raising funds in the language industry, especially throughout the pandemic, as they closed a Series B round worth USD 14.6m.
Stream Slator webinars, workshops, and conferences on the Slator Video-on-Demand channel.
SlatorCon Remote December 2021 | Super Early Bird $105
A rich online conference which brings together our research and network of industry leaders.
Florian: Smartcat is one of the most interesting companies in the language industry. You were originally a founder of an LSP, so tell us a bit more about your background, how you got started in this industry and how you started Smartcat?
Ivan: How I started in the industry is still a mystery to me because I graduated in physics and technology. I was doing hardcore research in physics, began with Samsung in Korea, then Canada and then somehow I found myself building the online translation agency. It was 2003, and at the time we did not have companies such as Gengo or One Hour Translation yet, all these online companies that are today still quite popular. At the time we were building a company like that. It was out of Moscow and just after two years, I sold a portion of that company to ABBYY group and then I continued running that company for another 10 years until 2016.
The name of the company was ABBYY Language Solutions. I owned a fair portion of the company and was running it very independently inside of ABBYY group and grew it into a top 50 LSP by CSA. The company is still around and it was rebranded into Awatera after I left. It is still one of the largest in Europe and in the top 50, maybe 60. I have not been doing anything with that business for the last five years now, so I have a pretty long history in the industry.
Florian: Tell us the origin story, the elevator pitch, how it got started.
Ivan: I always gravitated towards building something on the technology side and my LSP company was very much technology-powered back then. I found myself at some point understanding that you cannot do that much with a traditional LSP approach. That is how Smartcat was born in 2015. The idea was literally to build a new type of SaaS platform that sells. Back then I was thinking about three major problems. Firstly, it was to give a SaaS platform to everyone without charging them for licenses on a per-user basis, because I think it is pretty diminishing for the industry to sell and charge per seat, since we all deal with these freelancers and you never know how many you will need tomorrow.
Secondly, was the idea of combining the SaaS platform with a marketplace of suppliers, both freelancers and agencies. Sourcing freelancers, even back then with my agency, was always tedious because I remember we dealt with 2000 people all around the globe and sourcing them, testing them, verifying them was always a big thing. It was very time consuming, very labor-consuming, and expensive. I felt we could do much better, not just have this catalog of freelancers, but rather this combination of SaaS and marketplace of suppliers. Where we know much more about each supplier than you can read in the CVs because you know their performance, you know how if they deliver on time or not, you know the real feedback from the customers, how the customers are satisfied, and ultimately we can tell you if a supplier is a fit for particular content.
The third problem was dealing with payments with 2000 people all around the globe. I remember we had a team of people managing different payment methods so that freelancers are happy and not disappointed with us sending them wire transfers and them losing 10% of their money. These three problems motivated me to start building Smartcat back in 2015, and in 2016 I set it up and raised my first round, seed money. Then I had another round in the US in 2018 and then the last round almost a year ago.
Florian: Is there any part of your business that is services revenue, where you deliver the service to a client? Or is it just a platform?
Ivan: It is both. We do not deliver service as an agency in the traditional way, but you can buy services from us through our platform. The difference is that you always know who suppliers are. We never hide who suppliers are, regardless if they are agencies or freelancers. As a buyer, and a buyer can also be both agencies and freelancers, you always know who your suppliers are. You literally call them ‘My Team’ inside of our platform and that helps you to signal what you want from them. Usually for the long supply chain, it is a large agency, then a small agency, then freelancers, real producers never get the actual signal from a customer again and cannot communicate back effectively.
In our case, it is all open. It is transparent all the time, and you as a buyer see the rates of all suppliers, and you can see how much you spent and you can optimize that. You can juggle suppliers quickly. It is just a couple of clicks because you do not have a barrier to switch to another supplier. You can start working with a new supplier in a couple of clicks, but at the same time, you accumulate data inside of your account. The transition is smooth and quick. It is definitely a new model for the industry because, on one hand, you have the convenience of a single agreement. You have this convenience of having all the marketplace in front of you, all the suppliers. You can choose them easily and quickly. You can optimize your costs. You can optimize quality. You can train people and they are motivated to really be educated by you because they have this direct relationship.
Florian: Smartcat would not have a project management team that would help facilitate that, correct?
