How to Analyze Employee Feedback
min read - Updated on Nov 15, 2026You can analyze employee feedback in Kimola by turning survey comments, exit interview notes or Glassdoor reviews into a Report that groups what employees write under themes, labels how positive or negative they are about each theme and lists what goes wrong and what is missing at work. Kimola asks research Dimensions, such as Pain Points and Unmet Needs, of every comment, ranks the themes employees are least satisfied with and shows where those themes concentrate, by department, location or period. This guide follows four steps: collect the feedback, create the Report, read the themes and pain points, and compare groups of employees.
Sign in to your Kimola account and have your employee feedback ready: the comments of a survey as a spreadsheet file or a Google Sheets spreadsheet, or the Glassdoor page of your company's reviews. Everything in this guide is available in every package.
The Goal
This guide answers the question that HR, people analytics and leadership teams ask after every survey cycle: why do employees feel the way they do, and what should be fixed first. An engagement score tells you how employees rate their workplace. The comments next to the score tell you the reason, in the employees' own words, and they are the part that is slow to read and rarely counted. Kimola reads all of them, groups the same point across hundreds or thousands of comments and counts the comments behind each one, so you see which problems affect the most employees and where they concentrate.
Themes and sentiment show what employees talk about and how they feel about it. Pain Points names what goes wrong, such as "unpredictable shift schedules", and Unmet Needs names what employees wish they had, such as "a clear path to promotion", in the words employees use. Every answer stays tied to the comments behind it, so you can open it and read them.
A retail chain runs a yearly engagement survey with one open question, "What would make your job better?", across its stores and its warehouse. The Report might show workload and career growth as the largest themes. Pain Points might name unpredictable shift schedules and understaffed weekends, and Unmet Needs might name a clear path to team lead. The chain now has a ranked list of problems, each with its share of the comments.
The method works best on comments in which employees describe their experience in their own words: engagement and pulse survey comments, onboarding and exit survey answers, and employer reviews on Glassdoor. Comments that only praise carry no problem to find, and Kimola leaves them without a Pain Points answer. The Reports, Dimensions, Performance, Compare, Tables and Signals in this guide are available in every package.
Step 1: Collect Employee Feedback
Collect the feedback in one place before you create the Report. Kimola reads feedback from a file, from a spreadsheet in Google Sheets or from an Airset, a dataset collected with the Airset Generator. Internal survey comments come as a file or a spreadsheet, and Glassdoor reviews come as an Airset.
Export Survey Comments
Start from the results of your survey or HR tool. Most tools can export responses to an Excel or CSV file, and where the export button is depends on the tool. Kimola accepts .xls, .xlsx, .csv and .tsv files, one file per Report, up to 30 MB. Put the column names in the first row and one comment in every row below it. Two columns matter to Kimola.
Text Column (required): the column with the comments. Keep the comments of one question in this one column.
Date Column (optional): the date and time of each comment. With it, the Report can show how topics change over time, for example from one survey wave to the next.
Kimola doesn't analyze the other columns, but they become filters in the Report. Keep the columns that describe a group of employees, such as a department, a location, a job level, a tenure band, a survey wave or a score. Step 4 uses them to compare groups. Leave out names, email addresses and employee IDs: Kimola doesn't need them, and every column you keep becomes a filter in the Report.
Keep columns such as department, location, tenure and score in the file. They become the groups you compare in Step 4.
A Report analyzes one Text Column. When a survey has several open-ended questions, upload the file once for each question and choose a different Text Column each time. Employees who skipped a question leave its cell empty. Rows with an empty Text Column are left out of the analysis and don't count in queries, so you don't need to delete them.
A retail chain exports its engagement survey with the columns Comment, Department, Store Region, Tenure, Submitted On and Score. Comment is the Text Column and Submitted On is the Date Column. Department, Store Region, Tenure and Score stay in the file as filters, so the team can read and compare the comments of each department and each region.
If the responses are already in Google Sheets, for example the response sheet of a Google Forms survey, you can skip the export. Connect Google Sheets and pick the spreadsheet in Google's file picker. It opens in the same report wizard as an uploaded file. The integration is available in the paid packages, not in Starter, and Kimola On-Premises doesn't offer it. Where it isn't available, download the spreadsheet as an .xlsx or .csv file and upload that.
