Lessons in Localization: SAP

The original lesson by Tucker Johnson was published on November 9, 2020. The present report was updated in June 2023 to reflect the latest developments at SAP.

About SAP

SAP is a German software company specializing in enterprise software that allows to manage business operations and customer relations. SAP is especially known for its ERP software solutions, but offers an entire suite of products to customers around the world. SAP is the largest European software company by revenue and is the world's third-largest publicly-traded software company. Nimdzi talked to Markus Meisl, a member of the management team of SAP's language technology and services department about the company's localization program.

TL;DR

Large internal team: The Language Experience Lab (LX Lab) team consists of over 250 people in 7 locations and rolls up into the User Enablement unit, which is part of global Product Engineering.

Focus on niche vendors: Smaller, in-country single language service providers are used for all languages. There is a strong focus on using vendors that have a very involved senior leadership team and have their primary operations located in the local markets.

Local user groups: In priority markets, there are robust local user groups for SAP products that play a role in providing user experience feedback to the LX Lab team.

Internally trained MT: Based on decades of experience with rules-based and statistical machine translation, LX Lab today manages and trains neural engines in-house, using their large corpus of bilingual data from years of translation memory updates.

Fast figures

Included in this lesson in localization

• Program overview
• Project and process management
• Supply chain
• Quality control
• Technology
• Interesting initiatives

Program overview

Localization, as the world knows the term, is not quite the same as localization at SAP.

From SAP’s early days in the 1970s, they have been localizing. The first customer, Siemens, had requested the product in German and in French to accommodate their teams abroad. Things have not changed since then, and there is still a strong focus on localizing built into the company culture at SAP.

With an exponentially increasing global demand for its products since the early 1990s and numerous acquisitions after 2000, the volumes and complexity of the localization efforts have continuously increased, even through periods where budgets are being cut, meaning that there is always a strong need for the localization team to be advocating for the budget to support the growing volumes. There are literally thousands of products requiring localization, each with different requirements.

The localization team at SAP supports the translation of approximately 10 billion (source) words each year. About 90% of these words are processed via the TMs and MT, so may not have seen a translator more than once.

Culture

As stated above, localization, as the world knows the term, is not quite the same as localization at SAP. In the SAP world, localization has a much broader meaning than translation and related services and includes the adaptation of the company’s software products for local legal requirements and business processes. In other words: Through its Globalization Services unit, SAP delivers country versions of the standard software that is built for the US and German markets.

The Language Experience Lab (LX Lab), which is in charge of all localization-related activities designed to shape the language experience of SAP’s customers, manages a lot of activities and technologies in house, and they have a firm strategy of using single language vendors for each language. This is a model that may not work for other companies, but it fits the needs of SAP regarding cost, quality, and time.

Localization teams are set up somewhat like Scrum teams and there is a strong “Agile Development” culture. The majority of HR line teams within LX Lab consists of 20-40 team members and is structured so that there are people managers who can focus on team management, while individual service owners focus on the services provided and product managers and product owners focus on the products and tools that are developed in the engineering team.

Organization

Within the User Enablement unit, the LX Lab is a specialized team of 280 people responsible for localizing SAP’s software and non-coded assets such as documentation or training or marketing materials into up to 75 languages for the worldwide market. LX Lab focuses on traditional localization services and also develops their own tools, managing machine translation, artificial intelligence, and other initiatives.

There are separate teams to manage various functions required for the overall localization program. The LX Engineering team looks after tooling, automation, and the comprehensive technical infrastructure. The Business Partner Management team manages the supply chain. The various Service teams represent the core of what they do and manage the diverse localization requests.

SAP team locations. Source: SAP

While the localization efforts are centralized into the LX Lab team, team members themselves are spread across time zones in seven separate offices, including Vancouver, Bangalore, Shanghai, Galway, Walldorf, São Leopoldo, and Tokyo. Each location has developed because of a strong need to be local. As SAP grows in each market, the local offices grow accordingly. Over the past few years, there has been a focus on adding technical engineering and development expertise into the team.

Languages

SAP’s flagship ERP product, S/4HANA, is localized into ~40 languages. Other products have a different language scope. In cases of acquisitions made by SAP, some product lines support different languages than existing SAP products, which means the number of languages depends on which product is being localized. Core SAP products are available in seven primary languages, but language support usually goes beyond these.

