Whether you’re the developer of an app or a daily user, you have one thing in common: you want it to work well and make your life easier. But behind a great user experience is not just well-written code; it’s the ability to understand what happens when your app reaches the hands—and opinions—of millions of people.
Among these opinions, often hidden among thousands of ratings and comments, there are clear clues about what’s wrong, what’s great, and what could be improved. And, recurrently, a theme emerges that has a direct impact on the quality of the product: how the app deals with gender.
An international research team analyzed 7 million reviews of the 70 most popular apps on Google Play to discover what users are saying about this topic and what implications it has for creating stronger, more highly rated digital products. The result? A data-driven guide to understanding where we gain and lose the trust of app users.
A research team formed by Mojtaba Shahin (RMIT University), Mansooreh Zahedi (University of Melbourne), Hourieh Khalajzadeh (Deakin University) and Ali Rezaei Nasab (Shiraz University) has analyzed how the topic of gender appears in mobile app reviews.
They have done this by combining massive data analysis, advanced artificial intelligence techniques and manual review. The goal: to understand what real needs and problems users detect and how this information can be used to create quality digital products.
From massive volumes of data to useful information
The first step was to download 7 million reviews of the 70 most popular Android apps on Google Play, spread across 22 categories ranging from communication and entertainment to finance.
To locate reviews that discussed gender, the team created an initial set of 45 keywords, extracted from literature and previous sources, which they then expanded to 396 terms using the KeyBERT technique. This process allowed them to filter out over 12,000 reviews that were potentially gender-related.
After two rounds of pilot analysis to eliminate terms that generated false positives, the set was reduced to 4,885 reviews with a high probability of addressing the topic.
The role of automatic classification
From this set, a final dataset was created with 620 reviews on genre and 820 unrelated reviews. This material was used to train several automatic classification models. The best result was given by RoBERTa, with:
- F1 score of 90.77%
- Accuracy of 86.64%
This means it is very efficient at identifying reviews that contain gender conversations. But detecting them was only the first step: they needed to be analyzed to extract useful information for the product.
What users really say
The qualitative analysis of 388 reviews on gender allowed them to be grouped into six major blocks of product improvement.
Functionalities
Many people are asking for more options and flexibility when configuring the application. There are also cases of products that do not recognize female voices well or that force the gender to be displayed without the option to hide it.
Appearance
From emojis to colors and designs that exclude or box in, visual elements influence the user experience.
Content
Do books, series, games… represent everyone or perpetuate stereotypes? What the app offers matters.
Politics and censorship
Who decides what is shown and what is removed, and by what gender criteria? Moderation decisions affect the perception of the product.
Advertising
The barrage of ads telling you who you should be based on your gender. Misdirected ads degrade the experience.
Community
From hostile groups to spaces where inclusion and respect reign. The atmosphere within the app is part of the product.
Keys to achieving good digital products
Give more control
The study shows that many people are frustrated because they cannot adjust gender-related options within the app: from deciding how the options are displayed to configuring what type of content or advertising they receive. Developing clear and easy-to-use mechanisms to customize these preferences can prevent the user from feeling that the app is imposing an experience they do not want, and improve the overall perception of the product.
Incorporate diversity into the team
When the development team is homogeneous, there is a greater risk of overlooking the needs of other profiles. Including people with diverse experiences and identities helps detect biases before they reach production, better interpret user opinions, and propose solutions that respond to real and varied needs.
Conduct research with a diverse user base
Gender expectations and concerns are not the same everywhere. Factors such as culture, age, or social context influence how an app’s functionality or policy is perceived. Involving people from different backgrounds in product research and testing allows for a richer perspective and avoids decisions that may be appropriate for one group but alienating for another.
Constantly monitor and analyze feedback
Reading reviews once is not enough. Opinions change over time and with product updates. Establishing a continuous monitoring and analysis process helps detect patterns, evaluate whether implemented improvements are working, and act quickly to address new needs or problems.
Explore automated analysis tools
With millions of reviews and interactions, manual reading is unfeasible. Artificial intelligence tools like those used in this study can help identify and summarize conversations about gender at scale, not only in reviews, but also in forums, support tickets, and other user communication channels.
Consider the app category
Gender conversations are not the same across all app categories. In video apps, the focus may be on representation and content; in games and social networks, the emphasis often falls on community; in transportation services, safety may be central. Understanding these differences helps prioritize improvement efforts where they will have the most impact.