The Ethics of the Data We Collect

Teresa Torres argues that product builders rarely stop to consider the ethics of the data they collect, even though user tracking has grown far beyond the cookies that alarmed people in the late 1990s. She notes that most product people intend to help users, but good intent doesn't settle whether users would accept the data collection or its uses. Using the backlash to Facebook's emotional contagion study as a case, she shows how a data practice can be technically approved and well-meant yet still cross a line for users. She concludes that the industry must proactively help users understand what is collected and how it is used, and respond when they say it has gone too far.

01Key takeaways

• Audit all the tracking your product does, including analytics, ad, and experimentation tools, not just what sits in your database.
• Ask whether users would be comfortable if they knew the full extent of data collected about them.
• Ask whether they would still be comfortable knowing exactly how that data is used to shape their experience.
• Recognize that business targets like revenue per user or engagement can conflict with what users actually want.
• Assume positive intent in colleagues, but don't let good intent replace careful review of user reactions.
• Treat data-collection policies like any product feature: test them, gather feedback, and adjust when users push back.

02Key sections

From cookies to constant tracking

The essay contrasts the public panic over early web cookies with today's far more extensive tracking that most people never consider. It highlights how industry insiders accept tracking as normal.

Good intent isn't the whole story

Torres acknowledges that collecting data usually aims to improve products and that most builders would never intend harm. She argues that this good intent is exactly why teams fail to examine broader implications.

Two questions to ask

She proposes asking whether users would be comfortable with the full scope of data collected and with exactly how it is used. She also questions whether business metrics like ARPU and monthly active users always align with user interests.

The Facebook study as a cautionary example

The essay examines how Facebook adjusted its newsfeed based on emotional word frequency and why users reacted negatively. It stresses that the people involved were ordinary product teams, not villains.

Our shared responsibility

Torres calls on the industry to apply the same iterative testing and feedback practices to data policies, helping users understand data use and listening when they object.

03From the post

“Do you remember the uproar about cookies in the late 90s? I don’t mean the sugar-filled kind. I’m talking about web cookies - the bite-sized bits of information that your browser stores locally so that when you return to a website it already knows who you are.”

Summary and takeaways written by PM Atlas; quotes are short excerpts. © the original author.