Siemens Finds New Value in Employee Surveys Using AWS Machine Learning Services

In today's fast-moving business world, understanding what’s on employees’ minds is important for all companies and crucial for those in the midst of digital transformations. Have new strategies been communicated clearly? Are work/life imbalances or childcare difficulties keeping people from doing their best work? Are front-line employees noticing friction points or opportunities that aren't visible from the top?

At a global enterprise like Siemens—which employs 377,000 people who speak almost 50 different languages—obtaining and making effective use of employee feedback was a challenge. The company experimented with enterprise-wide employee surveys, spending as much as 3 euros per interview, but it struggled to realize all possible value in the results because of language barriers and the volume of responses.

"Most of our headquarters are in either German- or English-speaking countries, and human translation was too costly, so surveys that weren't in those languages weren't much use at upper levels of the organization," says Christoph Malassa, the head of analytics and intelligence solutions at Siemens. "Surveys returned in English or German were at least readable, but some department heads might be responsible for tens of thousands of employees. That's too many responses to read through even once a year—and once a year is not often enough to survey employees given the pace of business these days."

So how did Siemens get to the point where it could launch a quarterly employee surveying program, confident that results would be presented in a timely manner, most language barriers would be overcome, and each interview would cost less than a euro? By building a fully automated solution for survey-response processing and translation based on Amazon Web Services (AWS) technologies including Amazon Translate, a neural machine translation service; Amazon SageMaker, a managed machine learning service; and Amazon Comprehend, a natural-language processing service that finds relationships in text.

"With a solution using AWS machine learning technologies, we went from waiting months for human survey processing to just two weeks' turnaround time," says Malassa. "In addition to seeing results faster, we are also finding much more value because most responses can be translated and then automatically sorted by topic for faster absorption by busy executives."

"By using Amazon Translate and other AWS services, we are getting translation, processing, and analysis for less than one euro per interview."

– Christoph Malassa, Head of Analytics and Intelligence Solutions, Siemens

 

  • About Siemens
  • Benefits
  • AWS Services Used
  • About Siemens
  • Siemens AG is a global electrification, automation, and digitization leader. The company provides solutions for power generation and transmission, medical imaging, laboratory diagnostics, and industrial infrastructure and drive systems.

  • Benefits
    • Reduced employee survey cost by 66%
    • Cut time for survey processing by 75%
    • Automatically organizes responses by topic
    • Runs automated translating with three lines of code
  • AWS Services Used

Simple Survey Processing on AWS

The idea for this solution originated at AWS re:Invent in 2017, when Siemens learned of the imminent release of Amazon Translate. The company obtained preview access and immediately began testing the service. "Because Amazon Translate offers such a flexible API, we found it super-easy to use," says Malassa. "We also liked being able to opt out of letting AWS store our translations. From a data privacy perspective, it's essential that we keep survey responses private at all times."

After also testing the accuracy of the service, Malassa’s team built an AWS survey-response processing solution that sends completed surveys to Amazon Comprehend for language identification and then to Amazon Translate to execute translations. After Amazon Comprehend anonymizes any names, Amazon SageMaker detects and organizes responses into categories and topics. The solution also uses AWS Lambda, a serverless computing platform that runs code in response to events, to orchestrate this workflow. "By using AWS Lambda functions, we are running translations with just three lines of code," says Malassa.

A Powerful Platform Using AWS Machine Learning Tools

In addition to returning analyzed, sorted survey results at least 75 percent faster than before, the AWS solution makes the surveying program much less expensive. "Procuring human processing and analysis of past employee surveys cost multiple euros per interview," says Malassa. "By using Amazon Translate and other AWS services, we are getting translation, processing, and analysis for less than one euro per interview."

The ease of automation and integration with other AWS services, as well as the velocity with which AWS releases new tools and features, makes Siemens confident it can further reduce the time needed for translation and analysis. "In previous survey projects, we've spent at least a hundred hours manually creating neural networks to sort survey responses," says Malassa. "Now, Amazon SageMaker and Amazon Comprehend automate that step. As more and more automation becomes possible, we believe we can achieve one-day turnarounds for translation and analysis. And smaller surveys benefit because the one-time effort for large-scale surveys can be reused without any significant additional costs in smaller projects"

Siemens also appreciates how fast Amazon Translate is growing. "When we started designing this solution, Amazon Translate offered translations from eight languages into English," says Malassa. "The number of languages supported by Amazon Translate for translation into English is now more than 20 and continuing to climb, so our solution will only become more valuable over time without incurring any new costs."

The flexibility and power of AWS machine learning tools will enable Siemens to make the survey experience even simpler for employees. "In the future, we plan to eliminate the different text fields and just let employees enter their various responses into one," says Malassa. "We will use Amazon SageMaker to divide the comments up into topics, without imposing arbitrary restrictions on the front end."

The solution is integrated with Pulse, Siemens's internal analytics reporting platform, making its various capabilities available à la carte to any Siemens employee who needs to build a survey. "With the flexibility of Amazon Translate, Amazon SageMaker, and the other AWS tools we used, anyone at Siemens can now choose any or all of the translation, anonymization, and analysis features we built for this survey," says Malassa. "We're excited to see what kind of value other departments will discover by using AWS for fast and effective translation, processing, and analysis for other surveys of both internal and customer audiences."