[Screenshot: A survey export opened in a spreadsheet, with the comment column, the date column and the department and location columns]
Collect Glassdoor Reviews
Glassdoor reviews are employees' own reviews of an employer, and Kimola collects them with the Airset Generator, its Chrome extension. Each collection becomes an Airset in your account, a dataset that stores the raw reviews. You set up the extension once; see Set up the Airset Generator. Then open the Glassdoor page of the company whose reviews you want to analyze, go to its Reviews tab, click the Airset Generator icon in the browser toolbar and click Generate. The extension reads the reviews on the page and continues as you scroll and load more. When you have the reviews you need, click Continue in the extension window. The new Airset is sent to your Kimola account and appears under Airsets in the left menu. How to Scrape and Analyze Glassdoor Reviews describes this collection in detail, and What is an Airset? explains what an Airset holds.
An Airset has the columns of the platform it was collected from, usually Content, Date and URL. Before you create a Report, you can open the Airset with View and delete the rows you don't want to analyze, one by one or in bulk. See Manage Airsets.
Step 2: Create the Report
Both kinds of feedback end up in the same report wizard. For a survey file, open Reports in the left menu, drag the file onto the upload area or click the area and choose the file, then choose the Text Column and the Date Column in the Sheets & Columns step. For a Glassdoor Airset, open Airsets, click Create a Report on it and choose the same columns there; for most Airsets the Text Column is Content. See Create a Report by Uploading a Custom Dataset and Create a Report with an Airset for each way in detail. The steps after the columns are the same.
Choose Classifiers and Structure
In Classifiers, choose the Automatic Classifier and the Sentiment Classifier together. The Automatic Classifier discovers the themes from the comments, so you don't need a list of topics in advance, and the Sentiment Classifier labels each comment as positive, negative or none. Together they give the Report its themes with their sentiment, and the best and worst performing themes need both. If your HR team already codes comments with a fixed set of topics, a custom model with those topics can take the place of the Automatic Classifier. Custom models are available in the paid packages, not in Starter. See Create a Report Using Automatic Classifier.
In Structure, turn on Enable Multi-label Classification. One employee comment often mentions several things, such as workload and a manager, and with multi-label classification it is counted under each of them. Without it, each comment is grouped under one main theme. The option is on when the step opens. Because one comment can carry several labels, the shares of a Report's themes can add up to more than 100%. See Create a Report Using Multi-Label Classification.
[Screenshot: The Classifiers step with the Automatic Classifier and the Sentiment Classifier chosen]
Choose Dimensions
Dimensions are optional, and they turn the comments into research findings. For employee feedback, start with Pain Points, which asks "What goes wrong or creates difficulty?", and Unmet Needs, which asks "What is missing or desired?". Add Underlying Causes, which asks "Why does this occur?", when you want the reasons behind the problems. Motivations, which asks "Why is this chosen or preferred?", suits feedback in which employees say why they joined or why they stay.
In the Dimensions step, click a Dimension to select it, and click it again to remove it. The selected Dimensions appear under My List. You can also add Dimensions later from the Report's Dimensions page. See Add Dimensions to a Report.
[Screenshot: The Dimensions step with Pain Points and Unmet Needs selected and My List showing 2]
The Dimensions you select appear under My List, so you can review them before you create the Report.
Review and Create
The Review step shows your choices before anything is analyzed. The Report Title is optional, and a title that names the survey and its wave, such as "Engagement Survey 2026", makes the Report easy to find later. Choose the Report Language: English, Spanish, French, German, Portuguese, Dutch or Turkish. It is the language in which Kimola writes theme names, Dimension answers and summaries. The comments can be in any of the 30+ languages Kimola analyzes, so a survey answered in several languages by employees in several countries gives one Report. See Language Options in Reports.
Check Dataset and Required Query, then click Create Report. The number of queries follows the rule queries = rows × models × Dimensions: every analyzed row is multiplied by the number of models and the number of Dimensions you chose. A survey of 800 comments analyzed with the Automatic Classifier, the Sentiment Classifier, Pain Points and Unmet Needs uses 800 × 2 × 2 = 3,200 queries. See How to Calculate Queries. When the analysis is complete, Kimola shows Here we go!, and you can open the Report with View Report.
Every Dimension you choose adds to the queries of the Report: queries = rows × models × Dimensions.
