To learn about the science of Drivers read this article.
What are Drivers?
Drivers are a way for you to easily see what is driving up and down your numerical subset (NPS Score, Star Rating, etc).
A ‘driver’ can either be concepts found automatically by Luminoso's artificial intelligence, or can be concepts of interest that you select.
How does it work?
Drivers is a feature that examines a set of documents at the level of individual concepts (words or phrases and their conceptual matches), and reports when the presence of those concepts are correlated with high or lower scores.
The Drivers tool takes in the set of exact matches, their concept vectors, metadata, individual relevance, and determines the top drivers within a set of documents (aka a project). If a concept score driver’s confidence is too low it will be excluded as a score driver to avoid presenting misleading results.
Bear in mind that Drivers is not a causal model. Meaning, if a very positive review mentions “free wifi” that it’s not the sole reason for the positive review. Rather a hotel guest had a very positive experience at the hotel and given the opportunity to describe their positive experience called out the “free wifi”.
Drivers are correlations. These correlations are very helpful to understand a domain.
How can I use Drivers?
Daylight project create permissions
Data must have a driver field that is either a numeric value (number_[FIELD]) or a score (score_[FIELD]).
To get meaningful driver results it is recommended to upload a minimum of 300 documents that are comprised of text surveys/reviews and a minimum of one numeric score.
Open Drivers by selecting the star-shaped icon from the left side sub-navigation bar.
If the user provided a score type data field the Drivers tool will immediately start to calculate drivers, if no score field was designated the user will be prompted to select one from the numeric fields of data available. To the left of this, you can choose whether you would like to see Luminoso Suggested Drivers, or view your topic list (saved concepts) as drivers.
Once drivers are done being calculated they will be displayed
The y-axis shows driver impact score , the x-axis the driver Total Match count
Impact Score = How much a concept is driving up your numerical subset.
Total Match Count = How many documents that driver was in.
The Concept detail panel on the right side will give you details around the drivers. You can see the top associated concepts, matching documents. Conceptual matches are considered in the calculation of drivers.
Why aren’t all the project’s suggested or saved concepts showing as drivers?Not all top concepts are project Score drivers. Just because a concept is mentioned frequently does not mean that it drives score.
Why do negative drivers typically have a larger range than positive? Typically customers write more vocally about negative drivers than positive.