Introducing Text Insights: Your AI-Powered Employee Feedback Decoder

Text Insights is an advanced AI-driven feature that enables HR executives to quickly understand engagement levels. It goes beyond surface-level analysis by meticulously extracting and breaking down key topics and sub-topics from open-ended feedback or text responses. It detects the underlying reasons for employee sentiment and highlights areas that require attention, allowing you to make significant changes and improve employee engagement. 

Topics are broad themes from employee feedback, such as communication or leadership, while sub-topics provide detailed insights, like clarity of instructions or manager approachability. Analyzing both levels helps HR professionals uncover nuanced insights that might be missed with a more general approach. This deeper understanding of engagement challenges allows you to pinpoint the most impactful areas to elevate your engagement metrics. 

Key Benefits

• Quick Insights: Rapidly gain insights into employee sentiments, including sentiment and impact.
• Identify Concerns: Easily spot and prioritize employee concerns for prompt action.
• Informed decisions: Obtain a concise overview of topics and sub-topics to facilitate swift decision-making.

Analyzing Responses with Text Insights:

• Step 1: Navigate to the 'Text Insights' section to dive deep into text responses after launching an engagement or pulse survey.

• Click on “Start Analyzing” - This option can only be enabled with a minimum of 10 responses.
• Based on sentiment and impact scores, access an organized display of topics and sub-topics. For pulse surveys, you can use a global filter based on survey rounds.

A quick summary:

• Step 2: Select a particular topic (eg. Manager Supportiveness) and you’ll be zoomed in on its corresponding sub-topics for a more detailed understanding. 
• Whenever you hover over any topic or sub-topic, you can instantly access detailed insights/summaries, including sentiment and impact scores, helping you grasp the core insights at a glance. 

Impact by 

Step 3: To analyze your data further, use the “Impact By” option. This option allows you to analyze the impact of the topics based on favorability score, normalized score, mention count, and eNPS. This visualization helps you see how different factors contribute to overall sentiment and impact, giving you a clear path to actionable insights.

Group by: 

Step 4: Click on the “Group By” button as this option lets you visualize and analyze data either by topics or reporting factors. Topics are broad themes, with sub-topics and related responses nested underneath them. 

Text responses under each reporting factor in the survey are analyzed to identify sub-topics, which are then grouped and mapped to corresponding reporting factors. The mentions count of responses for each topic/sub-topic is the number of times a particular topic/sub-topic has been mentioned in the responses. The size of sub-topics or reporting factors is proportional to their mention counts, allowing for clear visualization of patterns and concerns. They are plotted on the graph as distinct data points, highlighting patterns and concerns. This visualization enables admins to quickly pinpoint specific issues and tailor actions to specific factors, improving organizational outcomes. 

Step 5: You can also get a more spaced out and decluttered view of topics in a particular quadrant for better understanding, by double-clicking on the area you want to focus on.

Responses: 

Step 6: Select the specific subtopic to view its responses.

Note: Text response counts in positive, negative and neutral categories are determined by their sentiment scores.

Step 7: Click on the “+ Add Filter” button to narrow down responses by location, gender, department, or job title.

Step 8: Toggle the “Highlight sentiment” feature to see keywords highlighted by color, making it easy to identify sentiment trends.

X-Axis: Sentiment Score

The sentiment score on the X-Axis indicates the sentiments of employee responses under each topic/sub-topic. This helps you quickly assess whether the overall sentiment is positive or negative. The sentiment for each sub-topic is derived from the average sentiment score of its grouped responses. 

• Negative Sentiment: Scores range from -1 to 0, reflecting topics/ subtopics with a negative sentiment score.
• Positive Sentiment: Scores range from 0 to 1, indicating topics/ subtopics with a positive sentiment score.

Calculation with AI:

The sentiment score is calculated using AI-driven natural language processing (NLP) models. Here's how it works:

Text Processing:

• The AI analyzes all text responses from surveys.
• It breaks down the text into short phrases.

Sentiment Analysis:

• The AI evaluates phrases using a sentiment analysis model.
• It assigns a sentiment score based on the meaning and context.
  • Positive Example: "The support from my manager has been excellent" might score +0.8.
  • Negative Example: "I’m frustrated by the lack of recognition" might score -0.7.
  • Neutral Example: "I'm glad to contribute, but the workload could be fairer".f

Scoring:

• Positive Words/Phrases:
  • These are scored between 0 and +1.
  • Example: "The training sessions were beneficial" might score +0.7.
  • Example: "I feel appreciated by my team" might score +0.9.

• Negative Words/Phrases:
  • These are scored between -1 and 0.
  • Example: "There’s a lack of clear communication" might score -0.6.
  • Example: "I’m unhappy with the current workload" might score -0.8.