Step 3: Read Themes and Pain Points
Read the Report from the whole to the detail. Overview gives the whole picture, Performance ranks the themes, Dimensions shows what goes wrong and what is missing, and Content shows the comments behind the numbers.
Start with Overview
Overview is the first page of the Report. TOTAL DATA shows how many comments Kimola analyzed, and SENTIMENT shows the share of positive, negative and none. BEST PERFORMING THEMES and WORST PERFORMING THEMES list the three themes employees are most and least satisfied with. A card for every Dimension shows its most frequent answers, and the EXECUTIVE SUMMARY writes up what all the feedback says, followed by a summary of every large theme. The SIGNALS card at the top shows the strongest Signals, when the Report has any. See Overview in Reports.
[Screenshot: The Overview of an employee feedback Report with the TOTAL DATA, SENTIMENT, BEST PERFORMING THEMES and WORST PERFORMING THEMES cards]
Rank the Themes in Performance
Click Performance in the menu of the Report. Every theme gets a Satisfaction Score and a Dissatisfaction Score from 0 to 100, and the page lists the Worst Performing Themes by Dissatisfaction Score, highest first. A score of 80-100 is Critical: most of the theme's feedback is negative, across enough comments to be sure of it. A score of 60-79 is Moderate: the theme draws a lot of negative feedback and is worth watching. Below 60 is Acceptable. The scores are adjusted for the number of comments, so a theme with three negative comments can't outrank a theme that hundreds of employees write about, and the same comments always give the same scores. See Performance in Reports.
[Screenshot: The Performance page with the Worst Performing Themes by Dissatisfaction Score list]
Performance needs a content classifier and a sentiment classifier. If the page shows no scores, check that you chose the Automatic Classifier and the Sentiment Classifier in Step 2.
Read Pain Points and Unmet Needs in Dimensions
Click Dimensions in the menu of the Report. A row of tabs runs across the top: All, then one tab for each Dimension, with a check mark on the ones the Report has analyzed. Open the Pain Points tab first. The summary card shows how many comments answer the question and their share of all comments, and a few paragraphs say what employees struggle with across the whole Report, with the key terms in bold. Kimola counts the answers first, and an AI model writes the summary afterwards from numbers that are already computed, so the text never decides which problem is larger.
Below the summary, each pain point is a cluster card, largest first. A card shows the pain point's name, a one-line finding, its number of records and its share, a short summary and quotations: excerpts in employees' own words. Click the eye icon next to a quotation to open the whole record. Cluster Scale shows each pain point's share of the records that answer Pain Points, and Dataset Scale shows its share of all the records in the Report. One comment can name several pain points, so the shares on Cluster Scale can add up to more than 100%. Use Copy as image or Download as image in the top right corner of a card to turn it into a picture for a presentation.
[Screenshot: The Pain Points tab with the summary card, the record counts and the cluster panel with Cluster Scale on the right]
Open the Unmet Needs tab next to read what employees ask for. Pain Points lists what employees complain about, and Unmet Needs lists what they wish they had. To see why a problem happens, click Dimensions at the end of a pain point's card and drag Underlying Causes into the box that says "Drop columns and dimensions to create tables". The card then shows the causes within this pain point, each with its number of records, its share and its sentiment. This works when the Report has Underlying Causes, and a table on a card is a quick look that isn't saved. See Dimensions in Reports.
For the retail chain, the Pain Points tab might show unpredictable shift schedules as the largest cluster, with 240 records. Dragging Underlying Causes into that card might show schedules published two days ahead as the largest cause behind it, ahead of no way to swap shifts. The Unmet Needs tab might show a clear path to team lead at the top. The chain now has a problem, its size and a cause it can test.
Check the Verbatims in Content
Numbers tell you how much, and the comments tell you why. Content lists every comment with its column values, its Dimension answers, its sentiment and its themes. Pick a theme or a Dimension answer from the filters, or click a word in Most frequent words, to read only the comments behind it. The search box finds comments by the words they contain, and an asterisk at the end of a word finds every word that starts with it, so manage* finds manage, manager and management. Reading a few comments of a theme is the quickest way to see whether its label fits what employees wrote. See Filter Records in Reports.
