The advance of AI as an emerging technology has challenged the fundamentals of the localization industry. Traditional roles no longer address the problems that they once did. Although, I don’t think this is unique to localization, I do believe the field of localization is going through a dramatic shift from traditional working models to more technical ones. Business models of companies are evolving, as they should. Adaptability has become a matter of survival – as it always has been. Today, I’d like to focus on the traditional role of language managers and explain why I think the industry needs a new role; a more modern and hands on one. One that I call a Language Growth Manager.
I’ll use the next few sections to provide background and context, then more specific information at the end of the article related to why you should hire Language Growth Managers (LGMs) per language for your team.
Why traditional models are failing
Traditional localization models are crumbling, right before our eyes. Language Service Providers (LSPs) are scrambling, Execs are screaming AI, linguists are stuck in the middle holding the bag. Yet, simply the arrival of artificial intelligence doesn’t quite seem to explain it. Why is localization changing?
In my opinion, it is because we built an industry around transactions. You know the story, the ‘normal’ model of localization, company needs ‘translations’, finds an LSP vendor, gives them some strings, hopes for the best. Sound familiar? It’s familiar to all of us, but it’s also why we have gotten to this point of asking ourselves why. The traditional model frequently led us through a transactional relationship, rather than a partnership. LSPs were simply go-betweens or brokers managing linguists who were getting scraps, being pushed to an increasingly lower price point. That’s right, the very people who are responsible for your entire international brand voice and user experience, are lowest on the totem pole in those practices.
Why LSPs aren’t ideal for new workflows
LSPs, specifically in the traditional client-vendor relationship, are less effective in an AI world than maybe 10 years ago. There was a time and place when their model worked, and it worked reasonably well. But, it is safe to say, things have changed.
Historically, LSPs were effectively set up to triage, assign and return. They would receive an assignment from the client, assess the needs, then assign it to their pool of linguists. As a broker of services, LSPs make money by charging the client a premium in exchange for managing the linguists. So if an LSP is charging $0.20 per word, the linguist might be getting $0.10 on a good day. In effect, Clients are paying three costs: the LSP, the linguists and typically a computer-assisted translation (CAT) tool. They were known costs, and were accepted as the cost of doing business. This should not be surprising for anyone. The alternative was, simply, not having translations, which wasn’t acceptable for global companies.
Other options have emerged with time. While I wouldn’t advocate for a full replacement (yet), it is commonplace for companies to be questioning whether the ‘old way’ still works. Through emergent technologies, come new workflows and while companies are paying for AI tools in addition to tools, vendors and sometimes freelancers, they are starting to ask themselves: who they should be paying… and for what.
The reality is the old model has become inefficient. It is still effective, just slower and unnecessary (in most cases).
Linguists are already using AI and machine translation anyway
To add to this, it goes without saying, that linguists are already using AI to complete their work. We would be naïve to assume otherwise. In some cases, LSPs are already applying AI work to try to increase their margins too. When done behind closed doors, this practice is both unethical and even downright shady. This is not the partnership that you want for you business.
Is there still a place for LSPs?
In short, yes. I don’t think it is a doomsday scenario for anyone working at an LSP. Furthermore, I’ve had great working relationships with vendors over the years – there’s definitely something to be said for having an outsourced team behind you when you need it. That being said, LSPs which fail to adapt will slowly (or abruptly) find themselves irrelevant.
LSPs still have a place and can fill a number of impactful roles in the relationship with their clients, namely:
- filling language needs which are not priority markets
- LLM training data
- sourcing, it’s hard to find quality candidates in the space
- market intel
- technology stack infrastructure
- even, all in one packages
That being said, those hoping for the traditional & standard ‘per word’ rate interaction with their clients to prevail should likely look into new business opportunities as they may be out of business soon. Whether you need an LSP or not, depends a lot on your stage of company, scale, and budget/resources.
Next, I’m going to walk you through some thoughts to help you better frame language in a growth and product context.
Language is strategy
If you know me, then it would be no surprise to hear that I believe languages are strategic. Saying less, words matter. And they matter a lot. Following a study done by Nimdzi, Project Underwear, “9 out of 10 global users will ignore your product if it’s not in their native language“. Although there may be variance in that statistic situationally, AI is not going to change user expectations for product quality, at least not in my current opinion.
If your organization still thinks of linguists as ‘just translators’ you are missing the boat… and a boatload of growth opportunities for your company to grow. Why you might say? Read on my friend.
UI is not enough
Localization should not be limited to the user interface. That’s only one step in a long list of areas where localization really matters. The impact of localization can be felt from the app stores to customer support and every in between. It’s a full funnel experience, not just a few strings in your app.
