The Privacy Panic: Analyzing User Sentiment and Data Trust in FemTech Applications

FemTech apps designed to track periods and fertility collect some of the most personal and intimate health data available on mobile devices, but users often do not understand or fully realize the privacy practices tied to that data. This thesis looks at the Say–Do Gap in FemTech privacy, which refers to the gap between what the applications claim in their privacy policies and what users actually believe and trust in. Using a corpus of 65,322 user comments from Google Play, the App Store, and Twitter/X relating to major apps - Flo, Clue, Natural Cycles, Premom, and Stardust - we determine what users talk about using BERTopic and assess the sentiment with GPT-4o-mini, a result that has been independently checked against a RoBERTa baseline (Cohen's κ = 0.73). The privacy policies are evaluated on six different criteria using a GPT-4o-mini pipeline, and the SAY and DO measures are then combined to form a standardized (z-score) Say–Do Gap. The policies reveal consistent flaws in terms of data retention and sharing with third parties; user sentiment is strongly affected by specific events, such as the 2022 Dobbs decision related to cancelling the right for abortions, which causes it to spike, and the quality of the policies does not correlate with user trust - Clue has the best policy but still has a large positive gap, while Natural Cycles is trusted to a greater extent than its policy alone would suggest. When users take action, they mostly switch to another app or delete it rather than file a complaint or try to resolve it through legal channels. A complementary survey (n = 101) provides self-report evidence that supports these findings and confirms a clear privacy paradox: high levels of concern go hand in hand with a lack of protective action.