Before using Drivers
Your text analysis project must contain an appropriate numeric metadata field. Drivers can analyze fields such as:
Ratings or satisfaction scores
NPS
Purchase intent
Employee engagement scores
Binary values such as 0 and 1
Time-to-resolution measurements
Other linear numeric measures relevant to the text analysis project
Balance reach and score difference
Drivers is most useful when both axes are interpreted together. A high-volume concept close to the average may be common across positive and negative experiences. For example, price may be mentioned by almost everyone and therefore show little difference from the text analysis project average.
A lower-volume concept farther from the average may represent a more distinctive “wow factor” or pain point. For example, watery may appear less often but be consistently associated with lower product ratings. Consider whether a concept is:
Widespread and strongly associated
Widespread but close to average
Less common but strongly associated
Less common and close to average
Understand conceptual grouping
When a driver is identified, conceptual matches help represent the broader idea.
For example, a driver such as perfect may also capture related language including excellent, wonderful, superb and perfection.
This can reveal an association that would be fragmented across multiple expressions if only exact matches were counted. Review the included conceptual matches when the result seems unusually broad or does not align with the source records.
Cross-check two numeric fields
Drivers can identify concepts linked to one score and plot them against another numeric field. For example, you might:
Identify concepts associated with a Customer Satisfaction score
Use Likelihood to Recommend on the visualization axis
Examine whether the concepts associated with taste also align with the overall experience
Keep Sync concepts score field and axis score field enabled for a straightforward analysis of one outcome. Disable synchronization only when you intentionally want to cross-check two different measures.
Compare Drivers over time
Use Compare period over period and select an appropriate date field to examine whether associations are changing. This can help you determine whether:
A recurring driver remains stable
A product or operational change altered the relationship
An emerging issue is becoming more strongly associated with poor outcomes
A previous strength is becoming less influential
Use comparable periods and review the number of records in each. Differences may reflect seasonality or a change in the population rather than a change in the experience itself.
Use the Concepts table
Switch from Visualization to Concepts when you need a more structured presentation of the results. The table provides a concise way to review concepts and their average scores without relying only on the scatter plot. This is useful when:
Comparing many concepts precisely
Preparing a stakeholder summary
Reviewing concepts with similar positions
Checking values before exporting the results
Concept colors can also distinguish themes or categories within the visualization and table.
Investigate before acting
Drivers identifies association, not causation. When a concept appears important:
Check its total matches
Compare its average with the overall average
Apply relevant filters to test whether the pattern holds across groups
Review several matching records in Records
Look for exceptions and alternative explanations
Use Volume to determine whether the concept is growing
Use Sentiment to understand the emotional context
A driver is best treated as an investigative signal. The source records and relevant business context determine what the result means and whether action is appropriate.