After we published our recent Q3 2024 update on market sizing for the language sector, which was also covered in a public webinar, this blog addresses some of the questions we have received from clients, prospects, and investors. This quarterly update for Q3 2024 is noteworthy because it includes our final annual market sizing numbers for 2023, which are based on a representative sample of LSPs and produced only after careful examination of vetted revenue data following the close of the year. The validated and verified results fell below our previous guidance for that year, which we had expected to end essentially flat over 2022, and instead declined by about 4% due to a variety of factors, including inflation, fears of a recession, and uncertainty about AI. After years in which language services seemed to defy gravity, perpetually outgrowing the economy as a whole or even netting gains when the economy went down, these results could have been a surprise to all except for CSA Research clients, who have been warned about these factors since 2021. In this post, we answer the main questions about the latest CSA Research quarterly market sizing update.
1. Is CSA Research Market Sizing Optimistic or Pessimistic?
When we publish our numbers for market sizing, we base them on carefully vetted data from a representative sample of LSPs. We exclude double and triple counting from sales between LSPs, and we statistically balance the results for the demographic makeup of the industry. To this we add economic analyses based on figures from NGOs and guidance from the quarterly confidence surveys we conduct with CEOs of LSPs from our representative sample.
The numbers are neither inherently pessimistic nor optimistic; rather, they represent the actual outcomes reported by the representative sample of LSPs. Between 2022 and 2023, some providers grew and some lost ground, but after we adjust for market share, service offerings, and geographic mix, the total revenue generated from language services decreased in that period.
Our forecast for the future provides a range of forecasts based on broader economic data in order to account for uncertainty about the economy. It also accounts for the confidence LSP CEOs express. However, we do characterize the forecasts in this range as “optimistic,” “pessimistic,” and “balanced.”
- The optimistic forecast assumes that the industry resumes a growth pattern that prevailed prior to 2020 in which it significantly outgrew the economy. In this view, events since 2020 have been temporary aberrations and the language sector will get back on track.
- The pessimistic forecast is one in which the industry grows at the same pace as the rest of the economy. Given past performance, this would define a lower bound for growth. However, for two of the past four years, the industry has fallen below this threshold. Nevertheless, we keep it as a mark for what “treading water” would achieve.
- The balanced forecast attempts to find a median between the two. In it, growth does not reach pre-2020 rates but does exceed overall economic expansion. It assumes the industry finds its footing with AI but assumes that lower unit costs and substitution effects depress growth compared to the optimistic model.
Which of these models will apply? In the end, none of them will be exactly right. Instead, these define a window of likely potential outcomes. The actual numbers will depend on a host of factors, both seen and unseen. For different reasons, 2020 and 2023 both marked years that fell below these ranges, but we anticipate that the final figures for 2024 will show resumed growth within the forecast range.
2. Does CSA Research exclude new AI-driven services in the August Update?
The numbers for our August updated market sizing fully account for AI-driven services because we asked the representative sample of LSPs for their total revenues from language services excluding reselling to LSPs. The “Forecasts do not include emerging NLP services and tech” note in the graphic simply explains that our forward-looking forecasts for 2024 and beyond cannot account for new services that do not yet exist or for technology that may yet emerge. Any attempt to do so would involve unsupported speculation.
In addition, AI in various forms – including neural MT and project management – is so integrated into language services, that it would be a fool’s errand to try to break out a separate figure for the total value of AI: It plays a role almost everywhere.
What we can say is that new AI services did not lead to overall growth in 2023. If they had, the range of LSPs we examined would have reported this growth in their revenue figures and – on average – they did not. Any market size figure that is predicated on increases based on these services contradicts the actual reporting of LSPs in general. Did some LSPs and tech vendors grow based on these services? Absolutely. But their growth did not counteract the overall decline in LSP revenue.
The real question is whether new services will lead to growth in the future. Here, CSA Research is optimistic because of the huge latent demand for translation that exists: This refers to content that remains untranslated – or translated only sporadically via free, online MT – due to cost or logistical issues. If AI enables LSPs to address this huge potential demand, revenues can rise dramatically. But as of today, this is potential, not something that has already led to growth. Predicting this sort of greenfield expansion in revenue is a fundamentally challenging economic problem.
