So Rohit has landed from the West Coast of the US this morning. Wow. That’s phenomenal. He’s also awake! More than I am! Let’s go!
Rohit’s driving aim is to make people happy. Sounds good. And the first question is about sentiment analysis. Not many people know about it (I certainly don’t’). He’s asking who is feeling positive, neutral and negative about the day. (No one was feeling negative). Sentiment analysis is also known as opinion mining. It’s a natural language processing technique that tries to extract opinions from text. Because people might say they’re happy, but write things that show they’re unhappy when they talk about the product.
We know that we have ways of asking for feedback (those damn pop ups everywhere). There are also plenty of other places where we can get textual feedback.
Rohit is involved in mobile testing, and platform usage shows that there is a lot of diversity and fragmentation – just too much to test. Looking at the ratings for their app in the play store, they are specifically interested in the 1, 2 and 3 star ratings.
Looking at some of the examples of bad feedback, they can be about bugs (unable to delete something). They can also contain updates and upvotes that show you how important this is for people. Another example was about the visibility of the app in the recent screen menu. Another example was improved after a developer fixed the issue they noticed. Different user groups (sellers for example) will find specific problems (and ones with monetary consequences). Many consumers ask for features that they still have to do online. One specific example was about the user experience (more than 100 characters are deleted because it goes over the limit – one of my personal pain points!).
Every comment had something that was reproducible and actually rather clear. (Some of these are better than some bug reports I’ve read!).
Rohit used to read new reviews after every release. He was spending about half a day every week doing that, then collecting and presenting data. The manual work was giving useful results, but could it not be automated?
The steps were to fetch daily feedback from the app, filter it to only have neutral or negative feedback, then to generate meaningful daily reports. He took this process and made it run as a part of a daily job. Once they have the feedback, he sorts them according to the features they speak about. Using the information, collected over a few days, they can make decisions on what to address next.