Page updated February, 2021.
Watch this quick clip and read the below details to learn more about Glint’s Comments and Narrative Intelligence features. If you want to share the video with users outside of the community, use this link.
Introduction
Narrative Intelligence at Glint helps you understand exactly what people are expressing in pulse comments. These are known in Glint as prescriptive comments and sentiments.
Open-ended responses are extremely useful for generating deeper insights that fuel initiatives to improve both organizational and people success. There is great benefit in giving employees opportunities to provide comments on a Pulse and exploring the rich insights contained within their comments.
Artificial Intelligence for Human Resources (AI-for-HRTM) makes it possible to transform employee comments into actionable insights. Glint’s Narrative Intelligence solves the business problem by turning valuable comment data into immediately consumable and understandable information and insights. Narrative Intelligence uses our Natural Language Processing (NLP) engine to connect the dots between what your employees are saying (topics), how they feel about it (sentiment), and how their comments relate to engagement, driver questions, and other outcomes.
Glint’s Narrative Intelligence is sophisticated, accurate, and easy to navigate. Glint’s state-of-the-art, Machine Learning technology is continuously listening and determining when new topics should be added to the platform. Using this technology, Glint improves upon the industry standard (which shows 92% accuracy) by matching 20% more comments and 24% more topics.
In combination with our reporting technology, companies can now reap the benefits of using a shorter, more predictive pulse, receive more frequent and actionable results, and simultaneously use complementary insights from open-ended commentary along with quantitative data.
Best Practices for Using Comments
Optimize Survey Design
Glint’s Methodology recommends reducing the number of questions per pulse and focusing on a broader set of topics. The comments section of each question allows employees to provide detail, color, explanation, and nuance to their answers. The strength of our narrative intelligence allows us to mine deeper insight from those comments without subjecting employees to longer surveys or requiring leaders from reading thousands of responses.
About one in three people leave comments per pulse, and they leave 3-4 comments on average. Over time and across multiple pulses, more than 80% of people in a company will leave comments at least once.
Explore Comments After Reviewing Scores
Comments can provide helpful context around scores you are exploring in the rest of the application. Wherever you see a comment count you can click on it to view the comments and go to a comment report with the full Narrative Intelligence experience. Here you can view comments by any population and/or question to get a better idea of what people are saying for a given score.
Consider Related Topics
Review how topics connect to each other in the bubble visualization for a new way to gain deeper understanding. For example, when someone is talking about “Culture,” Connected Topics show what other themes they mention in relation to “Culture,” and whether they talk about it with positive, neutral, or negative sentiment. Explore how a topic appears across questions, what the related topics are, and what keywords appear in the comments.
You can click on any of the topics to view the underlying comments or further drill down into the topic to see a new comment report focused on that topic. Drilling down into a topic also allows you to understand the context around that topic. Each connection is either a sub-theme or something that is mentioned in the context of that topic.
Take Action from Prescriptive Comments
Prescriptive Comments are those, identified by narrative intelligence, that may help offer specific actionable suggestions. Review the prescriptive comments that may offer helpful suggestions for how you can improve in an area. Because Narrative Intelligence is easily accessible from anywhere in the platform, you can access deeper qualitative insights from alerts, goals, or any other report. Continuous listening, as part of ongoing conversations, and jointly developing goals with employees, will lead to identifying the most appropriate actions that result in improvements in the workplace experience and the success of the business.
Beware Comment Bias
Note that it is natural to take comments personally, however, resist the urge to do so. Unlike in live conversations, pulse comments tend to be more negative than positive. Focus on what the comments tell you about root causes for the quantitative scores, and look at prescriptive comments for possible solutions.
Explore Different Populations
The Narrative Intelligence experience can be filtered by any population available elsewhere in Glint. We recommend filtering the report by different attributes to see what your key populations are talking about, such as specific teams, high performers, or new hires. This will give you a new view as to how the population feels in their comments overall (sentiment) and what they are talking about and how those things interrelate (topics and keywords).
