Using ChatGPT To Perform A Product Review of the UBA Mobile App

A little over a year ago I wrote an article on "How To Uncover User Problems Without Using In App Metrics". The main point was to demonstrate how product managers could perform a qualitative assessment of their product using reviews given by their users on the Google Playstore if for whatever reason in-app product analytics wasn't available or they simply didn't have access to the tools and applications to do so, some of which can be expensive. This approach was a hack and you can read the article to understand why.

Recently I have attempted to come up with a much more streamlined approach that takes up a lot less time. For instance, the initial approach I described used word clouds that involved a painstakingly manual process of "normalising" the data to improve the quality of the output after parsing all of the reviews through a word cloud. In all I probably spent about 4 to 5 hours working through this which included copying and pasting text of the reviews left by users from Google Playstore into a Google sheet.

Enter ChatGPT. I won't spend any time explaining what ChatGPT is, there is enough out there everywhere now that does this a lot better than I can. My goal using ChatGPT was the same as the initial review I did last year. Perform an analysis on the reviews left by users of a mobile app for a major bank in Nigeria and try to come up with a list of feature update priorities the engineering team can work on based on this. This time round, I opted to use user reviews of UBA's Mobile App on Google Playstore here instead of Sterling Bank's, which was what I used in the review from last year. My plan was to determine what the top issues or challenges were that UBA users faced and prioritise them for the engineering team to work on.

Here is how it started...

Something worth noting is the ability of ChatGPT to output quality responses has a great deal to do with the quality of your "prompt" (the instructions you give it). Tip: provide as much context as possible in your prompt about what you are trying to accomplish - this helps the AI to better understand what it is being asked to do. You can also even ask it to ask you questions to clarify its own understanding of your instruction. In previous instances of using ChatGPT, I have found this to be very helpful in the quality of responses I get.

And here is how it finished...

You can view my full conversation with ChatGPT here. Needless to say, it did NOT take me 5 hours to complete this task and I literally felt I still had enough mental energy to work on other things - which is one of the main points about using AI, right? It's meant to free up your time so that you can focus on more high-level / strategic stuff. If you are a product manager working for a rapidly scaling tech company, time is one of the luxuries you do most certainly not have and anything to free up more time is extremely valuable.

Whats even more impressive is that ChatGPT performed the analysis by going through all the 18.7k reviews and comments left by users in a matter of seconds without blinking. When you combine this heavy lift with its ability to present the information in way that is easy to digest such as using tables and representing the data as percentages (something that would take a human an amount of time to do) it's very easy to see where and how useful it is. What I also thought was very useful from a data accuracy point was that it was able to explain why the figures it reported did not add up to 100%.

Would you like me to help your business or product make the best use of ChatGPT? Then reach out by sending me a message here on LinkedIn or you can send an email to: [email protected]