Lorem ipsum dolor sit amet, consectetur adipiscing elit. Vivamus convallis sem tellus, vitae egestas felis vestibule ut.

Error message details.

Reuse Permissions

Request permission to republish or redistribute SHRM content and materials.

Employee Sentiment Analysis Shows HR All the Feels

A person pointing at a tablet with a smiley face on it.

​True employee engagement is a mystery to most organizations. One of the main reasons is companies' traditional engagement-measuring tools—employee surveys and feedback channels reviewed manually—are inadequate for getting a full picture of how employees really feel.

Cue sentiment analysis technology powered by natural language processing and machine learning. It interprets huge amounts of feedback to uncover organizational strengths and weaknesses and detect positive and negative feelings toward a new policy, a change in benefits, management practices or workplace culture.

HR can then make decisions and changes, based on the feedback, to address employees' concerns, which in turn encourages more open communication and higher engagement.

"We know how to analyze traditional human capital data, but we've been missing the other side, which is what I would call qualitative insights," said Armen Berjikly, senior director of growth strategy at Weston, Fla.-based Ultimate Software. He was the founder and CEO of Kanjoya, a leading software-as-a-service company that developed natural language processing capabilities to measure and improve employee experience. Kanjoya was acquired by Ultimate Software in 2016 and rebranded as UltiPro Perception.

"There's far more going on behind what can be learned from traditional quantitative metrics, and it's usually going on in the realm of 'what are people feeling?' and 'why are they feeling that way?' " Berjikly noted. "The goal of sentiment analysis technology is to have a machine try to understand what someone feels and why they feel that way."

[SHRM members-only toolkit: Managing Employee Surveys]

Smarter Surveying

Level Access, a Vienna, Va.-based company helping organizations ensure that technology is accessible to people with disabilities, uses UltiPro Perception to learn about its 200 employees quickly, without the human bias that comes from manual review.

Before the company implemented the tool, employee surveys were inconsistent, making it difficult for Chief People Officer Colleen Wood to gauge sentiment trends with any surety. 

The company had conducted exit surveys for years but hadn't used standard questions, she said. And general employee surveys were given at random times, with different questions, so there was no way to truly compare how things were changing.

Since Level Access was already using Ultimate's HR information system, adopting UltiPro Perception was a natural fit. Easily integrating the survey tool with the company's employee demographics and organizational information makes slicing and dicing the data that much easier, Wood said.

She fed 20 years of exit survey data into the tool, allowing her to study long-term trends. She also can now capture feedback from employees with short, routinely timed pulse surveys asking about relevant topics. "If it's goal-setting time, we might ask about goals," Wood said, "or we will ask employees to rate a recent all-hands meeting."

How It Works

The discipline of sentiment analysis based on language has been around for a long time, said Holger Mueller, vice president and principal analyst at Constellation Research, a technology research and advisory firm based in Cupertino, Calif. But the application of digital data analytics technology to sentiment is more recent.

Lisa Abbott, vice president of marketing at Wootric, a feedback management software company in San Francisco, said engagement surveys that solicit open-ended feedback return voluminous amounts of sentiment. "It quickly becomes impossible to read, let alone act on, every piece of open-ended feedback," she said.

In addition, there's always the potential to introduce bias into an evaluation with manual review.

Before implementing sentiment analysis technology, conclusions from a survey review at Level Access essentially came down to Wood's "gut feeling about what I was reading," she said.

A sentiment analytics platform uses machine learning to recognize emotion in the text, Abbott said. "As themes are recognized, each employee comment from the various surveys can be tagged with the relevant theme and sentiment—for example, positive, negative or neutral. The algorithms do the grunt work, reading the qualitative feedback and organizing comments into different buckets with respective tags."

While the technology helps HR managers understand general worker sentiment, Berjikly said Ultimate's tool also allows HR to view employees' emotions with more granularity.

"Previously, both confusion and concern would have fallen under the negative umbrella, but that doesn't help me improve employee engagement or job satisfaction," he said. "Even after drilling down and seeing that employees are sad or excited about something is not enough. You want to find out why they feel that way in order to solve business problems. Sentiment analytics can tie employees' confusion to their benefits, or their concern to pay equity."

But sentiment analysis will only be successful if enough employees submit enough text to be studied, Mueller said. "Technology has made the inputs easier, but employees don't really want to write long sections in response to surveys or performance reviews. People have survey fatigue. Sometimes the employer forces it and gets garbage in return, which brings into question the data and the validity of the analysis."

Best Practices

Here are some practical takeaways for HR professionals who are using sentiment analysis technology:

Survey genuine issues. "Employees very quickly find out if only … routine topics are being surveyed," Mueller said. "It's important that survey designers are in contact with operational teams that know what the true challenges, opportunities and hot topics are among employees."

Field short surveys less often. Surveys should not take more than five to seven minutes to complete, Mueller said, adding that employees will be quickly turned off by a tool that requires too much of their time. "Finding the right balance between frequency and necessity is crucial for the success of any survey tool."

Show the connection. Wood said employees at Level Access are thrilled to have a way to share feedback and receive the results of that feedback. "We close the feedback loop by outlining 'here's what we asked, here's what you said, and here's what we are going to do about it.' If you share the results and what you will do to respond to employees' concerns, you are able to drive trust and encourage future survey responses."

Commit to privacy. Employee privacy is critical. "Anything less is a dead end for a survey tool," Mueller said. "Ensuring privacy and anonymity of the survey tool is crucial for the success of its implementation because the right balance is the only way for employees to keep taking surveys and maintaining trust."


​An organization run by AI is not a futuristic concept. Such technology is already a part of many workplaces and will continue to shape the labor market and HR. Here's how employers and employees can successfully manage generative AI and other AI-powered systems.