How to conduct a sentiment analysis?
Artur Zbiejcik avatar
Written by Artur Zbiejcik
Updated over a week ago

If you have come to this point with us, you surely already know how to create tags, assign them to a specific answer of a respondent in the survey and mark their emotional color. You also know how to find out what and how much your respondents are saying about your brand thanks to quantitative analysis using cross tabs, rankings and trends.

Time for the next and last step - instructions explaining how to analyze the sentiment of customers' statements. Thanks to this analysis, we will find out not only how much, but also WHAT our customers say about a specific topic (previously defined as a tag, e.g. price, product quality, customer service, etc.)

How to conduct a sentiment analysis?

1. First, go to the Results tab, and then Sentiment Analysis

2. To start the analysis, the user must first select the date range to be analyzed (it can be the entire period of collecting the survey results, as well as a specific month, week, etc.). By default, the date range for the entire survey is set. The next step is to select tags, i.e. topics raised by respondents, that are to be presented in the analysis.

  1. After clicking the button: Show, we will get three charts that analyze the sentiments of the statements in detail.

a) a chart showing the distribution of sentiments

On the chart, we can easily see what the respondents say about a given trend. In this case, the colors represent the emotional tinge (sentiment) of the answer. Green is positive, orange is neutral, and red is negative. At the end of the scale, the Sentiment Score is also given, i.e. the average of sentiment, on a scale from 1 to -1, where 1 is fully positive and -1 is fully negative.

How do we calculate the Sentiment Score? For example, the last of this chart has a value of 0.67. Here we have 2 positive tags, totaling all 3, so โ…” gives us 0.67.

The chart is also ranked from the most frequently to the least selected tags at once.

b) Trend Sentiment Score chart

The chart shows when the respondents expressed a lot and good opinions about a given trend, and when little and bad ones. Or the opposite - what tags were talked about in a good way, but little, or much, and in a bad way. The bar chart shows the sentiment of the statement, i.e. how much was said about a given trend, and the linear one how much was said about it.

In the case of the line chart, the weighted average is calculated, where the value of 1 is achieved when all respondents expressed their opinion about a given trend positively, 0 neutral, and -1 negatively.

The user can at any time change the trend (drop-down list in the upper right corner of the chart), which is being analyzed, or choose a collective chart containing a summary of all previously selected trends.

The summary chart shows only the line chart, i.e. it shows the emotions involved in the answer on a given topic. So we can see which trend is associated with the most positive emotions at a given time (here: product-creation), and which causes negative customer feelings (here: price).

c) sentiment ranking

The ranking is used to clearly present trends, divided into those tags/areas which are most frequently talked about positively, neutrally and negatively.

The bar chart shows the number of people who expressed a positive, neutral and negative opinion on a given topic, thanks to which you can see at a glance which areas of the company's operations require improvement.

Do you have questions or concerns related to the verbatim analysis?

We are going to help you!

We recognize that the entire process of verbatim tagging and sentiment analysis takes time and requires a great deal of attention. If any part of the instruction is not clear or you have additional questions, please contact our support department, who will be happy to help you at every stage of the survey.

When creating a survey and carrying out an advanced analysis of results, you can also use the information collected on our blog and the instructions and articles in the guide.

Did you find the answer to your question?

Did this answer your question?