Review Keywords
Much more fine-grained than Topics (there are ~100 curated topics; there are > 16k keywords and we’re constantly adding more). As an example, things like “PPE” and “N-95 masks” show up here now, but would not exist as Topics.
Explore Representative Comments
We isolate a short list of comments that are representative of overall themes. Read just a few to get a sense of the whole.
Frequently Asked Questions
Q: Who has access to comments?
A: Each company sets their own rule for the level at which comments will be shared using our flexible role-based permissioning system. When available, comments should be reviewed for your own team, or the next available larger team. Glint’s confidentiality threshold protects individual respondents and restricts comment access and filtering to populations of 10 or more respondents (although this number varies by company).
Q: How do comments help us take the right actions?
A: After reviewing scores, take a look at what people are saying to understand and gain greater context around a score. Comments can be especially helpful in explaining the root causes for high impact, lower scoring items. Look at the prescriptive comments for suggestions on what can be improved. This is where you may get ideas for taking action.
The ability to identify suggestions or solutions is a powerful part of Narrative Intelligence. These comments may help you make decisions about what actions to take that will be perceived well and be effective because they come from people who are closest to the challenges. They can also help you formulate questions and topics for a team discussion.
Q: How do you generate comment sentiment?
A: Comment sentiment is tagged by our NLP Engine. Each comment is reviewed by an algorithm and the language used determines whether it falls into one of the following buckets: positive, neutral, negative, or mixed. The NLP tagging has been trained extensively over time and continues to improve. The comment sentiment is beneficial in that it can tell you how people feel in open-ended comments where there is no rating to infer someone’s favorability. Note that survey comments tend to be slightly biased towards negative language sentiment.
Q: Can the same topics be associated with more than one question?
A: Yes, a single topic may appear in response to more than one question. See if the comments that mention the same topic across the questions are similar or different. Similarities may indicate a systemic issue that needs to be addressed. Differences may highlight specific situations or areas where action is needed.
Q: How do you determine connected topics?
A: Connected topics are a result of many comments being tagged with both topics. For example a comment mentioning how their commute is ruining their work life balance would be tagged with both “commute” and “balance” topics. If enough comments had the same tagging then those topics would appear as more connected in the visualization. The stronger the connection, the more it means that when people are talking about balance, they really are talking about the commute. This provides the necessary context to drill deeper into a topic. Oftentimes, the strongest connections are sub-themes.
Q: What do the keywords represent?
A: Keywords are a way to quickly see what words people are using in their comments and can provide more detail beyond the topics. The size of the keyword indicates volume and the color indicates sentiment. When you click on a keyword, you can see the sentiment by question and all the comments, including prescriptive comments, related to that keyword. You can toggle between the cloud visualization and a sortable table view.
Q: What are representative comments?
A: Representative comments are samples of actual employee comments that are representative of the underlying data by using Glint’s Narrative Intelligence based on our AI-for-HRTM technology. The sampling is stratified based on our keyword cloud and sentiment distribution of the key phrases. The number of representative comments shown follows a square root pattern, e.g. 10,000 comments → 100 representative comments.
Q: How are Glint’s NLP algorithms kept up to date?
A: To ensure our NLP is providing you with the greatest benefit, we are constantly evolving our algorithms to improve accuracy. We do this through continued research and development, more comment annotations, and customer feedback. Any time we have an improvement in our NLP features, we put it through stringent tests to validate that it is more accurate than the previous algorithm before releasing.
Once released, the algorithm will run automatically against all previous comments. This ensures that classifications are more accurate and that trending is consistent over time. We will always announce NLP improvements in our release notes to provide more context around what is changing. This could include improved sentiment classifier, more and improved topics or changes in the keyword suggestions.
Q: What languages are supported for Narrative Intelligence Technology?
A: Our Narrative Intelligence currently includes support for over 50 languages. See the Multi-Language Support page for more information.