The cost of quality
INTERVIEW bY CAMERON RASMUSSON
European Institute of Business Administration
FAVORITE PLACE VISITED
Guanacaste, Costa Rica
In 1997, I did a road trip of 3,000+ kilometers from Delhi to Kerala in India. It was an experience for me as an Indian to experience transitioning people, clothing, food, and languages.
Few companies know more about globalization than Amazon. A company doesn’t grow into one of the biggest in the world without looking beyond its founding borders. And Amazon’s localization operations head Mahendra Muralidhar is one of the key players in ensuring Amazon’s goods and services are reaching people in a manner that respects their language, sensibilities, and culture.
We spoke with Muralidhar about the lessons he learned throughout his career — which has taken him from content creation to localization and beyond — and his vision for a future inevitably altered by emerging technologies.
Can you tell us a little about your experiences prior to joining Amazon and how they shaped you as a professional?
My educational background began with a degree in commerce, and then I did my postgraduation in visual communication from one of the top engineering colleges in India. To enhance my creative appetite, I found my love in graphic design. This skill set made me move domains, and I started working with one of the leading media houses in the country, the Times of India Group.
Now, this was ‘97 or ‘98 when the internet was only about creating websites. So, it was an organic move for me to use my graphic design skill to design websites. Soon I was given an opportunity to develop my skills in server infrastructure management. I helped the organization launch India’s first internet portal “Indiatimes.com.”
My job with them was to set up different portals into topics like travel, media, and music and then help them set up infrastructure to host those sites. That kept me excited for a while. But I was always in search of new challenges. I wanted to do something different. So, in 2000, I had an opportunity to work in the UK as a consultant in information technology and hosting services. In 2003, due to personal reasons, I had to move back to India. This was the time when international organizations were outsourcing customer service and back-office operations to India. One such organization was British Telecom (BT), which wanted to move its back-office operations to India through its partner HCL Technologies (HCL). HCL wanted someone who had experience working with UK customers and understood the nuances of business in the UK to transition BT’s operations to India. I led the transition as well as successfully operationalizing three business units for BT in India.
In 2007, I was hired by a leading computer manufacturer to take over their UK and Ireland tech support operations. At this firm, I had an opportunity to work for many teams, including as a principal program manager for one of the senior leaders. I was setting up mechanisms to monitor the health of operations in India, the US, and the UK. This helped me guide the organization’s leadership in making the right decision to optimize operations. One of the key projects that I worked on successfully was to identify opportunities to reduce costs and make customer service operations profitable.
Innovation was a common theme. Whenever I look at a problem, I like to find a solution. In this process I innovate. This was a time when I moved roles to lead the peripherals business for APJ. The challenge at that time was to increase margins on a laptop. The competition was stiff. Reflecting on the past when laptops were sold with telephone cables to connect to the internet, I thought, why not similarly bundle HDMI cables with laptops? This was the time when laptops started including HDMI ports to connect laptops to external high-definition displays. I remember I worked with cable manufacturers and the packaging team to include the HDMI cable inside every laptop box. The project was a success, and I was able to increase margins on the laptop.
Another innovation that I am proud of that changed the peripherals business for laptop manufacturers was hard bundling laptop bags. I started a customer-centric program that soon turned out to be a business success as well. Whenever a customer buys a laptop, they need a bag to carry as well as to physically protect it. I thought, why not ensure that when a customer buys a laptop, they get a bag from day one? I started to hard bundle a laptop bag with each laptop sold in the country. This program started in India as well as Asia Pacific and was a huge success adopted worldwide. Foreseeing the customer adoption and the profit margins this program can generate, other laptop manufacturers soon adopted this strategy.
When you moved into localization work, how did some of those prior experiences prepare you for a very different role?
That’s a great question. So, I think operations help me understand the way to make processes efficient. How can one utilize human expertise to enhance quality but use less human effort to optimize processes? That’s where one should use machines and automation. So that helped me a lot.
Another way that my experience in operations helped me is to embrace diversity. Operations is a constant learning process because when I moved back to Europe, I began working with people from various backgrounds. We had multiple nationalities in my team, and we were spread across several locations. This was fascinating. Let me tell you, one of the most difficult jobs in the world is to lead through others. In other words, how can you create a multiplier effect of the results that you would have achieved yourself, through your team? One must be really attuned to cultures to work effectively with people from various backgrounds and at the same time, understand the ways to engage them to get the best. I like the way I have transformed myself to work with various cultures and different skill sets. It keeps me thrilled.
