Your Data-Driven Marketing Is Harmful. I Should Know: I Ran Marketing at Google and Instagram Numbers can provide great insight, sure. But human connection is still key.

I didn't know it at the time, but I was about to join my very last "performance calibration session." This was late in 2018, when I was a managing director at Instagram, and these sessions had been a common part of my life -- just like they are at many big companies. They're a time for senior managers to discuss the performance of their individual team members, applying common organizational standards across job levels. HR specialists moderate performance discussions on a biannual, sometimes quarterly, basis. Conceptually, there's nothing all that weird about calibration. But that's conceptually.

I have held senior-level marketing roles at YouTube, Spotify, Google, and Instagram, so I've sat in on a lot of these. And the reality is this: A group of highly opinionated, often outspoken managers get together in a room shielded from prying eyes. Most managers gather in a physical conference room; others dial in by phone or video, making it nearly impossible for everyone to weigh in equally. The HR representative says a few obligatory words toeing the company line, and then the verbal battle swords come out. For the next several hours, we go around the room, screens, and phone lines making the best case for why one manager's team member deserves an "exceeds expectations" rating ("She's a rock star!"), while another's should be a "meets expectations" ("He's solid but hasn't gone to the next level") or, worse, a dreaded "meets most expectations" ("Her peers sometimes find her difficult to work with"). During one particularly memorable calibration session at Google, a young man's rating was under scrutiny because his manager argued that this employee needed to "grow a pair of balls."

Every so often, HR will step in to suggest that the group bump a few people down because we're aiming to hit a normal distribution of ratings. We're not looking for a perfect bell curve -- very few people are rated at the lowest or highest ends -- but the bell can't be too top-heavy. Although positioned as an objective method to evaluate employee performance, I have found calibration to be an almost entirely subjective experience, with sometimes dire consequences. One below-average rating means less bonus money; two in a row triggers a performance improvement plan that routinely ends with getting fired.