Farfetch Head of Linguistic Services, Alex Katsambas, joins SlatorPod and talks about the online luxury fashion platform’s localization model.
Alex talks about Farfetch’s internal teams, churning out billions of words annually, the importance of sustainability in content messaging, and why return-on-investment for localization cannot be measured immediately.
He shares insights into Farfetch’s approach to content differentiation, and explains why treating various content verticals differently is essential, particularly when thinking about integrating machine translation (MT), which Farfetch began to do in 2020.
Having already rolled out MT for two languages, Alex tackles expanding Farfetch’s use of MT in the coming months, and explains why MT is good news for human experts. He also discusses the challenges and opportunities presented by multilingual SEO, noting how there is more room for specialist LSPs.
First up, Florian and Esther discuss the language industry news of the week. Florian talks about DeepL smashing it in global traffic rankings (including the deluge of web visits from Japan), and adding 13 languages.
Esther shares some takeaways from a TechCrunch feature piece profiling media localization giant Iyuno, with a spotlight on their subtitling workflow, tools, and tech stack, and picks out some key growth stats from an analysis of the Slator 2021 LSPI.
The two also send their congratulations to the recent Juvenes Translatores award winners and to Boostlingo — for being the first interpreting provider to make it into the Zoom App Marketplace.
Subscribe to SlatorPod on YouTube, Apple Podcasts, Spotify, Google Podcasts.
Stream Slator webinars, workshops, and conferences on the Slator Video-on-Demand channel.
Pro Guide: Translation Pricing and Procurement
Transcription
Florian: Tell us a bit more about Farfetch. What does the company do? What does it sell? How does it operate?
Alex: The platform is a luxury marketplace. We are in the fashion sphere so what we do is we connect about 2000 brands and boutiques from around the world so they can sell their luxury fashion through the platform. That is the essence I guess for the large scale that Farfetch operates nowadays, but I think the essence of that is connecting curators and designers and brands and making sure that it is done properly.
Florian: Where is the company headquartered? Where is the bulk of the staff except when they are all at home?
Alex: The company is headquartered in London but in the last five years there has been a global expansion. We have production centers in several locations around the world where we do the photo shooting for items for our product catalog but we also now have very large operations in China. We have a local China office that has local production capabilities as well. We have teams in Hong Kong, Tokyo, Dubai, Moscow. There is a big office in Brazil in Sao Paulo. They have local supply as well that supports all our Latin American operations and of course the US, so from West coast to East coast as well. There are a lot of operations, we are starting within Europe as well in the Netherlands production hub so lots of locations. There is not any specific location that I would say there is more people, everybody is very evenly spread across the globe.
Florian: Tell us a bit more about who is a typical customer of Farfetch and who is a typical partner of Farfetch that sells things on the platform?
Alex: Let us start with the partners. It is such a vague term, as well to say our brand partners, who they are. You have two types. One is either directly working with the brands so whether those are the big luxury fashion brands or they can sell directly their stock on the platform or we also work with boutiques and that is what our main USP is as a business as well. We can sell luxury goods through luxury boutiques from around the world, no matter what their location is so you are not highly dependent on where the brand warehouses are, where the stock is going to come from because immediately a boutique becomes a stock point around the world.
To visualize it a bit, imagine that I am a luxury consumer in South Africa, I want a specific Gucci bag. What is going to happen is, I am going to go on Farfetch, I am going to see that Gucci bag and then what the Farfetch platform is going to do, it is going to figure out where the closest possible location is where we can pick up the bag and we take care of the logistics ourselves. That is what we do because we have the logistics department as well so we collaborate with delivery services from around the world in order to make sure that bag is going to reach the consumer the fastest possible way.
Now, if the closest for that specific item is in Japan, then that bag is going to go from Japan. The logic is that basically, you give access to very curated items or luxury goods from around the world that otherwise, any consumer in the world would not be able to access unless they are based in Japan, in this case. I think that is how the whole system works. Now, all of this is done in scale, it is millions of consumers on the databases that visit the sites, which is obviously attracting brands as well to be our partners.
Florian: You mentioned photo shooting before so how does photo-shoot content play into the model?
