top of page
Writer's pictureBurley Kawasaki

Unbabel Product/Market Fit Retrospective

Challenge - Re-imagine Product/Market Fit to maintain market differentiation and accelerate business growth and scale, even in light of business and AI technology disruption.


With all the recent hype around ChatGPT, Large Language Models (LLMs) and Generative AI taking the industry by storm, you would think that the ability to dialog with computers using machine learning-powered, conversational AI had just emerged in the tech industry in late 2022. People sometimes forget that chatbots, natural language processing (NLP) and machine translation (MT) have been steadily improving for decades. However, while many of us may casually use Google Translate while travelling, or use any of the modern web browsers for basic translation of web page content for personal reading, the use of MT for complex translation of business content and communications (such as press releases, product catalogs or legal documents) has largely remained in the domain of specialized language service providers (LSPs). LSPs have historically been able to uniquely offer a business-level quality and sophistication of language translation that was unmatched by MT engines.


Unbabel sought to re-invent this model when it first launched the Language Operations (LangOps) Platform. Unbabel provided their customers (primarily multi-national global Enterprises) a unique hybrid of software and services that combined the speed, efficiency, and deep learning of machine translation with the polish of its global community of editors and human translators. Rather than try and fully replace humans, it delivered a blended model using powerful machine learning algorithms combined with an innovative "human in the loop" model that efficiently engaged human translator expertise as part of high-quality translation workflows that adapted to different content types and use cases. This was a unique innovation that caught the attention of the language translation industry and won Unbabel numerous awards, grants, and patents. Unbabel's unique approach to Quality Estimation (QE) - the process of automatically assessing the quality of a machine translation output - continues to rank "best in class" in annual competitions across both commercial software providers and research institutes.


However, it is foolish to assume that one's competitive advantage will necessarily persist in light of the rapid pace of technological advances. As we've observed in prior blogs, Product/Market Fit is not a one-time milestone you can simply achieve and move past. Instead, we often need to repeatedly optimize P/MF to keep pace with changes in the market and/or our business. Unbabel has experienced this first-hand over the past few years in striving to maintain market-leadership even in light of the blinding speed of continued AI technological advances. This blog post is about telling that story…


INITIAL P/MF: Unbabel thrives in Language Translation Market


In the early years at Unbabel saw rapid success in pioneering their hybrid translation model. They had assembled a diverse team of experts in natural language processing (NLP), machine translation (MT) and other AI fields, whose unparalleled research and award-winning breakthroughs set numerous industry standards. Their model allowed them to efficiently power multi-language contact centers for major brands worldwide including Booking.com, Nestle, Panasonic, Patagonia, and UPS. This enables customer service agents to instantly become multilingual so they can effortlessly communicate with any customer, anywhere in the world. Unbabel's hybrid MT and human approach reduces cost, meets surges in contact center demand, supports growth, and ultimately delivers exceptional service.


Unbabel successfully found its initial Product/Market Fit around this customer support use-case. Based upon this initial Product/Market Fit, Unbabel experienced not only significant sales and customer growth, but also achieved accolades as a leader in the translation market by industry analysts.


DISRUPTION: Machine Translation becomes “good enough”


However, as the author Marshall Goldsmith famously observed, “What got you here won’t get you there.” The unique competitive advantage that Unbabel had achieved would need to be re-evaluated as AI advancements continued to mature. In particular, the power and affordability of MT solutions continued to improve and was steadily starting to commoditize the lower-end of the translation market and be “good enough” for many types of simple language translation projects. Customer service solutions also began to offer plug-ins to basic MT models which enabled multi-lingual capabilities for certain segments of the market and simple products. This led customers, and the market more broadly, to look for MT-powered capabilities as part of their existing SaaS products.


This kind of market shift doesn’t happen overnight and was not immediately an existential threat for Unbabel. However, if they were to continue to maintain a position of leadership in the market and demonstrate sales and business results, Unbabel knew they would need to hone their Product/Market Fit. The P/MF that had once enabled the company to grow quite successfully would not continue to be sufficient, nor would it yield a successful return for investors.