Step 4: Compare Groups of Employees
The columns you kept in the file turn the Report into a comparison between groups of employees: one department against another, one location, one tenure band or one survey wave. A Report from a Glassdoor Airset has fewer such columns, and its Date column lets you compare periods instead. Compare groups large enough that a comment can't be traced to one person. Four pages answer these questions, each for a different kind of comparison.
Compare Groups Side by Side
Compare sets up to 5 groups side by side. With Compare by Columns, choose a column and the values you want, such as three departments or two locations. With Compare by Filters, choose saved filters instead; see Save Filters in a Report. Every group gets the same measures: its volume, its trend over time, its sentiment, its themes, its Dimension answers and its popular terms. A theme that one group has and the first group doesn't is listed under Mismatches, which is often the most telling result. Compare reads the analysis the Report already has, so it uses no queries. See Compare in Reports.
[Screenshot: The Compare page with four departments side by side]
Cross Themes with Columns in Tables
Tables is where you cross-tab the comments. Drag a column, a classifier or a Dimension into the table, and Kimola groups the comments by it, with the count, the share and the sentiment of each group. Drag a second one to split every group again. A table grouped by Department and then by Pain Points shows the pain points of each department. A column of numbers, such as a score, can be measured instead of grouped, with Average, Sum, Minimum, Maximum or Median for every group. See Tables in Reports.
For the retail chain, a table grouped by Department and then by Pain Points might show unpredictable shift schedules as 38% of the pain points in the stores and under 10% in the warehouse. Measuring Score with Average on the same table shows which department rates its workplace lowest.
Score the Themes for One Group
Performance takes the same column filters. Pick a value from Columns, such as one department or one location, and both lists of themes are calculated again for the comments of that group only. This shows which themes upset one group as well as the whole survey.
Let Signals Point to Differences
Signals are statistically tested findings. A Differences signal says that a topic is mentioned more, or more negatively, in one group than in the others, where a group is a value of one of your columns or an answer of a Dimension. A Trends signal says that mentions of a topic changed over time, and it needs the dates of the comments. Kimola generates a Report's Signals after the analysis, and Open in Tables on a card shows the records behind it. A Report with too few comments, or too few columns that describe the employees, may have no Signals; more comments and columns such as a department or a location give the tests more to compare. See Signals in Reports and What is a Signal?.
[Screenshot: A Differences signal card with Why this is shown opened and Open in Tables at the bottom]
For the retail chain, a Differences signal might say that understaffed weekends is mentioned in 31% of the comments of one region and in 11% of the comments of the others, and a Trends signal might show onboarding rising in the last months. The chain now knows which region to visit and which problem is new.
Trends need the dates of your comments. Choose a Date Column when you create the Report.
A Report with too few comments or columns may have no Signals.
What You Get
The Report gives you the following from the comments themselves.
Themes with sentiment. Every comment is grouped under the themes it mentions and labeled positive, negative or none, with each theme's share of the feedback.
Scores. A Satisfaction Score and a Dissatisfaction Score for every theme, and the best and worst performing themes of all the feedback or of one group.
Pain Points and Unmet Needs. The problems employees describe and the things they ask for, grouped as clusters with record counts, an AI-written summary and quotations in employees' own words, and the causes behind each problem when you add Underlying Causes.
Group comparisons. The same measures for each department, location or tenure band in Compare, a table of themes crossed with any column in Tables, and Performance scores for one group.
Signals. The statistically tested differences between groups and changes over time, when the Report has enough data.
Exports. The Export menu of the Overview offers Excel, CSV, PDF, PowerPoint and Email, and every summary card can be copied or downloaded as an image.
Next Steps
To turn the findings into work, open Actions in the menu of the Report and create an Action Plan. Kimola writes up to 12 evidence-checked actions for a Profile, each with a confidence level, an owner and a metric. The goal is Balanced when the step opens. Action Plans written from a Report are available in every package. See Create an Action Plan. To share the findings with leadership, export the Report as a PowerPoint file or copy the cards you need as images.
If your survey asks a 0-10 recommendation question with a comment, see How to Analyze NPS Survey Comments. For the whole workflow with any survey file, see How to Analyze Open-Ended Survey Responses. A survey that repeats every month or quarter can also be followed file by file in a Space.
Spaces are available in the Enterprise package. They are also available in Kimola On-Premises.
For more on how teams use this kind of research, see Employee Feedback Analysis.