Alex: As a business Farfetch is also responsible for the eCommerce experience on farfetch.com. This means that we will do the photo shooting of the items ourselves rather than using the brand imagery unless we have an agreement with the brand that we are going to use their imagery accordingly. Which explains why we have multiple production centers around the world where basically the brands or the boutiques are going to the stock points, to ship those items in advance to the production centers. They are going to go through the production process of building the eCommerce assets, such as stills, life model, et cetera. We are going to collect information, data. What is the composition? What are the washing instructions? Anything that you would need to advertise? A lot of things have legal requirements in terms of if there is hazardous materials or any things might be blocked from traveling from one country to the other without costs.
There is a lot of complexity in that area, which is why we would rather do the collection, bring the items to the production centers, process them internally, collect all the information, do the imagery and then also make sure that we build those items on site. Now, you would imagine that you would have the same bag in five different boutiques around the world. What we are going to do is we are not going to receive that bag from five different locations in the production center. We are just going to make sure that you receive it once and it is probably going to come from the closest possible stock point to photo shoot and that is part of the process, which is essential to reduce costs, logistics and all that stuff.
Florian: Tell us a bit more about your own background? How did you start at Farfetch? What was your career prior to Farfetch?
Alex: Humble beginnings as a project coordinator in a translation agency so it has been the path I always wanted to follow. Language was always a passion and I am actually specialized in translation technologies, that is what my masters was from many years ago. I started as a project coordinator, and then I worked my way up. Previously I was at a marketing agency. I was working the Adidas and the Unilever accounts through that agency, which explains a bit how my progress through Farfetch came about because Adidas follows a very similar model in terms of how they localize their content and they are very big in the eCommerce sphere as well.
Building resources and working through this back in the day with a very unique model of trying to build localization teams internally. A lot of businesses could not see the value of doing something like that. I think the experience was not there and the knowledge was not there in terms of, okay, you can do it better, you can probably do it faster. That is when I personally fell in love with this idea of if a business can, they should be building resources for localization and I think this is the direction. Then Farfetch came about, about three years ago. Farfetch started doing something like that out of necessity because of how much the business grew, being able to be everywhere in the world really and to start finding where the luxury consumer is in this case. How do you make sure you provide the best luxury experience, but also the necessity comes from volumes as well.
One of the biggest challenges is very few suppliers out there and by suppliers, I mean, translation agencies or localization agencies in this case are able to afford really having a customer that has more than 2 billion words a year, doing it efficiently and quickly and flexibly and also potentially with the least cost possible. I think that is a necessity and that is how I came about Farfetch in this case and we have been building since, so we have been doing some brilliant stuff within the teams. We are expanding that model pretty much every year. The teams are growing, we are growing this year as well, consistently every year.
Esther: Let us talk a bit more about the team set up then that you have got internally and also your localization model. What roles do you have internally? Who is responsible for doing what?
Alex: In terms of what the model is, excluding English since there is a very large English content team as well that produces content, it is more than a hundred people across the business that cover all sorts of areas. It all starts from understanding what the content verticals are, as we call them, we split them in verticals because it is the best way of dealing with each one of them separately.
Talking about the model itself there are 14 teams, 14 languages that we produce content for and when you look at the global scale, you are going to have content managers that oversee a certain number of markets. There is a manager that covers the APAC region. There are two managers for Europe because there are a lot of languages. There is someone for Latin America as well and then under that, you are going to have leads for each market. We have content leads and then within each team, we have writers, senior writers, sub-editors, some teams have specialists depending on the region. They also have content assistants and obviously, all of that is supported and kept together by a coordination team so we have project managers and project coordinators that support the team. That makes a very large group of people including, especially the product writers, which is at the moment the vast majority of them are not internal resources. The vast majority of them are external resources that we retain, we train, we obtain directly so we do not go via a third party to find those resources so they are a natural extension of the internal team.
Slator 2020 Language Industry M&A and Funding Report
Esther: About ROI, so the return of investment on localization for Farfetch. How do you look at ROI? What do you think about it in terms of your coverage and localization?
Alex: It is one of the most interesting things for us to explore. Part of the big question is, is localization worth it? Do we need it? It is easier to answer that question when you look at markets like China, Germany or France, these big markets that have a lot of extensive localization capabilities within the country. Everything is localized or at least there is a preference to localize but what happens with smaller markets? Markets that have been traditionally considered that English is enough, and I think that is always been the biggest question.
