Entity Sentiment Analysis: The big insights unlock for mobile app development.
Sentiment analysis unlocks all kinds of information and insights into what people think about something. In this article, I’m applying it to a mobile app. Specifically, the user reviews. Yes, that eclectic mix of rants, raves, gripes, and directives can be evaluated and organized into information that guides us forward. Shedding light on both where we should look to optimize our app as well as what exactly we should be doing.
So let’s look at 3 examples of how sentiment analysis can help with mobile app analysis and optimization. We’ll review mobile apps from 4 major healthcare brands. Aetna, Cigna, Blue Shield of CA, and Stanford Health. And we’re using reviews posted to the Google (Android) app store during the 2022 calendar year.
Identify app areas with the greatest need for improvement
Sentiment analysis data for mobile apps can, at first glance, feel just as messy as the unstructured text it’s derived from. There are 2 keys to uncovering the value. 1. Extracting entities from the text, and then running sentiment analysis on the individual entities. This is usually referred to as entity sentiment analysis. And 2. Classifying those entities (and their surrounding user commentary) to align with your conversion funnels. That last step involves some AI assistance to do at scale, but being able to measure sentiment for an entire conversion funnel makes all the difference in providing the big-picture POV.
Let’s dig into this.
The chart below identifies some of the key conversion funnels for the Aetna Heath brand’s app. Note that these conversion funnels are highly customizable. In this case, I’m going with broader funnels, but these could also be applied to more specific user flows.
Sentiment ratings go from -1 to +1. In this case, it takes the sentiment scores for all the touchpoints (the entities we’ve extracted) within this conversion funnel, then calculates the average score. That indicates the overall user satisfaction, or dissatisfaction, for that conversion funnel.
It’s pretty easy to see that ‘Onboarding’, with its lowest sentiment score, is ruffling some users’ feathers. So if you’re wanting to optimize this app’s user experience, that would be one of the first areas I’d take a deeper dive into. What do I mean by a ‘deeper dive’? Because we’ve run entity sentiment analysis on the reviews, we can then go and assess all those entities, which are often user touchpoints, that fall within a given conversion funnel. That will reveal in more detail what the users’ issues are. I’m not going to delve into that here, but if you’d like to see how that works, check out this article.
Determine competitive advantages or weaknesses
Taking this same conversion funnel sentiment analysis to another degree, we can compare our analysis to our competitors.
In the chart below, that’s what we’ve done, combining the data for the Cigna, Blue Shield of CA, and Stanford Health apps.
Comparing Figures 01 and 02, a couple of observations stand out. One is that it seems like all these apps are causing user dissatisfaction at the onboarding stage. The brand that tackles this first should have a significant opportunity. In marketing an app, onboarding is pretty much the first impression. Making a bad first impression leaves a brand open to competitors poaching their unhappy users.
Furthermore, if you’re hemorrhaging users at this key conversion point, it’s going to make the return on your marketing budget less efficient. Fixing this means you can spend more on marketing than your competitors and still have a positive conversion-based ROI.
Taking this to yet another level, you can also assess each individual brand’s Onboarding conversion funnel, as well as the specific user feedback around each entity involved.
The end goal here is to identify ways to have a competitive advantage over the experience your competitors are offering.
Metrics for benchmarking future progress
Looking at sentiment analysis over time brings a new dimension to our analysis: Progress. This gives us a better understanding of two areas. 1. Whether changes we’re making over time are showing improvement in user experience, and 2. Whether significant shifts are happening in user experience ratings for our competitors.
For our own data, we can average user sentiment scores from a given period, make optimizations based on what we’re seeing in the user feedback, and then measure the change in user sentiment going forward. That’s a simple but valuable way to measure the impact of our work.
For competitors, we can look at points where they’re showing a marked change in sentiment, either positive or negative, and then review the user comments from that conversion funnel during that time period. That can reveal what the issue was. Reviewing client comments going forward can reveal what the client did, and how well it worked.
Better data views for better optimization results
The data we’ve gone over here is really just the beginning of the entity sentiment analysis process, but highlights how user reviews can provide actionable brand and competitive analysis, and give clear direction on where to look for additional insights and direction.
Entity sentiment analysis also helps us benchmark changes in user experience. Identifying the trends, and informing you what users are saying about those trends. That information helps you both more efficiently allocate development resources, and build deeper user empathy for everyone involved with the project.
Interested in reviews analysis for app design projects? Let’s chat.