To illustrate that, let’s take a look at a traditional funnel (applying localization):
Localization is part of nearly all parts of your product user experience in some way or another. And for that reason, it among other things, should be an integral part of your international product growth strategy. Delivering in localization at your company requires deep linguistic expertise, context-based strategic thinking and finesse – a job fit for – a human.
But… why can’t I just use AI?
You can use AI; and you should. If anything, in this article I’m advocating for the use of AI. Think of AI as tool, made to be leveraged to improve the workflow efficiency of your operators; yet AI is not the operator itself. The next owner operator who will steer your strategy or write your language-specific prompts can be a Language Growth Manager.
Hallucinations, commonplace in the AI world
It’s a widely known fact that large language models (LLMs) hallucinate. While we may not know much LLMs hallucinate on localization assignments yet, we do know it is happening. It just isn’t widely documented yet.
There are several documented research papers that speak to level of hallucinations that different LLMs make. Virtually all models make errors and even print occasionally gibberish. In localization, that can mean misinterpretation of words, misspellings, omissions, out of context translations or even technical issues such as placeholder formatting. While misspellings might be easily overlooked, technical issues can actually break your app or website and prevent it from compiling accurately. That can lead to developmental and sometimes even release slowdowns, which costs the business money.
At some point, you have to ask, is the risk of a hallucination on an important string or piece of content ‘worth’ the amount that you think you are saving by over reliance on AI? Further, if and when there is a hallucination, who is going to fix it? Because there definitely will be. At the very least, you should have a mitigation plan in place.
Well, don’t humans make mistakes too?
Absolutely. In fact, if you scanned over this article, I bet you can find a typo. The difference between me and AI though is when AI hallucinates or worse, is trained on bad data, it doesn’t know. Nor will it proactively go back and fix it, especially if it was working with bad data. As a career professional with extensive domain experience, I’m able to recognize something is off and find the best solution to fix it all while keeping the goals of the task at hand in context. AI doesn’t inherently care about your big picture or company – it cares about responding to a prompt. Although some might disagree, this has been my experience with AI thus far.
Not all languages are data’ed equally
One last point that I want to call out on AI, is the data pipeline. As many of you know already, training data varies greatly depending on the subject matter. Localization is no exception. While languages like French or Spanish might have a plethora of quality historical data from which to train models, languages like Bulgarian, Hindi or Laotian likely do not. Since source training data needs volume in addition to quality to train models, it could be inferred that LLMs are simply just not great for a lot of languages. The reason is not the methodolgy for training LLMs per se, but rather the amount of available data, the quality and relevance of it.
To take this a step further, in talking about localisation, optimisation or even customisation, consider the difference between French (France) compared to French (Canada) and even Portuguese (Portugal) vs Portuguese (Brasil), or even the generational norms and speaking patterns found in each language. If you think a native speaker won’t immediately notice a spelling difference, awkward phrasing, or even mistranslation – think again. LLMs are trained by masses of data and although they can be in theory, they typically are not locale specific nor can they naturally speak to a specific niche group. (By the way, if you noticed the “z” to “s” spelling change in this paragraph, that was intentional and to make a point. Although it might not steer you away, you noticed.)
Model training
Yes, you can train your own models. If you have the resources, data and means, then why not. It’ll potentially save a lot of headache with irrelevance. That being said, I still don’t think it’ll fix all your problems.
Concluding AI
The level at which you implement AI is ultimately up to you. At the end of the day, what really matters most are your goals as a team and company strategy. Is your goal to save money? Perhaps increase speed? If you just need simple straightforward content and want to save money, perhaps AI can do wonders for you. If you are trying to sell a premium user experience and save money – I’d caution you to think again or minimally at least slow down before you make too many decisions.
To those who think AI will replace the need for humans in the localization process, I would challenge you to apply your logic to other positions. Would you replace your current brand marketing team, product team or even engineers with AI? Likely not right? AI can already write a product spec overview for you, but it doesn’t mean product managers are obsolete. AI can already generate code, but it doesn’t mean you should liquidate your dev team.
Even if AI can work well 95% of the time, who is around to fix things when they go wrong for the 5%. Typically, no one is equipped to handle this on the team. Are you going to rely on Jack from customer support who spent a summer abroad in Italy to correct your Italian brand marketing campaign? I certainly wouldn’t; I’m sure Jack is lovely though. Grazie, ma no.
Ultimately if your risk isn’t that great, then it might not matter. But if it does, then it’s something to consider and think over. In my own company, I’d gladly pay a specialist to minimally cover the 5% rather than trying to save a few dollars on translations. That’s an easy LGM hire for me, even as a freelancer.
But… where is the ROI for localization?
As many of us know, ‘the ROI’ question is a never-ending conversation. For good reason, executives are concerned with the return on this investment. Thinking conservatively and even acting discerningly are important characteristics in strategic level thinking. It’s their job. It’s our job as product leads.