For us, there has always been a question, is it worth it? Was it worth it when we launched in Denmark, in Sweden, very small populations where the luxury market is relatively small? Did we need it? Was it necessary? That is where we started looking into what could potentially be the return on investment and I think you need to see the bigger picture.
What we do is we have to define what the metrics are. What do we consider as return on investment? Our analysts look into conversion rate, number of visitors and also converted visitors so this is someone who has visited but also buys through the site. That is how they did this analysis where basically we take the data and we have a launch date, and then we go a year in advance and then a year after launch so it is a bit of a long term analysis. It is not like you need to monitor it every day, but what needs to happen is every quarter you need to refresh the data to figure out whether it makes sense or not. We go a year in advance from launch, and then we start comparing the data.
What would have happened if you have a control group where it does not include the market that you are monitoring and you are analyzing the data for, and then you say that market did not get localized content. How would it perform? Probably would follow what the control group would be. Now that control group needs to have a very stable performance and when we say stable performance, it means that the markets that are included in the control group is markets that they either call localization content for years before or is obviously all the English content as well, so something that has been pre-existed for a very long period of time. That is what the control group is.
Then you get the data from both areas and then you start comparing to see, did we see an increase in conversion rate after we launched? Did we see an increase in visitors after we launched and also did we see an increase in converted visitors as well, and that is where it gets super interesting. It is worth mentioning that because everybody is probably going to be wondering how did that go? I have to be honest with you. It was quite scary at the beginning because the data that we had at the beginning was not good. We did not have good performance in those markets to start with. The first few months there was a big drop in those markets.
Now, SEO becomes Danish and it is not English anymore so the search engines start seeing something else with a click of a button and I think that confuses the algorithms. That confuses the search engines and search engines are really important. Performance marketing changes, the strategy needs to change there and that also becomes quite complex for the whole operation to just say, from day one, we started selling like crazy in Denmark, that really did not happen. It takes time, and from a brand perspective the Danish consumer started seeing something in Danish and this is part of the conversations with our Danish leads and the Swedish leads within Farfetch that the local consumer is not used to it. They are not used to having big websites that are trying to break into the market, targeting in the local language and that takes a little bit for everyone to get used to.
That is brand awareness and that builds up and the second half of the year after we launched everything started going up and I think it is just because everything was in place. Outside Farfetch, search engines started getting used to us so obviously, Google had to figure out in this case, is our content good enough? Are we producing good content? And that matters.
Florian: To me, SEO is a key point, right? I did not even think about it, you switch a flick but then you are not going to get immediately recognized so that is super interesting.
Alex: We always keep saying, content is king, this is the idea. This is our strategy and if good content is good content, then the engines will know. We should not underestimate Google for example, or any search engine because it is not just Google, there are different search engines in Russia, different search engines in China and local search engines work better. Consumers might be looking through different platforms. They do not just necessarily use Google but if you take Google as an example, Google becomes smarter and smarter. Google knows what good content is. To the extent that sometimes it is almost scary how they can recognize it, but that ranks you within the market and good content is going to help you rank better. Fresh content, refreshing the content, these kind of key areas goes really beyond translation and localization.
Florian: When you saw that dip, does this then escalate up quite high and you have to go into quarterly meetings and defend your strategy? Or is it that you have been given quite a long leeway and they understand the long term criticality of putting out good content in the local language?
Alex: We did not have to do the first one for sure so it does not go to that extent. I think it would take much longer of low performance to have to go very senior and say, actually we need to make a decision. Let us invest but not overinvest. Let us not grow the teams immediately to 15 people from day one. Maybe let us make this a long-term plan where we keep monitoring, but this needs to probably be a three to five-year plan more than within a year. We need to just have 15 writers for Denmark and 15 writers for Sweden, and I know we keep mentioning those two markets, but there are other markets as well.