P/MF OPTIMIZATION: Unbabel focuses on industry domain expertise and targeted workloads


Tachyon Solutions helped guide Unbabel through a strategic planning process using our Product/Market Fit Diagnostic Workshop and supporting strategic advisory services. This is based upon Tachyon Solution's proven methodology and practices for rapidly evaluating P/MF to identify opportunities and gaps and develop an action plan for an actionable Blueprint and execution playbook that optimizes both Product strategy and GTM alignment to drive accelerated growth.


For Unbabel, the P/MF diagnostic process identified the need to look beyond their traditional customer service domain and “skate to where the puck is going” (to quote the hockey great Wayne Gretsky). This required defining and creating new LangOps Platform that would continue to benefit from the unique high-quality translation capabilities that Unbabel could uniquely provide using their blended software and services model. This would help ensure differentiation in value even considering continued advances in AI-powered solutions.


Second, the P/MF diagnostic recommended re-visiting their prior assumptions around LSPs and viewing them instead as an accelerant to their growth strategy. The LSPs were not simply competitors - but in fact were potential ways to acquire and fill in gaps in new industry domains and skillsets due to their deep subject matter knowledge and expertise in areas that Unbabel needed to acquire. This highlighted the fact that Unbabel's human expertise network needed to grow and expand, not shrink - so that the ML-powered models could be deeply trained and optimized into new vertical domains.

Finally, while the initial purchase of Unbabel's products was likely to stay a traditional Enterprise B2B sale, the ability to expand and grow consumption of Unbabel's platform within a large Enterprise account was ripe for use of Product-Led Growth (PLG) techniques. The use of PLG approach for Unbabel could enable faster customer adoption and expansion in a self-service manner, thereby accelerating adoption and driving higher revenues per account.


OUTCOME: Strategic P/MF Pivot Supports New Growth Strategies


These insights into Unbabel's Product/Market Fit helped guide and support new growth strategies. Since completion of the P/MF workshop, Unbabel pursued a series of targeted LSP acquisitions (Lingo24 and EVS Translations) to accelerate the acquisition of deeper expertise into different vertical domains. By acquiring these providers, and shifting them over to the Unbabel LangOps platform, it allowed Unbabel access to the data and skillsets needed to train its unique ML models for supporting enhanced new and differentiated vertical domains. The acquisitions also simultaneously grew both revenues and margins since services provided via those LSPs could take advantage of the greater efficiency and scale of Unbabel's platform. Unbabel also acquired Bablic, which gained them access to unique software IP in a strategic vertical - namely localization and translation management requirements of websites and web applications. Taken together this targeted M&A activity aligned with their new P/MF accelerated their time to market for more highly differentiated solutions.


Simultaneously, Unbabel continued to invest in their underlying platform to add new Self-Service and Quality Intelligence capabilities to enable greater accessibility to their product. This enabled them to begin pursuing a PLG motion to support expanded adoption of their platform services within larger accounts. The investments in Quality Intelligence also help Unbabel showcase their value proposition as part of a larger AI architecture, by being able to provide high quality translation capabilities that complement but extend the types of scenarios where MT and Large Language Models (like GPT 4) may not fully yet deliver consistent and reliable quality. The quality intelligence that Unbabel provides can confidently predict the accuracy of any translation and identify when highly tuned domain models (or human translation expertise) is needed to deliver the highest-level results.


"With the ever-accelerating pace of AI innovation, you don’t win by simply having the most advanced large language models or the newest algorithm on the block” said Alex Cobb, Chief Product Officer of Unbabel. “Instead, it’s your ability to unlock unique and differentiated value for customers that is fundamental to maintaining a long-term, sustainable competitive advantage. At Unbabel, we believe that the unmatched experience we built up over the past decade in selected vertical domains, plus our unique and patented approach to quality intelligence, gives us the ability not only to meet customer demands but do so with unparalleled quality and user experience. Our hybrid approach leveraging our global community of editors and human translators also allows us an unparalleled way to constantly refine and improve the quality of our LangsOps models and processes with efficiency and scale. This is what allows Unbabel to continue to reinforce our position as a trailblazer in an increasingly competitive market."


This was a great opportunity to help advise and support a highly talented team in a strategic P/MF pivot - surely it won't be their last, but they now have the process and tools necessary to ensure that they can continue to adapt to continued change in a rapidly changing market.


124 views0 comments

Comments


bottom of page