SAP is very cognizant of the fact that once new languages are added, it is very difficult to stop them, as local customers have come to expect localized versions. Once a language is supported for a given product, it will most likely continue to be supported indefinitely. For this reason, product teams have to be very careful in selecting each language, as they know that once they ship a product in a certain language, they will most likely be shipping it until the end of time (or until the product is retired).

Project and process management

SAP develops business process management software, so it should be no surprise that most of the work being done by the localization team is being done largely on internally developed systems. As many of SAP’s core products such as S/4HANA are primarily based on their proprietary ABAP technology, they continue to use the corresponding translation environment SE63. Additionally, they are using third-party software.

There are dozens of different internal teams that request localization through the LX Lab team. While the team includes support services like engineering and vendor management, they are essentially a centralized project management organization, funneling requests from multiple stakeholders through the tools to dozens of single-language vendors that perform the translation.

Because of the diverse (and large) set of stakeholders, localization managers need to be very organized. Some products ship daily, whereas others ship on a weekly basis. Each product has its own language scope. Each internal team has their own unique way of working, and the project management team needs to adapt to each. Whatever cannot be managed through technology falls on the project management teams.

When it comes to KPIs that measure the success of their work, the team tracks the tried and tested localization KPIs such as being on time. The primary barometer of their team’s work remains making sure the requests that come in get done. At the same time, localization is seen as an enabler for product and sales teams to deliver a stellar user experience — while being only one small cog in the whole supply chain of getting the product into the hands of the customer, their work is no less vital.

Supply chain

A team of vendor managers manages separate single-language service providers for each language. Typically, all work is outsourced to the vendors, with only a few in-house resources for the priority languages German and English (German, since they have a strong German customer base).

SAP has chosen a strategy to only work with single-language vendors with in-country operations. There are a few multiple-language service providers in the supply chain, but these are typically used for specific target languages. When choosing a vendor, an emphasis is placed on local presence (in-country operations) for each market, and strong involvement from senior management of each vendor company.

Highlight: Annual vendor conferences

Each year, there is an annual conference or a series of separate vendor conferences in different markets. These meetings help to build stronger relationships with each vendor and share knowledge and best practices across teams. In addition to these conferences, SAP hosts a regular Supplier Advisory Board, which includes representatives from their supply chain that can advise and provide feedback on any changes happening to the overall localization processes or technology at SAP.

Other considerations when choosing a new vendor are price and feedback from the project managers. SAP uses surveys and feedback channels to collect data on how satisfied the project management team is with each vendor. Generally, feedback is positive. It is rare that they have to replace any vendor, and this is a step that is reserved for cases where the same issues are being reported repeatedly after multiple attempts to improve.

The most frequent challenges they have with the supply chain are with low-volume languages. SAP has an expectation that each agency should have qualified translators, who are familiar with the translation environment and with the individual products. Each vendor has one or more “super users” who are very familiar with the SAP localization technology and processes.

Quality control

SAP expects a lot from the agencies they work with. There are a lot of quality steps required from them before the translator clicks the “back to SAP” button and submits the translations. Since they have been hosting their translation environment for decades, they have had a lot of opportunity over the years to build in checks and analytics so that the central localization managers have full visibility into who is touching what files and making which changes, which allows them to maintain a level of control over the linguistic tasks being performed by the supply chain.

There are internal quality checks, as well. For marketing content, there is an in-country market review process. For software localization, some technical checks may be performed for complex non-Latin languages like Arabic, Hebrew, Japanese, or Hindi on localized builds. The focus, in particular for new products, is on language acceptance tests, which are based on test cases provided by the product team. Testers in the local markets, as well as linguists at the respective localization agencies manage this process and are tasked with fixing all language issues before each shipment. The last two years especially have seen a renewed focus on expanding the language testing operation, both working within existing SAP-developed tools as well as leveraging third-party tech solutions such as rigi.io to harmonize language acceptance procedures.

There are also very strong local user groups in some key countries that report issues. This feedback, combined with feedback from regular users, is constantly reviewed by the localization team, with feedback from primary markets being prioritized.