We started because you never know how the performance is going to be. We might end up being the number one online platform for clothing in Sweden within the next five years, it is very difficult to predict things like that. I think the business understands that. We launched Mexican Spanish about three years ago and that was one of our smallest markets and Mexico this year is projected to surpass the hundred million dollars as a market itself. That is really cool to be able to see because you could not foresee that three years ago when we were launching, but suddenly you can see that market somehow started ranking. It is all the steps, the SEO side of things since the performance marketing, but it is also good quality content and that really matters because that is security as well. There are also lots of other things. I know we talk about content, but we should not forget the logistics. We should not forget the payments, the security on-site to purchase freely and securely. There is a lot of that as well.
Esther: Taking a look back at Covid-19 and where we are now, do you think anything has changed in terms of eCommerce? What has changed because of Covid-19 or has not? What do you think the impact has been?
Alex: A lot has changed. Online retailers have benefited, obviously immensely. Anyone who has been in the online sphere, anyone who has a good service online as well, they have benefited. At least I know that Farfetch, it was our first year that we actually announced profitability as well while growing up 40 something per cent year on year. That has really pushed a lot of people who could not get out and go into physical shopping, towards online shopping and if you have a good product, then you will attract that consumer. The first half of the year we added about half a million new customers on the platform but I think that was not the biggest change.
The biggest change is how the consumer sees fashion and luxury nowadays. That was the biggest change that Covid-19 has brought. There was a huge shift towards sustainability. We have seen that and felt that within the content we produce. 2020 for us was a year that our biggest focus in terms of content was towards that fear of sustainability, sustainable fashion, zeroing out the carbon emissions through our logistics. There are a lot of those initiatives, but from a content sphere, we started doing analysis in markets that need to know more about sustainable products and that also requires us to research and understand how we speak about sustainability in Germany, is it the same in Mexico, is it the same in China? Adjusting the content and making sure that it really speaks to the local consumers.
We have initiatives that we are just launching in specific markets that require content support in this case, which is pushing our work more into content creation as well, which goes again beyond the translation, localization side of things. I think that was the biggest change. Now from a team perspective, obviously, we have been at home for a year now, and that definitely had an impact in terms of how close the teams are. Part of having an internal localization model is to have everybody in the same space to interact, to meet each other, to have teams that work together and collaborate. All of that we had to compartmentalize, ignore, pretend. We never did that before and just move on, hoping that all of this is going to go away soon and we are going to be able to go back.
Pro Guide: Sales and Marketing for Language Service Providers
Esther: I am interested to hear from your side, as somebody who might get approached by prospective vendors wanting to work with you. How do you deal with vendor inquiries? These kinds of cold emails, cold LinkedIn approaches. What kind of thing do you think works from your perspective? What would you want to see more of from those vendors?
Alex: The cold calling and all that stuff, obviously does not work. Honestly, we do not really get too much of that. Probably because we are not very much out there if that makes sense. Usually in a polite way we can always say that we are not interested. On what we would expect to see more of, I think what the industry needs is more transparency. This is my experience and I am talking as someone who started from a translation agency, moved into somewhere in between, which is marketing with the localization capabilities, and then eventually moved into an in-house mode so I feel that I have seen pretty much all sides and I understand all sides very well, and I understand where they are coming from.
We need transparency and what I mean with transparency is we need to understand the LSPs. The LSPs probably need to understand what is their USP, their unique selling point. What do they have to offer that is different or makes them different to start with? There are some brilliant LSPs out there. I do not think trying to cover every single area in the localization sphere is realistic. I do not think that it creates confidence from someone on my end when you are basically telling me that you can do everything, I think we need more specialization.
LSPs should also probably spend a little bit more time understanding new content requirements, especially in the eCommerce sphere, move away from just translation. Everything moves so fast nowadays, translation is not good enough. This is the reality when you are looking into the content sphere, especially online, just translation is not good enough. The consumer is very tech-savvy nowadays, the consumer knows how to navigate social media. More and more people are multilingual as well, so we need companies that go and specialize in specific areas and there is room for them in the industry for sure. There is room for everyone nowadays. Especially in eCommerce, the balance between eCommerce and physical retail is still 25 to 75 per cent. The online sphere is still relatively small compared to what happens in a physical store. There is room for people to grow careers or LSPs to grow into specific areas and specialize. We are going to need this in the future for sure.
Florian: What about video as a channel for your advertising sales? Are you doing it a lot? Is there a particular challenge to localize it?
