Phrasee's language insights give you a glimpse into the language experiments that the Phrasee Brain designs for your brand. The automatic insights and data visualizations allow you to see what is being tested and resonating with your audience. Language insights are currently available for broadcast and trigger projects.
This article covers the language insights available at an individual experiment level. Check out this article for information about language insight reports at a project level.
In this article, we cover the following about your experiment-level language insights (click on the link to jump to each section):
Where can I find them?
Language insights are automatically displayed at the experiment level once the results for your experiment are received (i.e. completed experiments).
To view the insights for a given experiment, simply open the completed experiment and navigate to the Insights tab at the top of the page.
What am I looking at?
Experiment-level language insights give you a look under the hood at what sentiments were tested in each send. Phrasee analyzes language against seven key positive sentiments found within marketing language.
After Phrasee has received the results of the language test you can:
Explore the levels of the seven sentiments within your test variants.
Use the tooltips to gain insight into how you might apply these sentiments across your marketing copy outside of Phrasee.
Click on the sentiment titles surrounding the chart to see which words in your test variants the Phrasee Brain thinks contributed most to each sentiment.
Phrasee sentiments
Sentiment is one of the features of language that Phrasee loves to test in each experiment.
Sentiments can be dialed up or down and still be within a given brand's tone of voice. So by testing a wide range, you’re able to see which distribution engaged your customers most and how this changes from experiment to experiment.
The seven key positive sentiments found within marketing language are:
Sentiment | Common sentiment features |
Impressed |
|
Helpful |
|
Curious |
|
Excited |
|
Surprising |
|
Appreciative |
|
Urgent |
|
Sentiment radar graph
The radar graph displays the distribution of sentiments within the language variants that you've tested with Phrasee. This lets you see what Phrasee is testing and what your audience is engaging with.
You can hover over the tooltip next to each sentiment to view the common features of that sentiment.
By default, Phrasee displays the three top-performing variants on the radar graph. You can change the graph view by selecting or unselecting the checkboxes next to the variants in the table below the graph.
Sentiment variant table
The variant table underneath the radar graph displays the sentiment scores across all seven sentiments for each variant. The score is out of 10. The higher the score for a particular sentiment, the more prevalent it is in the variant.
You can also click the sentiment titles in the radar graph to highlight the words in the variants that the Phrasee Brain interprets to contribute to the variant sentiment scores.
How does it work?
The sentiment scores are assigned through the use of a multi-class classification model. The function is similar to the model Phrasee uses power and optimize your split tests - but instead of being trained on performance data, this model is trained to recognize patterns within data that contribute to the underlying meaning of language.
Each tested variant is run through the model, which calculates the degree to which the variant aligns with each of the sentiments - the higher the sentiment score, the more the variant fits into the sentiment.
The second calculation is run to determine which words within the variants are contributing most to the sentiments. This can be seen by clicking on the sentiment titles surrounding the chart.