Text Drivers
Text Drivers is a predictive analysis that identifies which themes in your text responses are most strongly associated with a specific outcome — for example, which topics most drive high or low satisfaction scores. It bridges qualitative text analysis and quantitative measurement, showing you not just what people say but what it means for the numbers.
Adding a Text Drivers analysis
Section titled “Adding a Text Drivers analysis”- In the Workbench toolbar, click Text Drivers.
- The Text Drivers configuration panel opens. Work through the Options and Advanced tabs.
- Click Save to add the chart to your output library.
Options tab
Section titled “Options tab”Variable to predict
Section titled “Variable to predict”Choose the outcome variable you want to predict — typically a rating scale, NPS score, or satisfaction metric.
- In the Variable to predict selector, choose a supporting variable from your project.
The model identifies which text themes most strongly correlate with each category of this variable.
Chart types
Section titled “Chart types”| Chart type | Best for |
|---|---|
| Bubble | Showing predictive strength and frequency as a scatter-style plot |
| Confusion | A confusion matrix view showing how accurately themes predict each category |
Minimum mentions
Section titled “Minimum mentions”Set the minimum number of times a theme must appear in the text to be included in the analysis. Themes mentioned too rarely may produce unreliable predictive scores.
Advanced tab
Section titled “Advanced tab”Define two categories
Section titled “Define two categories”By default, Text Drivers compares responses at the high end of your chosen variable against those at the low end. In the Advanced tab, you can define exactly which values constitute each category — giving you more control over the comparison.
- Under Category 1, select the values from your variable that represent one group (for example, scores of 8, 9, 10 for Promoters).
- Under Category 2, select the values that represent the comparison group (for example, scores of 0–6 for Detractors).
- Give each category a meaningful label to make the chart easier to read.
Defining clear, contrasting categories typically produces more actionable driver insights than using the default high/low split.
Reading the results
Section titled “Reading the results”A Text Drivers chart shows themes ranked by their predictive strength. Themes near the top of the chart for Category 1 are strongly associated with that outcome — when respondents mention these themes, they tend to fall into that category.
Look for themes that appear strongly on one side but not the other — these are your most actionable drivers.
Frequently asked questions
Section titled “Frequently asked questions”What kind of variable should I use as the “variable to predict”?
Any supporting variable in your project can be used. Rating scales, NPS scores, and categorical groupings (such as customer type or satisfaction tier) are common choices. Define the two categories in the Advanced tab to control exactly which values are compared.
How is predictive strength calculated?
Text Drivers uses a statistical association model that measures how much more (or less) likely a theme is to appear in responses from each category. It is not a full regression model, but it provides directionally reliable driver rankings for most business research purposes.
Can I use Text Drivers with an NPS variable?
Yes — NPS is a common use case. Define Category 1 as Promoters (9–10) and Category 2 as Detractors (0–6) in the Advanced tab for a standard driver analysis against NPS.
My Text Drivers chart shows very few themes — why?
This can happen if the minimum mentions threshold is too high, or if there is not much variation in the text between your two categories. Try lowering the minimum mentions threshold, or check that your variable to predict actually has meaningful variance in the dataset.