Aggregation:

• The AI combines the scores from all responses.
• It calculates an average sentiment score for the entire response.

Final Sentiment Score:

• The X-axis of the sentiment graph is segmented according to the average sentiment scores, calculated by averaging all responses related to a specific topic or reporting factor. 
• For example, if the responses regarding "manager support" are largely positive, the overall sentiment score will approach +1.

By leveraging advanced AI techniques, sentiment analysis provides a nuanced understanding of employee feedback, allowing organizations to identify key areas for improvement and celebrate positive feedback effectively.

Y-Axis: Impact Score

The Y-Axis, labeled as Impact Score, divides the topics/subtopics into two clear zones. The midpoint, based on the highest overall score, splits the axis in half:

• Low Impact: Below the midpoint, indicating less significant/less impactful topics.
• High Impact: Above the midpoint, showing more critical/more impactful topics.

Calculation:

The impact score is calculated using a dominance analysis algorithm that considers several factors. Here's how it is done:

1. Normalized Score:
   • The average percentage scores of eNPS (employee Net Promoter Score) and rating scale values are being calculated.
   • These normalized scores are then analyzed to determine their impact.
2. eNPS:
   • The eNPS score is passed directly to the algorithm for impact identification.
3. Favorability:
   • Favorability scores measure how positively or negatively employees feel about specific topics. This approach allows for a more nuanced understanding of which areas have the most positive or negative influence on employee sentiment, making it easier to identify and prioritize key issues.
4. Mentions Count:
   • Mention count reflects how often a topic or sub-topic appears in employee responses, highlighting the frequency and importance of specific issues. This metric helps HR professionals quickly identify recurring themes that require attention.
   • This distribution helps to determine the impact score based on the frequency and prevalence of mentions.

By incorporating these factors into the dominance analysis algorithm, the impact score is providing a comprehensive measure of how significant each sub-topic or topic is in influencing overall employee sentiment and satisfaction.

Quadrants

Based on the X and Y axes, topics and sub-topics are categorized into one of four quadrants to help prioritize actions:

1. Address Immediately: High Impact with Negative Sentiment

These issues have a significant negative impact on overall employee engagement. Action here is urgent, as unresolved issues in this quadrant can quickly lead to disengagement among employees. 

2. Minimize & Reassess: Low Impact with Negative Sentiment

These issues are not currently critical, as they have a lower overall impact on employee engagement. However, these negative sentiments should be monitored over time to ensure they don't escalate or start affecting other areas.

3. Focus & Leverage: Low Impact with Positive Sentiment

These are areas where employees have a positive perception, but the overall impact on engagement is lower. While these may not require immediate action, they present opportunities for further improvement and can be leveraged to build a more positive work environment.

4. Maintain & Monitor: High Impact with Positive Sentiment

These are high-impact areas that are currently performing well and contributing positively to employee engagement. The focus here is on maintaining this success through continuous monitoring, ensuring these critical areas remain strong and effective.

5. Key overall action:

The goal is to move items from the 'Address Immediately' quadrant to the 'Maintain and Monitor' quadrant, ensuring they stay there through continuous monitoring. Similarly, for the 'Minimize & Reassess' quadrant and 'Focus & Leverage' quadrant, the focus is on improving these areas to eventually shift them into the 'Maintain and Monitor' quadrant. The aim is to continuously elevate issues into a stable, well-maintained state.

Additional Information

• Real-Time Updates: It syncs fresh responses every 15 minutes to keep your insights sharp and up-to-date.
• No Text Response: If a reporting factor has no text responses, it will not appear in the chart. This ensures the data remains relevant and actionable.
• Edited Reporting Factor: When a reporting factor is edited, users need to refresh the data to reflect the latest changes. This ensures that the analysis remains accurate and accurate and up-to-date.

Summary

Text Insights is a powerful tool for HR teams to analyze employee feedback, prioritize necessary actions, and improve engagement strategies effectively. By understanding the sentiment and impact scores, HR professionals can make informed decisions to enhance employee satisfaction and address concerns promptly.

FAQ

1. What is the sentiment score? How is it calculated? How is it relevant?

The sentiment score in Text Insights is calculated by averaging all sentiment instance scores, ranging from -1 (most negative) to 1 (most positive). This score helps identify how employees feel about specific topics, guiding strategic actions. e.g., “poor quality” might score -0.3, while “worst experience” could score -0.9. 

2. I changed a survey question included in Text Insights. Will it affect my analysis?

Yes, changing a question can impact text insights as rephrasing may affect contextual relevance and reduce accuracy.

3. What’s the difference between mentions count and response count?

• Mentions Count: The number of times a topic/sub-topic has been mentioned in the responses. 
• Response Count: The total number of text responses.