Technology

SAP was one of the very first Trados customers. Being early adopters, they have been in a position over the years to influence the features of the tool and have become very integrated with their current Trados-based solution. An integration between SE63, SAP’s proprietary translation environment, basically serves as a “wrapper” over Trados to help them integrate the systems that they are using. This unique TMS/CAT solution has helped to increase efficiency internally, though it has added some difficulties finding and training translators who are able to work within this special technology.

The teams at SAP are always looking for new tools and processes that will provide the highest possible level of automation, a high degree of tool integration, and more context to the translators and reviewers during translation.

The LX Lab development team has designed and developed their own middleware, Translation Enablement Workbench, which automates the data exchange between development repositories such as Github, and the LX Lab translation systems.

To further support the above-mentioned objectives, LX Lab has over the last 3 years also partnered with XTM to integrate SAP Translation Hub (STH), including SAP’s in-house MT infrastructure, and other tools and processes with XTM Cloud as the infrastructure for the translation of products developed outside the ABAP environment.

In addition to Trados and XTM, SAP also works in Smartling to handle marketing content. This tool was selected specifically for marketing content because of the in-context features offered.

On top of their TMS/CAT solutions, SAP uses an internally developed terminology solution called simply SAPterm, which is available as a public online resource. These glossaries have been managed and updated for years and currently boast thousands of terms. When evaluating third-party cloud-based systems, they place a strong focus on interoperability, AI features (including MT), and integrations. Out-of-box connectors, access to MT options, in-context review features, and flexibility are all priorities for SAP when choosing a technology.

Machine translation

Internal departments are always expressing a desire for translation into more and more languages, so there is a strong internal demand for machine translation to process the high volumes from the product, marketing, and education teams.

Having been an early adopter of rules-based machine translation, SAP decided several years ago to switch to statistical machine translation and hired a small team of experts. An external vendor supported them with the infrastructure, because they wanted to build momentum around rolling out MT but were not quite ready to manage it in-house right away.

Today, SAP has been working with NMT for several years, which is set up and managed internally. Extensive translation memories were used to help train the engines. After the engines were trained, SAP NMT engine quality was compared to major brand engines and found to be superior for SAP domain translations. While public MT solutions may still provide better output for very general texts, SAP has a very specific terminology and style, which is what necessitates the internal management of the MT engine training.

The decision of whether to provide post-editing or not depends on the use case and the related quality requirements. For some cases, the team is investigating posting raw MT wherever it would not risk potential legal or regulatory issues.

Overall, SAP has a robust language automation program. 90% of the content that flows through the LX Lab team is completely translated by the TMS and MT, never even being touched by a translator.

Interesting initiatives

There is a lot going on at SAP. Here are some interesting initiatives in the works at the time of this publication.

Language platform evaluation

As we already reported, SAP was one of the very first customers of Trados, and they continue to use it today, albeit with plenty of integrations built on top of it. While the current custom-built tool set meets their needs, the LX Lab team is investigating potential additions to their tech stack that bring out-of-box integrations, increased flexibility, options for in-context review, and other beneficial features.

MT is a key focus

Given the sensitive nature of the business SAP operates in, security and confidentiality of data remains primordial. So, when it comes to new tech solutions such as generative AI, until the fundamental concerns related to AI are addressed and SAP as a company has developed a general strategy around the use of Large Language Models in its products, the team continues to invest in related research activities internally and to rely on its ever-expanding MT operation. SAP has been leveraging the power of MT for a while now, and it allows them to address a key requirement for the organization — the need for speed. However, the team is now looking to push the boundaries of MT use cases, seeing how it can be embedded directly in the products customers use.

One of the key integration scenarios realized in 2023 is the one with SAP Enable Now, SAP’s digital adoption platform for providing in-application assistance and enablement content where and when end users need it. Document Translation for SAP Translation Hub allows customers to expedite their content distribution without any manual effort and with minimal effort from the technology side. Users can also access machine translation to translate tutorials and e-learning content on their own.

Special thanks to Markus Meisl, a member of the management team of SAP's language technology and services department for sharing a valuable lesson in localization.

Special thanks to Kerstin Bier, Solution Owner and Product Manager at SAP for sharing a valuable lesson in localization.

This Lesson in Localization was prepared by Nimdzi's Co-founder and Strategic Advisor, Tucker Johnson. If you wish to learn more, reach out to Tucker at [email protected].