To consider the return on investment, I’d like us to think in terms of opportunity and risks (of not doing) rather than strictly in terms of costs or speed. Costs and speed ARE important, but in reality, we need to have a paradigm shift in our thinking towards international product. I’ll cover ROI deeper in another post, but I do want to cover some relevant points in this article.
Opportunity ‘think’
As a company, you have many opportunities every day. Here are some ways to think of localization as an opportunity rather than a traditional cost center or obstacle to moving faster:
| What? | Opportunity it presents |
| Localizing your product into a new language | – make a good first impression, a lasting one – new market growth and discoverability – setting yourself up to be in the right place at the right time |
| ASO localization | – improved conversion rates and clearer value propositions – more conversions, more installs, more viral loops |
| SEO | – free organic traffic when indexed by search engines – market seeding, see where you are popular before spending too much in growth |
| Personalization | – better retention rates and brand loyalty |
Thus, in this level of thinking, I focus on what localization does for you rather than what it doesn’t. Would you rather be available in Thailand for 71M people, or not?
Risk ‘think’
Alternatively, we can consider this from the angle of risks, opportunity costs, and risk mitigation.
What are the opportunity costs of not localizing? What are the risks of not localizing well? The risks to your business can be far reaching, perhaps more so than you think.
A few common risks: the ones you might already be thinking of:
| Risk / Cost | What? | Why it matters? |
| Financial | Mistranslating financial information communicating products or even incorrect payment info | misrepresenting or misinterpreting financial info can lead to detrimental costs to an org |
| Legal, Regulatory or Compliance | Context errors within legal documents | legal matters can be costly, and lead to lengthy legal processes |
| Cultural | Gaffes | can quickly lose trust with users, trust is hard to build |
Some less common risks: the ones you probably didn’t consider
| Risk / Cost | What? | Why it matters? |
| Technical | breaking placeholders, formatting | engineering resources used, breaks app, slows down go-to-market time |
| Strategic/Reputational/Brand | brand perception through digital exposure | poor quality experience can lead to a poor reputation of the brand |
| Revenue | loss of revenue for not being ‘available’ | as a global company, being available in a language means you can be understood. The value can be conveyed and a user can purchase – missing this means missing revenue |
| Growth (or lack there of) | lack of effective localization in key funnel areas | Missing content is missing opportunities. Keywords can be crucial to organic discovery, if you don’t fill the gap, someone else certainly will |
| Loss of competitive advantage | first mover advantage is a crucial gap step | The risk of losing first mover advantage is critical to early startups |
In the business world, mistranslations can lead to business loss. In the medical field, the preciseness of words can be literally mean life or death. Localization is ‘the right thing to do’.
Ying Xinlong (President of the Shanghai Maritime Court), once stated “Although such cases account for less than 5 percent of all cases, the avoidable mistakes brought about significant losses, especially to the Chinese side.” He was referencing Maritime commerce in Shanghai. Why avoid business loss when you don’t have to?
If you think it is expensive to localize, try calculating the cost of reacquiring churned users after they’ve had a subpar user experience with your product. Doing things the right way from the beginning may be uncomfortable and time consuming but it’ll pay off in the long run. Remember, what got you to this point (i.e. A to B) might not be what gets you to the next (i.e. B to D).
Localization costs less than you think
In line with the last few points I’ve made. I would say localization costs, when compared to others are quite modest. This assumes normal volume or work but considering the normal business flows and how much money is filtered into marketing campaigns and ads, the addition of localization is really not that expensive.
When you are young (pre-seed, seed, Series A) frugality is important and a few thousand dollars DO matter. However, Series B and above, it’s much easier to justify the localization spend if you consider it in relevant terms.
Enterprise level localization programs for 50+ languages can be run for just a few hundred thousand dollars a year. Factor in all the other areas of product and marketing that localization impact, the return on investment pays dividends from top of funnel all the way through.
Great localization is also just ‘the right thing to do’
I’d like to conclude this section with a sentiment on accessibility. Language is very much a part of accessibility on the internet, just as much as features might be for the blind. Sure, it’s a different range of accessibility, but they are both addressing a root issue or challenge: access.
Accessibility for the blind helps people to ‘see’ through hearing by making tools to guide and help them read. Tool tips tell them which buttons to press and help them navigate through a digital space. In the same vain, although a user in another language may be able to see, their ability to comprehend what they are seeing may still be limited. In the west, we’ve far too long categorized the world into buckets by what they speak. We also assumed people should just speak the languages we serve them. This rigid categorization has marginalized a lot of communities, cultures and even languages.
