Nestlé Says Translation Costs Must Be Seen in A Wider Context Than Cents per Word

Doing quality translation is far cheaper than developing original content and reaches more users, according to Dr. Germán Basterra, Translation Manager, Nestlé GLOBE, the world’s largest food and beverage company with annual revenues of CHF 90bn (USD 90.7bn).

Speaking at SlatorCon Zurich on December 5, 2017, Basterra discussed a large enterprise buyer’s perspective of the need for translation, building a strong business case around it, using automated tools for certain use cases, and streamlining processes to save on cost.

Advertisement

“In the development of global brands, localization obviously has been there from the beginning. The Milkmaid brand becomes La Lechera in Spain or in Spanish-speaking countries. So I want to address this question of centralization and framing it (translation and localization) within two dimensions — the dimension of quality and of content lifecycle,” he said.

Basterra, who has been with Nestlé GLOBE for the past 15 years, leads the language and translation competency center. A major focus of his work is the integration of language technologies and authoring and training tools within the context of a corporate system landscape.

Watch Dr. Basterra discuss content centralization under the Nestlé GLOBE program with Slator co-founder Andrew Smart.

KitKat By Any Other Name

Home to the world’s most iconic brands from coffee to breakfast cereals to baby food to nutrition products, Nestlé’s requirement for translation is huge.

“They [translation requirements] involve not only packaging, marketing or consumer-facing websites but also general business documentation, internal HR processes, employee training, etc. And all of this for quite different product categories and functional areas,” he said.

“In the development of global brands, localization obviously has been there from the beginning” — Dr. Germán Basterra, Translation Manager, Nestlé GLOBE

The challenge, he added, is both linguistic and technical. “Let’s say you develop a fancy brochure in InDesign with special fonts and then you translate it into Vietnamese but the fonts are not available in that language so the customers are not happy. It has nothing to do with the translation itself,” he explained.

Though translation is a major part of this process, he said the key is centralization and managing it end to end. “Following the guidelines of the latest Nestlé Business Excellence program (NBE), this means simplification, standardization, and sharing,” he said.

A Centralized Approach

Nestlé’s GLOBE program, according to Basterra, is a corporate program for sharing best practices, unifying data and standards management and harmonization of the IT landscape and system architecture. Hence, Nestlé has a single HR system and a single data management system.

Basterra said centralizing translation does not mean having a single language service provider or a single (translation) budget, it means centralizing the process — bringing in translation memory tools, workflows, and machine translation. “It is the integration between organizational and technical centralization which brings the benefits,” he explained.

“If we look at translation in this framework, we (also) need things like a common repository for the content, including versioning and single source, if possible, to avoid all of the dangers and costs caused by unnecessary translation,” he added. “Then we need to be able to manage the terminology cycle from scratch and to do quality assurance at source and then to have a fully integrated set of translation tools.”

Large-scale Machine Translation

Basterra said it is fairly easy to criticize machine translation. However, it has use cases in very large enterprise settings.

“Everyday, Nestlé receives calls, emails, Facebook posts, anything about the products from customers throughout the world. And obviously, this is very important for different departments, for quality, especially, if people report some defects or provide positive feedback,” he said. “If a mother in China comments about baby milk and the actual product has been manufactured perhaps in Germany or in Switzerland, how do these people in the factory get the feedback?”

By collecting customer feedback and categorizing it by country, by category, whether it is positive or negative and sending them to be translated by a machine translation engine, the company’s business users can get at least the gist of this feedback in bilingual form.

“There is no way we can post-edit all that amount of text”

He emphasized that this is straight baseline gist in and no post-editing. “We can’t go more than that because we’ve got several million words a day,” he said. “Obviously, the quality is often acceptable for basic understanding. However, it is not consistent for all languages and not good enough for some. So we know we need to improve that. But there is no way we can post-edit all that amount of text.”

Integrating base machine translation into existing processes is one thing. But asked whether, as a buyer, Nestlé would translate more if unit costs go down, he said definitely they would do more human translation or incorporate a human post-edit step in an advanced MT workflow.

“Definitely we would translate more. It’s like when you go shopping, if you get a 50% discount, you probably buy twice as much,” he said.

Translation for E-Learning

Translation at Nestlé GLOBE goes beyond the requirements of brand packaging, marketing materials or customer feedback. They also need to produce training materials for its 350,000 employees worldwide in multiple languages.

An example would be the e-learning they’ve done for their infant nutrition division where they trained thousands of employees about guidelines from the World Health Organization (WHO).

“Our Infant Nutrition unit has to train thousands of people following compliance requirements. So we had quite a few e-learnings courses and other training materials for translation into a large number of languages. It is very important to have the local review and acceptance,” he said. “Recently, we have introduced an in-context review step so people can see a draft translation of the e-learning and incorporate their comments. It is not even the linguistic part yet, but it helps to improve the overall quality of the translation.”

Beyond Cents Per Word

Basterra was candid enough to admit that there are people inside organizations that think translation is too expensive. “You can change that, you know,” he said. His advice: “Give them the idea of the value that you can bring.”

“Definitely we would translate more. It’s like when you go shopping, if you get a 50% discount, you probably buy twice as much.”

The Nestlé GLOBE executive provided a simple model comparing the costs of development against translation.

Assume, for example, that the cost of producing content for training 40,000 users around the world is EUR 4m (total amount of 1 million words in English). If we translated those documents into five languages at an additional cost of EUR 1m, we could reach 60,000 more users in non-English speaking countries.

“This means that your average cost if you produce your training [materials] only in English is going to be EUR 100 per user. If you translate it into five languages on top of the source, your cost is going to be reduced to EUR 50 per user,” he concluded.

Asked what universities
should focus on to train students for a future career in translation during the panel discussion, Basterra answered: “From the translator’s perspective, post-editing, reviewing, and working with the new technologies.”

For a copy of Dr. Germán Basterra’s presentation, register free of charge for a Slator membership and download a copy here.

Editor’s Note: The article’s title has been updated from its original version to better reflect the presentation’s main line of argument

Eden Estopace

IT journalist and Online Editor at Slator. Loves books, movies, and gadgets; writes for a living, but codes for fun.