employee hours saved
A household name, Dell is one of the world’s largest international computer tech companies, specialized in personal computers, servers, data storage devices, and other computer hardware and software.
Dell’s HR department conducts a yearly employee satisfaction survey, across their global offices, to understand their employees’ level of satisfaction with everything from their office environments and amenities to gym and parking facilities and more.
To best understand the feelings of their employees, many of the survey questions are open-ended, allowing employees to respond with their free and unrestricted opinions.
Previously, Dell analyzed its employee satisfaction surveys manually – taking two to three employees over a month to sort, categorize, annotate, and evaluate.
With over 10,000 open-ended comments to analyze, they needed a solution to save time and money that could perform at or above the accuracy of manual human analysis. They needed to sort the responses by custom category, then find out which were positive, neutral, or negative. And they wanted the results fast, so they could implement data-driven insights right away.
Open-ended survey responses are hard to analyze because they can’t simply be calculated on a star rating or number scale, and often reveal reactions and data results that the questioners may not have even considered.
Each response contained multiple opinions and diverging sentiments, so responses needed to be broken into two separate opinion units before being properly analyzed
Manually analyzing open-ended surveys is costly and time-consuming. Dell needed a fast solution that allowed them to gain insights right away.
The technology had to integrate with their platform and needed to be up and running within a month, just in time to handle high volumes of data during the high season.
Insights were available to facilities managers 4 months earlier than the previous year.
Most frequent complaints were bought to managers’ attention sooner.
The whole process took just one week (from specifying goals and methodology, training a custom topic analyzer, running topic and sentiment analysis, and then building a Tableau dashboard to display the results).
400+ employee hours saved since only one employee was needed over a period of one week.
10,000 comments were analyzed in just one week.
Once MonkeyLearn’s analysis framework is set up, it performs 24/7, in real time, and much more accurately than humans
Dell used MonkeyLearn’s suite of text analysis tools to perform aspect-based sentiment analysis
The comments were first broken into individual opinions using MonkeyLearn’s opinion unit extractor:
Then they built a custom topic analysis model to organize responses into predetermined categories: Break Area, Food Service, Gym, Meeting Room, etc.:
Topic analyzers only take a few steps to build and can be easily customized to the language and criteria of any business.
The above shows Dell’s topic analyzer. You can see the survey topic tags they used to tag open-ended responses on the left, topic classification statistics on the top (that show how well your model is working), and a word cloud on the bottom (that shows the most used and most important keywords from the survey analysis).
From here, Dell used MonkeyLearn’s pre-trained sentiment analysis model to understand the sentiment or “opinion polarity” of each of the opinion units:
The end result is aspect-based sentiment analysis – Dell was able to categorize each opinion unit into a predetermined topic and analyze each one for sentiment.
The whole process took around one week and just one employee, from specifying goals and methodology, training a custom topic analyzer, running topic and sentiment analysis, and then building a Tableau dashboard to display the results from over 10,000 comments.
Once MonkeyLearn’s analysis framework is set up, it performs 24/7, in real time, and much more accurately than humans.
“The amount of time and money saved was tremendous. The dashboard was available for facilities managers from August, while it was not until December last year.”
Business Analyst @ Dell