Admittedly idealistic, but we shouldn’t be limiting the internet to people who have enough money to buy something. Further, embracing those users today may create new markets tomorrow. As the next 1 Billion continue to join the internet, look for your new customers and meet them on their terms.
“When we work on making our devices accessible by the blind, I don’t consider the bloody ROI.”
– Apple CEO Tim Cook
To bridge these gaps, a Language Growth Manager, is the modern role that your company needs to be successful in localization. Invest in good people, not only tech.
So, what’s a Language Growth Manager and why do I need one?
In aggregation of this entire article, I’ll use this section to summarize what a Language Growth Manager is, and what it’s not. When done right, a Language Growth Manager (LGM) can be an extremely effective team member. One that I’d argue, you won’t regret spending money on.
What is a language growth manager?
To be clear, a language manager is not the owner of your localization program. They are only the owner of their specific language. For example, French Language Growth Manager or Japanese Language Growth Manager – alternatively, Language Growth Manager, Spanish would be a cleaner writing style.
In essence a language growth manager (LGM) is an evolution of a traditional language manager. I frame the LGMs as a ‘CEO of their language’ because I want them to think about their language as a business which they ‘own’ and operate. Akin to a traditional language manager, a language growth manager will create glossaries, style guides and do day-to-day translation work. Where these two positions differ is in the hands on approach. LGMs need to be able to think critically, look for opportunities, spearhead initiatives and be a voice of the customer. Language Managers often lack organizational context. LGMs are a part of the growth and/or product team.
Language Growth Managers are cross-functional in nature and should be able to easily plug and play into any team whether product, marketing or engineering.
Do LGMs already exist?
Yes and no. There are great product localization specialists out there with titles ranging across the board.
What does a typical language growth manager do?
Day to days might include:
- UI localization; a core unchanged role
- …and other forms of content, glossary development and maintenance
- App Store Optimization (ASO), Search engine optimization (SEO), Answer Engine Optimization (AEO)
- Social listening and report generation
- Cultural insights, tone and style direction
- Asset localization (video and static images)
- helping/crafting marketing, ad creatives
- Reviews analysis
- what I call ‘pitch and play’ initiatives. Share ideas about things you think might work, if there is budget for it, then let’s do it!
Why did I choose Language Growth Manager? Couldn’t Language Manager just work?
I chose the term Language Growth Manager for a couple a reasons. In my experience a Language Manager role traditionally is a relatively dead end job for a linguist. They might excel in their work but usually they don’t get meaningful career growth. In becoming a growth manager specified by language, linguists for the first time have the opportunity to learn tangible skills to help them grow in their careers as well as be more effective at their job. Also, being a part of this better incorporates them into the broader team and enables them to be more hands on.
Where does a Language Growth Manager fit in the org
Language growth managers can sit in several positions depending on your team structure.
For companies with traditional international product teams, here’s what I would recommend:
- Director of Product, International
- Product Manager, International
- Language Growth Manager, Italian
- Language Growth Manager, German
- Language Growth Manager, Korean
- Product Manager, International
For companies with in-market teams, you can sometimes put this under those:
- Country Manager, Vietnam
- Marketing Manager, Vietnam
- Language Growth Manager, Vietnam
- Communications Manager, Vietnam
In my opinion, if you have a centralized international product team, they should sit there with a dotted line to in-market teams. The reason for this is that they can keep a foot in each space and easily relay messages back and forth.
Hope that helps – happy to chat if not.
Hiring Language Growth Managers
What to look for in a Language Growth Manager
Foundational background/experience, passion and drive. You do not necessarily need a linguist with 25 years of translation experience, though that could work just fine. Having experience at a ‘big’ company, is nice, but isn’t always as valuable as you might think.
Is an LGM full-time, part-time or freelance
This is an important question. It also depends on the country, your stage of business and profitablilty on a per market basis.
If you are an earlier stage company rolling out new markets a freelancer or part-time works best who can grow with the company. I’d recommend creating some milestones. For example, once the market crosses a specific growth threshold the person can be incorporated into the team on a more permanent basis.
Larger companies with ongoing and pressing needs may benefit most from a full-time hire.
Content Stratification
Different types of content require different levels of scrutiny. Optimize your workflows with an LGM, use their time wisely.
Training for knowledge gaps
Remember, in general, no one person has all of the skills for any given role. If someone gets a 7 or 8 out of 10 on your marks, why not bring them on for the other 2-3 that they might be missing. They’ll gain something from the process by learning and you will win with coaching in the way that best fits the company. Win win.
Wrapping up
If you are still with me to this point, thank you. In the age of quick information, it’s easy to overlook a larger article.
Recent posts
- Language Growth Manager: the role your team needs
- App Store Category: Social Networking
- Why App Store Categories Matter for Global Growth