Ivan: Yes. At the same time, there are two components to it. Firstly, LSPs do project management on their own. End customers tend to have someone who is doing some part of project management. At the same time, step-by-step we are automating project management and that is good both for LSPs and end customers because for LSPs, it is the second-largest cost after they pay freelancers and we make project managers three times more productive. For end customers, once they integrate with the platform, all the processes are delivered continuously. They essentially do not do any project management because it is automatic, including the selection and management of suppliers.
Florian: You have three stakeholders here that are working on the platform, the enterprises, the freelance linguists and the language service providers. Tell us a bit more about that dynamic. It must be quite a complex ecosystem because there are some competitive elements in there.
Ivan: Absolutely, and that is both exciting and complicated because we need to go through this complexity all the time. We need to align the interests of stakeholders. We need to optimize products for them. Firstly, If a buyer chooses to work with an agency, the agency is not forced to open the supply chain. They can still work with freelancers and the buyer will not know it. It is a choice for both an agency and a buyer. We see many buyers working with freelancers directly on our platform and that is because we solve the problem of having multiple contracts and multiple invoices. We provide them with the benefits of dealing with freelancers, such as quality and transparency and everything.
At the same time, many buyers still prefer working with both agencies and freelancers, and that is good as well because, for some big complex projects, agencies are more qualified. They have some additional layer of expertise and that is a combination that is liberating for buyers because they can now afford to increase 10 or 20 or sometimes 50 suppliers. Sometimes we see buyers localizing into 20 languages and they have 10 agencies and 60 freelancers working altogether in combination, depending on the language pair and complexity of the project. We do our best to align interests and not to have any competition.
As for agencies, they use us as an all-in-one TMS platform and that is pretty unique. Usually, you need to combine the CAT platform and TMS product, and these two never work efficiently together because of different limitations in integration between traditional business or translation management systems and CAT platforms. If you have this limitation in integration, you cannot build an effective continuous delivery process. You need to transfer data manually, and that is where you have this waste of project management efforts. We combine that in a single platform. We provide them with all the integrations in a single platform. They can use it either for free or they can buy a subscription, which is flat and independent of the number of users, and that is a big benefit for them because they save a lot of money. They can optimize cost on suppliers thanks to our marketplace where we have 450,000 freelancers in different language pairs. Then they can go and buy and sell to buyers their services more effectively because they are now much more effective themselves. Also, we have this partnership program for them where they can resell Smartcat to the buyers and in this case, they do not have technology costs at all.
Florian: You mention subscriptions, how do you manage that? Is there a certain upper limit? Do people max this out consistently? Is it hard to sell translation as a subscription?
Ivan: We do not sell services as a subscription. In our case, it is a pretty sophisticated model, but that is our competitive advantage. Firstly, we have a very powerful freemium component. Any user can use Smartcat for free, freelancers do not have subscriptions at all. For agencies and end customers, we have a very large limit of what we call today Smartwords. It is essentially automatic translation, which we do for you. You can use it or you can avoid it. If you do not use the automatic translation that comes from Smartcat, you can avoid paying a subscription. All other features are available, no limitation on the number of projects or users, you name it.
Ivan: You see the costs transparently on the marketplace and then you can pick them for a particular project, you can combine your supply chain and you can modify it for your goal, essentially. You are not locked in as a buyer on a single supplier for another 12 or 24 months after you have just run the RFP process. That is the beauty of it. You pay for all supplier’s services with a single invoice at the end of the month with a net 30, net 45, depending on how you choose it. This invoice accumulates all supplier costs that you had in the previous month. It can be one supplier or 100 suppliers.
Florian: How hard is that to pull off on the back end? Logistically, how do you do the payments?
Ivan: There are three different products under the hood. Firstly, the TMS, but it is not a standard TMS. It has the complexity of three different target audiences and for all of them, what we are aiming to achieve is an all-in-one platform that solves all your problems. If you are on Smartcat you do not need to use other technologies such as an additional TMS platform for managing your customers or you do not need to use an additional CAT platform. Secondly, is the marketplace itself. The marketplace is not listing off freelancers or agencies. It is not static profiles. It is connected to the TMS platform and the TMS feeds these profiles with real actual performance data. Based on this performance data, we can match you automatically with the best suppliers, not just based on language pair, but on the content domain. When you upload some specific documents we can tell you who are the best suppliers from English into German in this specific domain field.
