By Gloria Lombardi

am“The important thing is to understand what causes an engaged employee to become disengaged, and to focus action on reducing the rate this happens.”

Andrew Marritt (pictured right) is the founder of OrganizationView, a prominent people analytics firm based in Switzerland.

He is also the creator of Workometry, a tool for capturing and understanding employee feedback. This technology uses machine learning techniques to gather and make sense of large volumes of employees’ comments in real-time.

In fact, the tool uses computational linguistics, a computer technique that identifies what employees talk about and makes sense of thousands of texts in different languages. Indeed, it has all the benefits that you would expect from a modern tool, such as the responsive design for the mobile age.

A tool for senior executives

Originally, Marritt’s clients were using Workometry to run regular pulse surveys. But, in fact, it has been as valuable for letting senior executives capture and understand important strategic questions about the business. analytics-925379_960_720

For example, a large retail organisation in the UK wanted to ask 60,000 employees working in the branches what the company could do to improve the customer experience. The retailer was going through a big strategic initiative as they were trying to reposition the business. They went live with a series of questions at the beginning of January. They had the survey period open for a week. The tool provided the break-down of information almost instantly.

In fact, the following day, the executive team had a detailed review from Marritt, which spotted patterns within the business. Workometry found out that many employees were engaged in the transformation. “There were some good examples of employees going significantly out of their way to help customers to a degree of things such as people driving certain customers back home at the end of their shift,” says Marritt.

But, the tool also revealed areas for improvement such as the need to invest in more staff. All of this was discussed during an important board meeting only one week after the survey was taken.


Other companies go to Marritt with their own data and ask for help. For example, the new CEO of a large enterprise sent an opening message to staff asking if they had any feedback on things he should be looking at. “Of course, he received a large number of emails!”

Marritt worked with the CEO’s team by feeding that information into the Workometry engine. Ultimately, the technology enabled them to identify and understand patterns within the business.

Employee feedback in the age of machine learning

Typically, organisations have conducted traditional surveys by gathering raw quantitative data. Then, they would do some qualitative research such as running focus groups or interviews to explore a particular issue in details.

However, Marritt believes that this methodology is expensive and time-consuming. And, “it increases the chance that the findings are out of date before they’re presented.”

Workometry starts with the qualitative data first. It asks a few open questions about a topic to capture large volumes of employee verbatim feedback. Powered by machine learning, the tool analyses the comments and identifies the key themes. It presents the findings through an interactive visualisation of themes. Typically, it spots between 30 and 50 different topics for each question.

“Whereas with traditional qualitative research analysis, the more responses that you receive the harder and more expensive the analysis becomes, with its algorithms as you capture more data Workometry continues to learn, spotting new signals in the ‘long tail’.”


Semantic understanding

The tool breaks the texts at a sentence level and explores the parts of a speech. It identifies the nouns – what people are talking about – as well as the adjectives that are commonly used in those areas. It takes an approach called ‘semantic understanding,’ which means that “it is not looking at words per se, but it at the meaning of words.” It groups similar sentences together. And, it summarises those sentences in a few short words. For example, ‘poor communication’, ‘poor decision-making’ or ‘staff shortages’.

“It automatically tags the sentences to one or more descriptions,” explains Marritt. “Often, it captures several themes as employees can talk about multiple things in response to one particular questions.”


Granular segmentation


After identifying the topics, the tool pinpoints the groups of employees who are most likely to discuss those topics. This is critical to find pockets of the organisation where things may go wrong.

It may not just be in terms of function or department, let’s say Finance. The segmentation goes deeper. “It may be that the people who are most likely to talk about a shortage of career development are the women Gen Y working at the headquarter.”

In fact, as well traditional fields such as work position, location, gender, and age, the technology brings in other business or HR data. For example, employee performance or attendance records.

Companies can zoom into pockets of interests and then zoom out to understand what else those employees are saying. Not only to that particular survey. The technology has the ability to look at any survey that this group has taken in the past. “It can compare feedback across time.”


Marritt works with the client organisations to roll-out the tool. In fact, presently, companies do not have the ability to set-up the survey by themselves. OrganizationView does it. Part of the reason in doing that, he says, is that he tunes the algorithms at the questions’ level.

“We know that the way companies ask questions to their employees makes a huge difference in terms of the quality of the output.”

For example, he has just worked with a bio-technology company in the U.S. It has taken about two weeks to set-up the survey as it was the first time they used the tool. “They want to survey the career development process that they have just gone through. It is a management performance conversation and they are looking at employees’ perceptions towards it.” During the two weeks there was some work around branding and setting up the technology in multiple languages. Plus, reshaping the questions to make sure that they capture as much rich data as possible in a way that the tool can analyse. For existing client organisations the roll-out can be within hours.

Investing in advanced analytics tools


The use case when the tool becomes important is when companies have many employees.

“If your organisation has 250 members of staff, you may still want to read the texts by yourself; it will not take too long. But, if you have 20,000 employees or even more, then it just becomes cumbersome.”

Most probably he is right. The ability to use algorithms can become a stronger factor in determining the effectiveness of the survey, particularly for companies with different languages.

Another use case comes from organisations that are advanced in dealing with data. “They use data across the whole business. They like to be able to tune into something and react quickly. They understand the benefits of using advanced analytics to increase performance.”

Worth mentioning is also the ability to reach remote employees who do not have an email address or access to a computer. For example, shop workers. Because the tool is mobile-friendly, they can easily comment from their smartphones.

Employee life-cycle

For Marritt, having the whole picture of the employee life-cycle enables a company to understand where to invest in employee activities. Ultimately, it becomes key to preventing disengagement.

“Whereas many businesses capture some feedback at certain stages. It is often sporadic, and done in a way which makes analysis difficult, or where the data collected is superficial.”

In fact, OrganizationView’s data suggests that it is hard to re-engage a disengaged person. He believes that it is much better to focus on slowing down the process of disengaging people. And, that can be done only if you understand how people perceptions change over time at the individual level.

“So, it is about stopping looking at the end state of disengagement versus engagement, and really focusing on prevention.”