Trustpilot Analytics: Review Insights - Topics
Review Insights uses machine learning technology to detect topics, trends, and sentiments within your business’s reviews. Use the Topics page to see how each of your topics are performing.
To find the Topics page, go to Analytics > Review insights > Topics.
The first thing you’ll see on your topics page is a table with all the topics we’re tracking for your business and their corresponding details.
We use a machine learning model to sort through millions of reviews to find clusters of related words. Those clusters are then used to shape and define a topic. Our topics cover subjects most frequently discussed across all the reviews on Trustpilot, such as product, customer service, delivery, etc.
We analyze all of your reviews, and each sentence is used to determine the topic(s) being discussed and their sentiment.
Topics are shown in order of most review mentions, with any favorites appearing first. You can apply filters and sort by using the arrows next to Topic, Sentiment score, or Reviews.
You'll see a star icon next to each topic. Click on it to mark it as a favorite and promote it to the top of the table.
In addition to our topics, you can create your own custom topics. Custom topics have a pencil icon next to the topic’s name.
What’s a sentiment score?
You’ll see that each topic has a sentiment score. When our algorithm detects one of your topics in a review, we run that sentence through our sentiment model, which labels it as positive, negative, or neutral.
The score is calculated by taking the percentage of positive mentions and subtracting the percentage of negative mentions, resulting in a score between 100 and -100. Next to the score, you can see how much it’s improved or decreased compared to the previous period.
Our sentiment model is regularly updated for improvement, so scores may change over time.
You’ll also be able to see the number of reviews mentioning the topic within the date range, alongside a percentage showing the increase or decrease in review volume compared to the previous period. We also show an excerpt from your latest review that mentions the topic.
Go from 4-stars to 5-star reviews
This section displays your most recent 4-star reviews with a negative sentiment score. Analyze them to get insights on what held them back from being 5-stars.
Topic mentions per star rating
Here you have heat maps that show the distribution of positive and negative mentions of your topics across 1 to 5-star reviews. Analyze them to identify your strengths, weaknesses, and points of friction where a sentiment contradicts a review’s star rating. Click a cell to be redirected to its corresponding page for the subset of reviews.
Can I view individual topics?
Yes! You can click on each topic to learn more about it. You’ll be taken to a page where you’ll find the following info:
- Sentiment score: The sentiment score for the topic’s reviews within the date range.
- Reviews: The number of reviews mentioning the topic within the date range.
- Of all reviews: A percentage that compares the currently selected subset of reviews against your total number of reviews within the date range.
- Sentiment score over time: A chart displaying the sentiment score for the topic at daily, weekly, or monthly intervals within the date range. The intervals are chosen based on the date range. Hover over a data point to view the sentiment score and number of reviews it’s based on.
- Sentiment per star rating. A chart displaying the number of times the topic was mentioned in reviews within the date range, and whether it was in a positive, negative, or neutral way.
- Locations: An overview of where your reviewers are located. Bar graphs display the top locations and total number of reviews for the period. Hover over the graphs to see the star rating distribution for each location.
Below all that, you can find all related reviews. Any sentences that our algorithm associates with the topic will be highlighted - green for positive, red for negative. We also include any additional topics mentioned and an option to mark the review as misclassified if you feel the topic classification is incorrect.
Export the reviews as a CSV file by clicking Export.
Filter results for a specific topic
-
In Trustpilot Business, go to Analytics > Review insights > Topics. Find the topic you want to filter and click on it.
-
Search for a specific keyword in the search bar, or select a filter for Date, Tags, Star rating, Sentiment, and/or Reviewer country. Click Apply.
Note: If you select more than one tag to filter with, reviews will only be filtered using one of them. For example, if you filter with tags A and B, you'll see reviews that have tag A, and reviews that have tag B. You won't see reviews that have both tags A and B.
Create a custom topic
You can also create your own custom topics to tailor insights more specifically to your business or industry.
-
In Trustpilot Business, go to Analytics > Review insights > Topics. Click + Add topic, at the top right. A pop-up may appear if it's your first time using this feature.
-
Under Add keywords to include in your topic, enter keywords or phrases you’d like to be associated with the topic. Toggle on Advanced to exclude reviews with specific keywords you don’t want to include for the topic.
Accepted characters are A-Z and 0-9. Keywords must be separated by a comma.
When you enter a keyword, a Topic preview appears showing the volume of reviews that match one or more of the keywords from the last 6 months. You can also see Recent reviews mentioning your keywords. These can help estimate the review volume you can expect from the topic.
- Click Save keywords.
- Enter a Topic name and Description, then click Save topic.
Edit or delete a custom topic
You can edit or delete custom topics at any time.
-
In Trustpilot Business, go to Analytics > Review insights > Topics. Find the topic you want to edit or delete.
-
Click the pencil icon next to the topic's name.
- To edit the topic, enter your changes and click Save topic.
- To delete the topic, click Delete topic. Click Yes, delete to confirm.