Another thing is marketplace complexity and matching technology. It is a separate big team inside of Smartcat. Marketplace works for all parties as well. It is for end buyers, agencies and freelancers. Freelancers are only consumers. Agencies are on both sides, they are consuming the marketplace and they are participants of the marketplace themselves, and end buyers are just buying from the marketplace. Thirdly, the FinTech product which is a different product. We can charge you in any currency, in any country with a single invoice. For example, we can give you an invoice in Japanese yen in Japan, and you will pay this invoice for 100 suppliers spread across the globe. These suppliers will go to their personal account in Smartcat, choose the preferred payment method, which will allow them to withdraw money from Smartcat in their currency and with their preferred method in their country. That is an additional big thing and it is also connected with the TMS platform because the TMS feeds it with billing data. Otherwise, you will not be able to accumulate this into a single invoice every month.
Florian: Are you exposed to any currency risks? How does that work?
Ivan: It is not a big deal for us at all because as a supplier you can choose to get money in US dollars, regardless of where you are. If a buyer pays with US dollars, you can mitigate your currency risks on your own. In the case of withdrawing money, usually, it is pretty straightforward. If, as a buyer, you just paid someone in US dollars, then we will convert it into your domestic currency if you prefer to withdraw it this way. Essentially you see this exchange rate at the moment of your withdrawal. We have never experienced any losses or issues with that.
Florian: No “horizontal” platform has ever managed to correct translation, localization. Why do not these large platforms like LinkedIn and Upwork manage a correct translation? What are the key reasons why they have failed so far?
Ivan: The top reason is that the catalog of suppliers does not present value for enterprise buyers because as an enterprise buyer to deal with this variety of suppliers you need the technology around it, which is very powerful and very vertical and specific. If you need a project management tool together with a catalog of freelancers or agencies, first, you need to automate this whole workflow. You need this automatic translation component in it, which we call Smartwords and it combines all the best machine translation engines, translation memory, glossary management, and all QE on top of it. That is the number one core value that any buyer in this industry, including enterprises and agencies, want to see.
Secondly, a catalog of freelancers will not help you if you need to source, select and test these freelancers yourself, so you need technology to integrate with this TMS platform and also help you to source and verify the suppliers. Otherwise, you have to do it yourself and then what is the point? That is what you usually pay an agency to do for you. That is when you have both pros and cons of this standard model. If you do not combine catalog of suppliers with vertical software that automates the entire workflow, plus sourcing, and testing, and payments, then it does not present significant value for enterprises. That is why “horizontal” catalogs of freelancers are not successful in this industry.
Florian: Which ones are the most popular categories that work?
Ivan: Software engineers. It is because the workflows for engineers are so different and as a buyer, you do not need anything, except for just talent to really make it work. In the case of our industry, it is very different. It is close to being successful with our designers.
Florian: You have hundreds of thousands of freelancers. How do you qualify and vet them? What are some of the key issues with scale and production? It is one of the hardest things to do. You find a qualified person, that person then tends to get very busy. You cannot use them anymore for a particular project because they are booked out in something else.
Ivan: Yes, that is exactly the problem we are solving. In both our agencies and end customers, because agencies appreciate that component. They are always in a situation where they have projects that do not have enough freelancers. They are always in a situation where they have a new project for a new customer and they just do not know where to source freelancers for them. In our case, this is multi-layer technology and that is where we apply machine learning or AI. It starts from just basic things, such as language pair and the number of words on our platform in the same domain field. The domain field can be something that the customer identified when uploading the content and freelancer indicated in the profile.
Then rates should be matched. It should be in your budget and you look into customer feedback, which we accumulate as well. After that, we can go into another layer where we can analyze the domain field of the content. If you upload the content, in reality, there is no such domain as purely legal. It is legal for something. If you have a contract for an oil and gas station, that is very different from your contract where you invest money in the company. The vocabulary will be very different as it will be two totally different types of content. This legal domain does not help a lot and that is why we help both enterprises and project managers inside of an agency to understand what is inside of this content and who are the best freelancers who have translated similar content before successfully.
These two things are very important and they are purely technology-based. This is similar content, which means that we can analyze the semantic field of this content and can tell you it is not just about legal, but it is about specific terminology, which prevails in the original content, and here are freelancers who did similar content successfully five times before in this language pair. That is not just based on the customer feedback, it is also based on the number of transactions to this freelancer. If a freelancer retains a customer for 15 months, and this is a reputable customer, that means something. We do not uncover this data because it is confidential as well as the content itself, but based on our technology, we can tell you who is the best candidate for that and that is very valuable.