It is vital to note that the present conditions do not determine the fate of LSPs. Some – particularly interpreting-centric ones – did extremely well in 2023. Others found tremendous success with data initiatives. Some even found that their traditional services expanded rapidly as they won large contracts. An analysis based on just the largest LSPs or ones that shared results publicly paints a skewed picture: A comprehensive view of the market only emerges by analyzing a representative sample of the full spectrum of LSPs using a statistical methodology.
3. Do CSA Research results point to the end of professional human translation?
In a word, no. While it is true that traditional human-only services are declining, that does not mean that human workers will disappear. Instead, their role is changing. But this has been continuously true since the mid-1990s. When translation memory first appeared, many workers thought it would replace them, but it led to 25 years of sustained growth in the sector as it made translation more cost effective, quicker, and more consistent. What was once novel quickly became accepted and even “traditional.”
Moving forward fifteen years, a similar dynamic played out as statistical – and later neural – MT led to a new round of concern that AI would replace human translators. Instead, they adopted these tools and MT editing is now part of what we consider “traditional language services.”
The point is not that technologies do not disrupt the language industry: They clearly do. Rather, they lead to a continual redefinition of what language professional do. Translators may no longer labor with a pen and a hard copy dictionary, but they are still there. Over time, LLMs and other disruptive technologies will just become part of a constellation of services that language professionals interact with in various ways. If they can learn to leverage them, they will find that these open up new business opportunities for the sector.
So, no, human translation is not going away. But it is becoming something different to what it was just a few years ago. And we should note that change is coming for interpreters too, even though it has taken a longer time to arrive.
4. Is CSA Research arguing that humans are going to all turn into post-editors?
No. As I discussed in the previous point, human translation is changing. It is becoming an “augmented” activity. CSA Research first proposed (or coined) the term “augmented translation” in 2016 to describe a paradigm in which AI works in tandem with humans to deliver faster, better, and cheaper services that are more responsive to enterprise goals and the needs of consumers. This goes beyond “human in the loop” models – a sort of window dressing for post-editing – which often relegate expert linguists to an essentially janitorial role, sweeping up “bad MT” (quality checking and correction) and cleaning up AI messes. Instead, CSA Research has shifted to describing augmented translation as “human at the core” because, at the end of the day, empowered linguists will be making the decisions, aided by technology.
It is precisely this model, which CSA Research first announced in 2016, that holds the key to the future in which language professionals should be able to earn more and perform more rewarding tasks. Because increasingly competent machines can replace humans for repetitive and low-value tasks at a much lower cost and speed, it frees them up to work on the interesting and complex tasks where they add value. These include content that requires cultural nuance and knowledge, adaptation, or domain expertise to spot and understand errors.
What this means is that even as more and more content is processed by machines, language professionals will have a more important role as they use AI to deliver greater value to their customers. They will go from translating a few thousands of words per day to supervising the translation of tens or hundreds of thousands, all while ensure that those words meet enterprise requirements.
Market Sizing Is a Snapshot in Time
Finally, keep in mind that forecasts represent an estimate based on one point in time. When all is settled for 2024, performance will adjust for changes and pressures we expected in 2023, but it will also include factors we could not foresee. Some of those will promote growth. Others will promote declines and replacement. In the end, the success of the language sector depends on how well it adapts to these forces and focuses on delivering real value and new services to its customers and to the world at large. The pace of change has increased, but the industry will go on, ever changing, as it responds and adapts.
Successful LSPs and Global Content Service Providers (GCSPs) have been collaborating with enterprises to address emerging enterprise needs, and develop new, non-traditional services. The danger in 2024 and 2025 for smaller LSPs and their clients is inertia. The good news is that emerging third-party automation and AI tools will help smaller LSPs overcome barriers and make needed changes to ensure their competitiveness in the Post-Localization Era.