This is the toughest principle to apply effectively. But consider the old maxim: Half the solution is in the statement of the problem. The purpose of PrD is to get key internal stakeholder assumptions out in the open, and get these assumptions rigorously tested out in the field without creating a bona fide solution.
What is the cost of a decision that harms a few thousand people compared to a decision that irritates millions? Ellen Pao talks about problems of scale to The Washington Post “If mistakes are made 0.01 percent of the time, that could mean tens of thousands of mistakes.” So much of an emotional message is nonverbal, and a computer can’t pick up those cues. Designers can’t see people’s expressions or body language, but we can try to understand reactions by paying very careful attention to what people do, particularly in aggregate. Numbers often don’t mean what we think they do, they tell us the what but not the why. Metrics are a great way to come up with assumptions to be challenged, but don’t say a lot without context. If someone is spending more time in a product, is it because the product is great or because the person can’t find what they need? A friend of mine used a feature on a professional networking site that scraped her address book for contacts. She didn’t know the service would automatically connect her to every person in her inbox, an inbox that spans a decade. She was suddenly connected to a person she never wanted to speak to again. Somewhere in a data warehouse, the volume of connections on this service appears to be going up, which superficially seems good. However, the quality of those connections is diminishing. How many of these people actually want to be connected? How many are actually going to return to the site because they were connected to someone they exchanged an email with 3 years ago? That number going up isn’t very meaningful.