Florian: If an enterprise client wanted to translate, edit and review everything by themselves, is that something that they could do and how often do you see them trying to do this?
Ivan: With all the experience I have had in the industry it was surprising for me to see how many buyers started editing content themselves over the last two years. That is why we have Smartwords and the difference with just raw MT is, in a nutshell, we combine multiple different MT engines plus translation memory, so you always get complete automatic translation with Smartcat. You can opt not to have it, but by default, you always will have automatic translation. It is a combination of translation memory and MT, and it always learns from how you edit it. It is not exchanged across accounts because of confidentiality, but inside of your account, we always learn from your content and you can upload more content or we can crawl your website and help you to build translation memory from scratch. It depends, but it is always populated with pre-translation and the quality of this automatic translation is always growing.
Surprisingly for me, a lot of enterprises during the last two years found that for many content types, they do not need to outsource it anymore. They just go with the internal reviewers who are the in-country marketing content writers or marketing managers. For some type of content, they just review themselves. It does not mean that they stop outsourcing at all and I believe that they do not even reduce the budget because overall they still want to spend the budget since they have more content than they could afford otherwise. For some portions of content, they can either use automatic translation without any editing or they go with what we call insourcing. They edit the content inside of the company and this market is easy once again because we do not count it. We can allow you to do it internally via a limited number of users. You can send invites to thousands of people in your company and we do not charge you a penny for that. That is convenient and that is different. Secondly, you can invite your volunteers or users. Sometimes we see people invite their users and they do reviews for them for free. You can invite agencies you worked with before. You can invite freelancers you know, or you can go to a marketplace. There are generally four different options with Smartcat.
Florian: As an industry, we should hope the budgets overall do not get cut, but the amount of content that gets localized keeps expanding so much more. The quality for the person consuming the content gets better and better.
Ivan: Absolutely. That is what I am seeing and that is what I hoped for five years ago when I started SmartCat. That is why I believe that agencies and freelancers should not worry. They should not be scared of automation because the total amount of work that is being outsourced is not shrinking. The business world produces more and more content. In my opinion, content becomes the driving force for sales and marketing. I see that sales and marketing converge into content strategy, content management, content delivery, and specifically, it seems like it is accelerated with Covid because you do not need this field presence anymore in different countries, and you can just go with effective content strategies. What it means is that you always have 10x or sometimes 100x more content in your company that you would like to localize than your budget can afford. It is not a problem of shrinking budget, it is about having more content translated in different ways, automatic, a little bit edited by users or your employees, edited with high-quality level by professional linguists and agencies, et cetera.
Florian: When you started, you were not fully focused on the enterprise, correct? Did you transition into the enterprise over the past three, four years more or less?
Ivan: We did not experience any change in our strategy. It is about the problem of platform complexity when you have three distinct target audiences and you develop under the hood three different products for them. It is hard. You just cannot address it all simultaneously and that is why we started with a product just for freelancers. We first got traction with freelancers because the simple CAT product back in 2016 was enough for them and that is how we populated our marketplace. Then we went to agencies and said, ‘we have this nice product and freelancers, and we are building payment components, would you like to try it?’ Then two to three years after 2016 we were focused primarily on agencies. We always had enterprises as our customers, but we did not do any marketing for them. We did not develop specific features for them, and then at the end of 2019, beginning of 2020, that is where we added more resources into both our product development and marketing for enterprises because we felt like both freelancers and agencies products are mature enough to allow us to focus more on the enterprises.
Florian: Selling to the enterprise is very hard, different and complex. How did you find the sales and marketing effort when you are getting into this vast field of the enterprise?
Ivan: Firstly, we are still learning. Honestly, I cannot boast here that we know how to do that at scale super effectively. Secondly, our marketing is not yet super cool and super effective. I think that we have huge room for improvement. LocFromHome was a great idea from our marketing team. I enjoyed it, and I did not even come up with this idea. That was just one month after lockdown. Between 4,000 or 5,000 people registered and we had 2,400 attendees. At least half of them were simultaneously present in a single room which was essentially the largest conference and we have kept growing since then. Still, I believe we have huge room for improvement in our marketing and enterprise sales.
