It's my first post in the Community – and I hope to receive your feedback. :)
While using some great features inside Zendesk we have encountered one and the same challenge – how to perform cheap but smart analytics for a big amount of our low structured text data by minimizing the manual involvement of our support agents.
There are in general 3 areas where we get this kind of the text data and analytics challenge:
We receive hundreds of comments and many of them are very informative for product development, customer service optimization, CX etc. But we have not enough time (honestly no time at all!) to go through all the comments and analyze them. A long time ago we have activated Satisfaction reasons but around 90% (!) of customers are leaving this field blank (see the screen below).
Comments people leaving by participating in NPS surveys are very useful too (besides NPS scores). But again, we get a lot of comments in 2-3 days when surveying thousands of customers at once. Characteristic for NPS surveys is the even bigger diversity of the feedback themes compared to satisfaction surveys.
Searches in Help Center
We are using search stats to optimize our Help Center content for better article findability. I'm sure you have noted that search strings in your Help Center are distributed as a typical "long tail" – there are less high-frequency but many and many low-frequency searches. But if we look deeper into the search strings we can note that themes of these searches are much less diversified. But who will spend the time to go through hundreds or thousands of these rare searches coding them into different themes?
Thus my questions to the Community are the next:
- Do you experience similar challenges like we?
- How do you solve these problems today?
- What solutions should be used (now or in the future)?
We ourselves are investigating manual analysis (it's a bad idea!) or automated analytics solutions (based on AI techniques, like Natural Language Processing) to cope with this challenges. The only problem with these solutions is that they are not adapted to analyze data better than simple keyword search – by reflecting the specifics of our business domain in general (i.e. telecom, e-commerce etc) or even of our unique business processes, products etc. Also, pricing question is always a challenge. :)
So any feedback and thoughts will be very useful and helpful.