I think some of the innovations I’ve accomplished are the results of my exploration of industries outside localization. I try to understand the transferable expertise I can bring in. So, when it came to operations, I thought of things like cost optimization — moving operations into different, more cost-effective locations. There were a lot of lessons from my past that I’ve utilized. And that’s what keeps me motivated to innovate in my current role.
With everything that your team is working on right now, what excites you the most?
What excites me is Amazon’s vision to be Earth’s most customer-centric company. The mission is to ensure that Amazon’s products and services for customers are understandable. We need to ensure language is not a barrier to meeting those customers’ needs. We are at the forefront of bridging those barriers. And it never stops. It’s been growing. We’ve been growing the number of languages we’ve been looking at, and with that, the number of challenges thrown at us grows too. In terms of content required and at what quality? So, we’re always looking to innovate the translation process. Now we are at the cusp of where we are considering the next steps. How can we further enhance our service and expertise across Amazon? And that’s something I’m looking forward to.
How has language-industry technology changed from the start of your career to the current day, and how have you and your various teams over the years responded to it?
So, where we started is text-to-text translation done by humans, taking up an enormously long time and possibly done on paper or a typewriter. From there, we went to a word-processing format that is highly manual with limited QA. The translations initially started within industries where compliance was most needed, especially legal, medical, product sales, and so on. We’re moving from that to human-to-human interaction needs or, in other words, basic communication needs. Like for example, even Facebook chats have real-time translation. There’s a need to scale translations because the volumes are growing. When someone asks me, I always give an example about a recent study done in the world of photography. More photographs are being taken every year that surpass the entire history of photographs. The same is for content. We as humans are generating more content every year compared to the history of content creation. Here, I am not talking about just books and documents, but text conversations, social media, and more. To effectively ensure common understanding, translations are needed, more now than before. In terms of technology, we went from standard text translation to statistical machine translation to neural machine translation. And now we’re moving into AI modeling where the machines are learning and suggesting words based on context.
The distance between machine and human translation is narrowing year after year, depending on the content and languages involved. At the same time, tools have evolved. Another key technological improvement that has helped the industry is how the content is moved between demand and supply — that is, the customers and the translator. Translation management systems have evolved. In the past, the content that requires translation was pulled by translators from translation management systems. This was highly administrative. Today there are Uber-like translation management platforms working on content push systems, where the platform pairs the job with a translator based on their language and content expertise. Linguists are not wasting time on administrative tasks and are more focused on producing translations with quality.
This is how the systems have evolved up until now. Looking forward, I see technologies like ChatGPT will help create content from source and may reduce the steps involved from source to translated content — for instance, in cases where source-language content needs to be created and then translated. So, the systems and processes are becoming more efficient, more powerful, and more focused on quality and turn-around time improvement.
We’re now in the middle of another technological revolution within localization with AI. Can you tell us how you’re responding to it in your current role?
I see technologies like ChatGPT are at a very nascent stage. It’s a new and less explored technology. Remember when Apple released the iPad for the first time? People didn’t know what to do with it at first. Over the years, apps were developed for iPad. People began recognizing its value as a professional tool: in the field of medicine, construction, logistics, and so forth. Today, it is a portable tool that is easy to carry around for doing business with efficiency.
So, the analogy for ChatGPT technology is similar. It’s an amazing tool, and I have personally experienced it. I see the potential. But in the localization world today, it’s too early to pinpoint the usage and benefits of this tool. Perhaps this tool can be used in content creation or transcreation, but basic translation still requires translation engines. Perhaps eventually ChatGPT may build internal capability to translate content while creating, and then its capability can be connected via APIs to translation management systems to process content.
Where do you see opportunity within the language industry right now?
I think a big opportunity for evolution in translation technology is in languages that have not been explored. The question we must ask ourselves is what percentage of the human population is not able to communicate with one another due to language barriers. For example, there are so many languages in Africa and dialects that are used as common business languages around the world. The goal of the localization industry should be to eliminate language barriers to communication. Another example: Due to various reasons, we have a mass migration of humans across the globe. Do we really understand the communication pain points these migrants experience when they land in a new location? How do they integrate into the new society? How are they overcoming basic communication barriers to live their lives? If we truly believe in these real-world challenges, then the question industry must ask is: should we focus on building machine translation engines to translate unexplored languages and dialects? Should we improve our language barrier coverage? And most importantly, should we use our localization and technical expertise to give back to society?
