Drivers helps you identify concepts associated with higher or lower values in a numeric metadata field. Instead of only showing what people discuss, it compares the scores of records matching each concept with the overall score for the text analysis project.
Use Drivers to answer questions such as:
Which concepts are associated with results above or below average?
Which common topics have little relationship with the selected score?
Which less common topics show a larger difference from the average?
Are these relationships changing over time?
Drivers requires at least one suitable numeric metadata field, such as a rating, satisfaction score, NPS, effort score, or another measurable result.
How Drivers works
For every concept shown, Journey Insights Hub identifies the records containing an exact or conceptual match. It then calculates the average value of the selected score field among those matching records.
That average is compared with the overall average for the text analysis project or currently selected data. The difference between these averages is the concept’s impact. For example, imagine that a hotel’s overall rating is 4.1 stars, while records matching “wait time” have an average rating of 3.8. The concept has an impact of −0.3 because its matching records average 0.3 stars below the overall result.
Drivers reveals an association between a concept and an outcome. It cannot prove that the concept caused the score to increase or decrease.
How concepts are prioritized
Journey Insights Hub considers several factors when determining which concepts provide the most meaningful driver signals:
Impact: The difference between the average score of matching records and the overall average
Confidence: How likely the observed difference is to represent a consistent association rather than random variation
Relevance: How meaningful the concept is within the text analysis project
Weight: The amount of matching evidence available for the calculation
These factors are combined into an internal importance value used to select and order drivers within the current calculation.
The importance value is relative to the current data and should be compared only within the same text analysis project, filters, and calculation. A concept’s position may change when the selected records, score field, or available concepts change.
Concepts require sufficient evidence and confidence for display. Some concepts remain hidden when their matching records have an average different from the overall result but the available evidence or confidence remains low.
Horizontal axis: Total matches
The horizontal axis shows the number of records that match the concept.
Concepts farther to the right appear in more records. Concepts farther to the left appear in fewer records.
Total matches combines exact and conceptual matches when Include conceptual matches is selected. Repeating a concept several times in one record creates a single record match.
Vertical axis: Average score
The vertical axis shows the average value of the selected score field among records matching the concept. A concept higher on the chart has a higher average score. A concept lower on the chart has a lower average score.
The meaning of “higher” depends on the selected field. A higher satisfaction rating may be favorable, while a higher resolution time or effort score may not be. Interpret the chart according to what the metric represents.
Overall-average line
The dashed horizontal line represents the overall average score for the selected data.
Concepts above the line are associated with an average above the overall result
Concepts below the line are associated with an average below the overall result
Concepts near the line have an average similar to the text analysis project overall
The farther a concept sits from the line, the larger the difference between the average for its matching records and the overall average.
Concept colors
The legend distinguishes concepts above the average from concepts below the average or identified as outliers. Color helps you scan the chart, but the concept’s position and underlying values provide the more complete interpretation.
View a concept’s values
Hover over or select a concept to review details such as:
The concept name
Its average score
Its difference from the overall score
Its total matches
The difference from the overall score shows the direction and size of the association. For example, a difference of −0.4 means that records matching the concept average 0.4 points below the overall result for the selected metric. This difference applies to the matching records as a group. It means that the average score across those records differs from the overall result.
Separate common topics from stronger signals
The position of a concept helps distinguish frequency from score difference:
Far right and near the average line: Frequently discussed, but associated with a result close to the overall average
Far right and far from the line: Frequently discussed and associated with a different average
Far left and far from the line: Less frequently discussed, but potentially important because its matching records have a noticeably different average
Far left and near the line: Less frequently discussed and associated with little difference from the average
A concept that is far from the line but appears in few records should be interpreted cautiously. A small group can produce a large difference that may not represent a stable pattern.
Confidence and match volume should be considered alongside impact. A large impact based on limited evidence may be less reliable than a smaller difference supported by many records.
Compare drivers across subsets
Apply metadata filters to calculate Drivers for a particular subset of records, such as:
A customer segment
A product or service
A region
A feedback channel
A time period
A range of satisfaction scores
When a filter is applied, Drivers recalculates the overall average, concept averages, impacts, and driver selection for the filtered records. For example, you could compare the drivers associated with low ratings before and after a product release or investigate whether different concepts are associated with satisfaction among new and established customers.
Because each filtered calculation uses a different set of records, compare the results as separate analyses. Treat impact values from different subsets as comparable only after considering the score distribution and available evidence.
Compare drivers across time periods
Use trend analysis to compare Drivers results between two time periods. This can help you:
Identify directional changes in concept impact and match volume
Compare consecutive periods or year-over-year results
See whether different concepts emerge as drivers over time
Monitor an established concept list consistently
Trend analysis compares two separate periods rather than displaying a continuous trend line. Each period appears in its own scatterplot.
Configure trend analysis
Open Drivers.
In Configure, select the numeric score field you want to analyze, and choose which concepts to display.
Select the edit icon beside Select date field to compare periods and choose a date field. Drivers displays two scatterplots. The upper chart initially uses the full date range available within the active filters.
Select the period you want to analyze in the upper chart.
The lower chart automatically displays the immediately preceding period of the same length. For example, if the upper chart covers 10 days, the lower chart displays the previous 10 days, ending one day before the upper period begins.
Adjust the lower date range when you want to compare a different period, such as the same dates from the previous year.
Use Active concepts when you want both charts to compare a consistent set of business metrics. With Unique to Filter, different concepts may appear in each chart because Journey Insights Hub evaluates which concepts are most meaningful for each period.
When comparing the charts, look for concepts that:
Move vertically, indicating a change in their average score or impact
Move horizontally, indicating a change in the number of matching records
Move closer to or farther from the overall-average line
Appear as a meaningful driver in only one period
A concept appearing in only one chart does not necessarily mean that it was absent from the other period. It may not have met the criteria for display. Check its match count, confidence, active filters, and supporting records before interpreting the change.
Understand why Drivers results may change
Drivers results can change when you:
Add or remove records
Apply or remove filters
Select a different numeric field
Include or exclude conceptual matches
Rebuild the text analysis project
Filters change which records are included in the calculation. Adding records or rebuilding the project may also affect the available concepts, conceptual matches, overall averages, and internal importance rankings. Review the active filters, selected score field, conceptual-match setting, and sample size before comparing results from different analyses.
Avoid treating association as causation
Drivers can show that records mentioning a concept tend to receive a different score. It cannot determine whether:
The concept caused the score
The score influenced how the person wrote the feedback
Another factor affected both the language and the result
A small or unusual subgroup created the pattern
Treat Drivers as a way to prioritize investigation. Review the matching records, relevant filters, sample size and business context before deciding what the result means.