Sentiment Analysis: Improved Confidence in Decision-Making
An infectious disease life sciences client needed follow-up answers on a previous campaign survey focused on treatment for a new indication on existing disease therapy. The client believed a pattern should exist separating two groups of healthcare consumers but initial research findings did not reveal one. High confidence in the split would justify additional marketing expense.
Knowvanta believed that Sentiment Analysis insights would reveal what conventional research did not. Focused on sentiment specific elements, a deeper nuance of the data analysis could be revealed.
A traditional approach to the new research would be too slow and still miss the data needed. Normally, a detailed follow-up line of questioning requires market research analysts to dig through the research interviews and notes using ad hoc tools and techniques with an indeterminant timeline to complete.
To implement quickly, Knowvanta leveraged technology and applied key word and tagging techniques specific to an infectious disease using opinion-mining algorithms. This allowed the machine learning model to test multiple outputs. The phrasing and key word combinations drove the ultimate threshold levels and results for the improved decision making.
Sentiment Analysis Healthcare Weightings
| THRESHOLD | OPINION MINING ACCURACY | SENTIMENT DIFFERENCE | STRENGTH OF SENTIMENT DIFFERENCE |
|---|---|---|---|
| to split Campaign | Based on Key Word and Subject | % | (1 weak -7 strong) |
50% Difference In Opinion > 3 strength of difference (1 – 7 scale) | 98% | 68% | 5.3 |
Sentiment Analysis Healthcare Strength
| KEY WORDS | KEY PHRASING | WEIGHTS |
|---|---|---|
| Insurance | Coverages, Medicare, reimbursements, co-pay, out-of-pocket | High |
| Friends, Family | Experience, positive, bad, negative, good | Medium |
| Doctor | Recently, changed often, trust, judged, misunderstood | High |
Sentiment Analysis provided a highly effective solution. Results were produced in a fraction of the time traditional methods would have required. The approach revealed a significant difference of opinion and high confidence in the data and the decision that would be made for additional cost and effort in marketing segmentation.