Perception Intelligence: Unraveling the Customer Feedback Conundrum

The CEO of a fortune five hundred firm found that the feedback her organization was receiving from top customers was negative. Metrics from their customer satisfaction surveys would indicate the opposite. It turns out, some customer service reps were selectively sending out surveys to customers they knew would respond positively. The CEO’s reaction? Collect feedback more often and more consistently. And that opened the flood gates. Their responses grew from a few hundred a month to over four thousand, collected across multiple channels, a good portion in foreign language. More than half of these responses contained open-ended feedback. The customer experience team struggled to keep up with the volume and create actionable advice for their product teams. The feedback was frustrating; data was too noisy, it was poorly trusted, examples were anecdotal, and no actionable changes were obvious.

Enter Perception Intelligence, which makes the advances in artificial intelligence accessible to customer service teams. We worked with the client to consolidate their feedback into a standard format, applied machine learning to identify key themes within open-ended responses, and prioritize the most relevant comments. We made sure the key themes were aligned with human expertise, as well as the solution with critical business objectives.

Because our client’s privacy is paramount, below is an illustrative dashboard only.

 

We used Google’s suite of language processing APIs to handle foreign language and identify references to specific products, people, and locations. Our state of the art machine learning algorithm learns from both human labeled examples and uncategorized feedback to reach human-level performance at categorizing future responses. Our accuracy at categorizing comments is up to 90%. It turns out the final step, prioritization, unlocked the most value. In a world where there are growing amounts of data, with limited trust,** it’s more important than ever to have trusted tools that can sift through the noise. The client can now identify the most relevant comments, positive or negative, for each customer service theme, broken out by specific product lines.

Because our client’s privacy is paramount, below is an illustrative dashboard only.

Whereas product owners previously viewed data shared with them as anecdotal, the organization now has regular conversations around this data. They are quicker to respond to evolving customer needs and can demonstrate repeated themes in qualitative data. Since using Perception Intelligence, their customer experience metrics have increased by over 20%.

Want to learn more about Perception Intelligence? Click to get started.

**Fewer than 16% of all managers full trust data they use to make decisions. (MIT Sloan Management Review, 2016)

Cal Al-Dhubaib is the Chief Data Scientist at Pandata.

Leave a Reply

Your email address will not be published. Required fields are marked *