What we do well, that is not marketing and sales, is the product. It is about having a really good quality, easy to use, effective product that anyone can start using for free and that creates a large funnel of users of all types, including enterprises. Also, our product, our design, which does not push you to buy and pay for any user, assumes easier sharing. There are no obstacles for you as an end buyer. You can jump on the platform, translate a single file automatically, then share it with your colleagues quickly, and then invite an agency that you know, and then an agency will invite 100 freelancers or these freelancers then will bring a couple of agencies they know, and so on. I cannot say that they work as effectively for end buyers as they work for freelancers and agencies, but still, it generates quite a lot of end customer acquisitions. We primarily continue focusing and doubling down on this growth product component, where we have a powerful product, a lot of value, easy sharing, and that is something that we believe in quite strongly. At the same time, we do want to strengthen our enterprise sales and marketing in parallel, but that is a work in progress.
Florian: Tell us a little bit about raising from VCs. How are the dynamics post-pandemic? How do the VCs that you speak with perceive the language industry at large?
Ivan: Firstly, I think that raising in the language industry is harder than some other software industries because for many VCs there are still lots of unknowns. How can you create such a value that would allow you, not necessarily in our case, to compete with agencies directly? We are not an AI agency whatsoever, but we provide value to both buyers and suppliers from the single platform and leverage how they essentially transact with each other. On one hand, that is a very appealing and interesting and intriguing strategy for VCs as well, and for me personally, that is why I am pursuing it. I believe it creates lots of value for the world and for both the buy-side and supply-side. On another hand, it is hard, it is complicated and in addition to this complexity, the language industry is very fragmented, and software agencies are all scattered all around the globe. We do not know how many buyers are out there so that makes fundraising harder.
During Covid, I think fundraising was hard for every startup. It is accelerated for some companies, such as FinTech companies for example, but for the majority, it is harder than before. It blossomed again six months ago in the US. About where the money goes, it will not be going to sales and marketing just yet. That is why I am saying there is still big room for improvement. We have 150 people on board today in-house, full-time employees and more than 50% of them are in the product. They are in engineering, product management, data analytics, but sales and marketing teams are very small at the moment.
Florian: In the next two to three years, where do you see Smartcat going? Where do you see the industry at large going?
Ivan: What we see as our major value in where we go is we see that enterprises really can get all technologies in a single platform. Not necessarily all of them are developed by Smartcat, but we accumulate lots of interesting technologies under the hood. We are thinking of integrating some content analysis engines such as GPT-3, et cetera. We see this Smartwords component is something that allows you to process content automatically and then decide if you want to edit this content at all, and that is not necessarily about translation. It can be the content in the original language. We can process speech-to-text for example, and then allow you to edit or not. Or if it can help you to improve your original content with GPT-3 and then process it via machine translation, which seems to be promising for support teams, for example, and that is one of our core components. We have added two more components that we see as valuable for enterprises and they are already incorporated into the platform.
Firstly, is this selective editing workflow. You can either go with your own team or you can invite suppliers you know, or you can source automatically through our large marketplace of agencies and freelancers, and that creates lots of flexibility. Secondly, you can deal with all that via a single contract with one invoice, at the same time benefiting from dealing with all these suppliers directly, because you see their rates, you can communicate with them, but contractual paperwork, invoicing, is removed from your responsibilities. These three components we see as a major benefit for all enterprises, in addition to integrations, IT, compliance, data security, and procurement compliance. That is all incorporated already. We plan to strengthen this automatic language processing component and make sourcing and vetting human resources, including agencies or freelancers, more and more robust so that you do not need to think about who are the best people who can do it for you. You can just click a button and it works.
In terms of the industry, I think that inevitably we will all go into that combination of automatic language processing and human workflow. Another thing is that service providers who do lots of manual work are not very competitive, and they should think about it. This manual work comes from managing files manually with help of the project managers, exchanging documents manually with the customers, then transferring files inside of the internal systems between TMS and CAT, then sending files to freelancers. That consumes a lot of money and time, and that is not competitive today and that will not be competitive tomorrow. I am very confident about it. The best way to go is to automate all of that and to become an expert consultant of selling to your customer, helping them to make the right choices with the content. Choosing what needs to be automatically translated or localized or transformed somehow, what does not need to be transformed at all and what needs to be edited and how that still requires some strategies, some thinking and some help from experts. That is where we enable lots of benefits for agencies as well, so they can resell the platform, becoming our partners and then automate everything on it and become trusted advisors for the customers, seeing how it all goes on autopilot. That is my view, at least for how it is going to evolve in the next couple of years.