The Art of Labeling User Feedback for Your Mobile App
Feedback is the single most valuable source of information about the app.; however, extracting meaningful data from it might be a challenge.
Join the DZone community and get the full member experience.Join For Free
Feedback is the single most valuable source of information about the app.; however, extracting meaningful data from it might be a challenge. The key to making use of this information is structure – you see, the answers to your app’s issues are already somewhere in that pile of text waiting to be discovered. To do it the right way, you need to understand the strengths and weaknesses of the options available.
The Importance of Feedback
The value of user feedback is difficult to overestimate. On the one hand, it is a major source of information on technical issues the app might have. On the other hand, it helps the dev team understand whether the project is on the right track and what can be done to enhance user experience. Finally, it serves an important function – namely, it shows customers that their voice is heard. The latter two are essential for winning the hearts of millennials, who form a massive chunk of the market’s demographic. This is why mobile app development companies are now integrating feedback systems within their products to encourage participation among their audience.
Once feedback channels are in place, new problems may surface. Apparently, simply receiving feedback is not enough to make customers happy – someone needs to actually do something with it. Once the amount of feedback hits a certain threshold, this process can quickly get out of hand. Trying to sift through several hundred tickets per day can clog the workflow quite quickly even for a large team. In fact, you still may find polarizing opinions on the value of feedback – some companies have a general understanding of its importance but lose faith once they get overwhelmed by an avalanche of disparate messages.
The Solution: Feedback Labeling
The most feasible way to get a hold of this is through labeling – basically, assigning categories to messages to make the management process easier. This can be done in a variety of ways, from doing everything by hand to entrusting the job to AI. In most cases, the selected method will fall under one of three categories:
- Manual labeling
- Partially automatic labeling (simple algorithms)
- Fully automated labeling (via machine learning)
Each category has its strengths and shortcomings, so let’s look into them in detail.
The most straightforward way of labeling feedback and, incidentally, the least efficient one (more on that later). Manual labeling can be done by support team members or the users themselves. In the first case, the team uses a pre-defined list of categories to mark what each message is about based on its contents. A good example is a system used on Github, where issues are assigned to one or more of the conveniently color-coded categories:
The problem with this approach is that it’s resource-heavy. Going through dozens, if not hundreds, of messages a day just to label them is a laborious task that would certainly not benefit project management. On top of that, such a monotonous job inevitably introduces errors, which in a sense defeats the purpose of categorizing them in the first place.
A good way to address the issue is to delegate the task to users by including a drop-down menu with a common label in the submission form. Such a tweak is a huge time saver and can make labeling more precise – at least in theory. The idea is that a person submitting feedback will know exactly whether they are writing a complaint about the pricing or have a problem with the UI.
In practice, people may have difficulty deciding on the appropriate label as well. Not everyone is familiar with the terminology, so they will choose whatever they feel is the closest to their subject. Not only that, having an extra field might feel confusing and restrictive, which is not an impression you would want to make on your audience.
Partially Automated Labeling
A more streamlined and less intrusive way to label feedback is through the use of algorithms that will analyze the text and come up with suggested tags. Such algorithms vary in complexity but are usually not too sophisticated. Their main strength is being intuitive – you just choose from suggestions as you type, nice and simple.
However, this method will work best for common word structures, tripping over more elaborate sentences. Not only that, it won’t be suitable for multilingual products, as each new language will need its own algorithm.
Automated Labeling Through Machine Learning
If simple algorithms don’t cut it for you, neural networks may be the solution. Machine learning is already used throughout mobile app development with excellent results. Neural networks have already demonstrated impressive results in text recognition, so using them for labeling is a logical step. Initially, this was the most complex and costly method of the three. Nowadays, the market is responding to the demand with ready-made solutions, so you have a good chance of finding the right tool for the right price.
Labeling may seem like a chore and, for the most part, it probably is. However, this should not be an excuse for ignoring it. In fact, with the abundance of AI-driven tools, managing feedback, and using it to your advantage is easier than ever before, so make sure you implement it in your project.
Opinions expressed by DZone contributors are their own.