Secondly, we need to embrace the fact that machine translation is evolving and the distance to post-editing is reducing. The localization industry should focus on putting more emphasis on training machine translation engines and at the same time invest in enhancing machine learning and AI capabilities.
Thirdly, the speed of translation and the potential for real-time translation can only be done through machines. If one can focus on enhancing the translation delivery system along with machine translation capability to produce native content in real time, this will eliminate the need for post-editing. This may not be possible today in industries like medical or legal where compliance is key. But it can address a large portion of the translation demand today.
Given how sophisticated some of these technologies could potentially become, where do you see humans fitting into the work?
I don’t foresee a scenario where humans will be redundant. The reason I say this is because the content is evolving every day. Take the Oxford Dictionary: They add words every year. And if you look at the evolution of dialects, it’s a similar story. Oxford has also added non-native English words to its dictionary over the years, which are now part of the English language. So, the content itself is evolving because there will always be new content types, terminologies, and associated meanings enhancing a language. So that’s where I see human expertise fitting in. In the future, humans will be spending more time evolving machine translation engines and less time bridging the distance to a high-quality translation.
Are there any areas of the technology conversation you think are overlooked?
Here’s food for thought for the industry: How can we enable and enhance collaboration? Due to a lack of collaboration, we have seen more and more new LSPs start operations, and over the years these LSPs are merging. I see a lot of opportunities for localization experts and LSPs to collaborate. For example, in the localization world, if there is a central term base that can help everyone, it’s not eating into anyone’s business. Why can’t we collaborate using a centralized, open-source method for term bases, translation memory, and more? That way the industry can focus on building better technology for driving efficiencies through TMS or better machine translation engines.
You mentioned the tension that exists between quality versus quantity in localization. What’s your perspective on that?
In the industry today, what is missing is a way to measure the cost of quality. What is the cost we are ready to bear for the quality one is asking for? Like, for example, a customer wants a translation to be done sooner. At the same time, the translation should be at 100% quality, and it needs to be done for multiple complex languages and at a lower cost. This is a fallacy. So I ask: Does the customer know what the right cost is for the translation they are asking for?
So, what the industry needs today is LSPs or localization experts who can guide the customer based on their expertise but proven with statistical analysis to find the right quality for the content they need translated. It’s important to be transparent about what that cost of quality means. Here’s an example: Say I want to do an English-to-French translation. I want it to be done in two days. Yes, it’s possible. And I want the best quality. If you go to any LSP, they’re happy to provide translation as a service but not being a consultant, not being a center of excellence. Here the LSPs should ask their customers what type of content they are trying to translate. LSPs should use their historical expertise to guide customers in what’s right for their content.
If the customer wants you to translate, for example, a website that is going to be seen by a million viewers, then yes, the quality needs to be high as it impacts the customer’s brand. But for internal content meant for employee consumption, why do you need super high quality? An example of an internal-consumption document might be flyers in the office cafeteria or breakout area.
This is why the cost of quality as a measure is important. The cost of quality is like a hockey stick: It’s low or relatively flat until 95% of human quality. But the cost to bridge the last 5% of quality to meet the human level is significantly high. So, I think this is what the industry needs today: defining the cost of quality. I think this is a gap. We should guide the customer on what’s right for them in terms of quality. This may not sound financially viable in the short term, but you have cemented a long-standing relationship based on trust with that customer.
On the topic of adding value to service, what do you think LSPs can do to be more attractive to the buyer market?
That is a great question. Customers today are always looking to optimize costs for their translation and select providers based on who is providing the lowest cost. This is business as usual. Unfortunately, cost optimization always incentivizes extracting the best out of humans at the lowest cost. This can either be a freelancer or an in-house linguist. This means the localization industry demands higher quality from linguists or higher efficiency in terms of words translated by them or renegotiating translation rates. I don’t think this is the best solution.
If LSPs want to provide more value to their customers, the way they could do this is by saying, I’m going to provide benefit on the total cost of translation and not on cost per word. For example, if a customer needs content for their document that needs to be translated and quality-checked, LSPs could tell the customer that their technology is optimized to find the best source for their translations. Or they can tell their customer that their technology (translation engine, translation memory, or term base) is fully optimized to provide a first-time, best-quality translation that eliminates or reduces the need for quality checks. This will help the customer to save costs for their translation. So, your net cost is not focused on squeezing the best out of the linguists, who are already squeezed in today’s economic environment. I understand one person or one LSP cannot bring this change. This is a collective effort, a strong collaboration across LSPs, and linguistic experts can bring a change in the way we think about cost in the industry.