Alex: We do not do a lot of localization with video. We do a lot of assets for different campaigns either subtitling or graphic that it is quite extensive at the moment. Obviously, there are a lot of languages and there are a lot of assets and there are a lot of platforms that need to be advertised so with performance marketing or brand teams, we do a lot of those, but they are seasonal usually. Every season we change our assets and imagery, et cetera so obviously a lot of features need to be redone. The team is going through a thousand videos right now actually. Look, it is a very big team and they are very efficient to put it this way. They do a lot of work every day but do not forget that they just do that. This is the key there, and this is why it is so difficult for us to work with an LSP because there is no LSP out there that you can send them a thousand videos and they can do it as quickly and as with a cost-effective way that an internal team is going to do it.
I think it is almost an inevitable approach of building our teams the way that we have built them at the moment. That is obviously the brand assets and all that stuff. There is a lot of work that started in the SEO, in the search engine sphere in terms of branding as well, but that is more managed locally as well so we have not looked properly into, or at least we have not been asked to build subtitles for that video, for example, because we want to show it. I am sure it will happen at some point, it will come.
Slator 2021 Data-for-AI Market Report
Esther: What is your current approach to machine translation and language technologies? How are you seeing those kinds of technologies within localization?
Alex: Absolutely, 2020 has been a year that really pushed us more into that sphere. It is inevitable that unless I have the team that I have now but twice as big, we are not going to be able to manage what is coming up. The planning, new categories, bigger product catalog, more localization and by more localization, you are looking into doing bespoke purchase journeys for different markets around the world, but that requires content to support that. That becomes a chain of content requirements.
If I want to collaborate with a partner in Germany because I want to build a bespoke experience because there is a loyalty program and it is going to bring another X amount of consumers on Farfetch, that immediately becomes a chain of content requirements. It is not just integration with that, it is the UX copies, the user experience, the customer service, the FAQ pages. It is potentially email, newsletters, and all of that immediately becomes a program. Now we ended up having a lot of those around the world, and obviously, we are never going to stop because you need to be as local as possible so that is where we started looking into the machine translation capabilities and AI engines.
We actually have two languages that are supported by machine translation at the moment and we are going to have a few more within this year launched as well. I am going to take you back to the original thing that we said about content verticals, about different content types and how we deal with each one of them. That is where machines and AI engines can support. We will look at the large volumes that we have built service levels now, what is the post-editing effort required? What is the level of quality that I am expecting from brand and marketing content and what is the level of quality that I am expecting from product information, from compositions, from countries of origin? Any content that is more regular, that is almost dynamic, that does not really need to be revisited and looked at by two or three different people.
We also do a three-step localization within the teams anyways so it is not just one or two people, and we separate that and then we activate machine translation on specific areas and we just do not activate it in other areas. That really already shows, I have the numbers from the teams when we did our POC’s with the third party that we are working on. I think we had a 45 per cent increase in efficiencies across product information because the copy is more standardized and that is the way forward. It is all about understanding how you can put the engines in use, in the right use, but at the same time having that critical eye and making sure that it is not a blanket approach. It needs to be approached depending on content, but it all starts from there eventually. What kind of service level do you want in each area of our website in this case? Then we define where we go from there.
I do not think there is any other option, it is an inevitable thing. The businesses who offer the services and are specialized in building engines, I think they are doing a great job and they are needed. I have to mention here we have not reduced any of our existing resources and I think that is the biggest fear. We should be mentioning that as well because there is always the fear of am I going to lose my job? Are we going to need less translators? No, we are always going to need more and more writers and translators.
It is more content in a shorter period of time because time to market just reduces considerably. To be honest with you, what you do is redirect the resources into optimizing content, doing the stuff that we are always thinking. I would love to be able to do that but I do not have the time because I need to do the other thing, and this is what we are doing now so eventually it is a win-win situation. The areas that you always wanted to increase the quality of content, you are actually increasing it by allowing the engines to backfill all the capacity constraints that otherwise you would stretch your teams to be able to cover. It is a balance and you need to have people who understand that they need to build that balance within their models. Otherwise, it can go really badly, really quickly, and I think that is the other end of the spectrum as well.