Start at the End: How to Build Products That Create Change

Since we’re picking on tech companies, let’s go with another fan favorite: Facebook. In June 2014 they, along with some Cornell researchers, published a paper10 that revealed a massive intervention in which the company manipulated the contents of users’ newsfeeds to contain either more positive or more negative content, which then caused users to p

When we focus on novelty, we lose sight of behavioral outcomes. And while using validated pressures to inform our designs can help guard against the natural tendency to pursue interventions that feel unique, they aren’t a guarantee; just as data can be used to justify bad business decisions, pressures can be used to justify bad interventions.

we can do this by ramping up or down all sorts of inhibiting pressures. Physical availability is one factor, but so is psychological availability.

There are also counterrational pressures at play on the inhibiting side, like whether or not branding fits a context. Lighthearted and fun is great for kids, but picture a romantic Valentine’s Day dinner. Candlelight, filet mignon, roses, red wine, and for dessert . . . M&M’s. It just doesn’t work. That’s a Lindt moment, a Ferrero Rocher moment

“How much more likely does the flu shot letter make people to get a shot?” and “How can we be sure that this will generalize beyond these two hundred people?” This is why there are two numbers your quantitative research is likely to report: an effect size and a p-value. Effect size answers the first question: it tells you whether the letter was rea

you can’t force insights, but you can create an environment in which they are more likely to happen.

If each researcher works independently before they come back together, there is less of a tendency to cheat and proceed with an unvalidated insight or to reach group consensus too early.

you can look at whether a pilot’s failure was singular or part of a larger pattern.

Research papers are good for external insights, but so is simply cross-pollinating with other industries and disciplines.

Pilots are tightly scoped interventions that we expect not to work (remember, we have to explicitly prove efficacy as a defense against confirmation bias), so we use small populations, focus on speed to market, and do them in an operationally dirty way.