When the number becomes the goal

The open rate went up for three straight quarters, and the business got worse the whole time.

We’ve watched a version of that sentence play out more than once, in more than one dashboard, and the strange part is that nobody involved did anything wrong. The subject lines got punchier. The send times got smarter. Every optimization was defensible on its own. And yet the emails, the actual emails, drifted toward whatever tricked the most thumbs into tapping, because that’s what the target rewarded, and the target was the only thing anyone was looking at anymore.

Charles Goodhart, a British economist, noticed this pattern in monetary policy back in the seventies: once a measure is used to steer behavior, the behavior reshapes itself around the measure and the measure stops meaning what it meant. The anthropologist Marilyn Strathern later compressed it into the version everyone quotes: when a measure becomes a target, it ceases to be a good measure. Central banks learned this the expensive way. Marketing teams are learning it now, one KPI at a time.

How a proxy dies

Here’s the mechanism, because it matters that it’s a mechanism and not a morality tale. Every metric starts life as a proxy. Open rate stood in for interest. Time on page stood in for attention. Marketing-qualified leads stood in for people who might actually buy something. The proxy was never the thing, but it correlated with the thing, so it was useful.

Then somebody put it on a slide with a goal next to it, tied a review cycle to it, maybe a bonus, and the correlation began to dissolve. Not because anyone gamed it cynically. Because smart people, asked to move a number, will find the shortest path to moving the number, and the shortest path almost never runs through the slow, unmeasurable thing the number was standing in for.

A cost-per-lead target produces cheaper leads, not better ones. A content-volume target produces volume. An engagement target, on most platforms, quietly rewards being annoying, since arguments engage beautifully. In every case the team can show you a chart going the right direction while the customer relationship goes the other way, and both charts are telling the truth.

The good faith is what makes it hard to catch. If someone were fudging spreadsheets you could fire them. Instead you have a diligent team hitting the goals you set, and the only thing broken is the connection between those goals and reality.

The scoreboard works fine. It’s just no longer attached to the game.

Keeping the signal alive

We don’t think the answer is fewer numbers. Flying blind isn’t a philosophy. But a few habits protect a dashboard from eating itself.

• Keep some metrics observed and never targeted. The moment a number carries no incentive, it becomes trustworthy again, because nobody has a reason to bend it. Some of the most honest data in any company is the stuff nobody’s comp depends on, and it’s worth deciding on purpose which numbers stay in that category.
• Pair every target with a counterweight. If you’re pushing open rate, watch unsubscribes and spam complaints beside it. If you’re pushing lead volume, watch what sales says about lead quality, in words, not scores. A metric alone can be gamed. A metric and its natural opposite have to be gamed in two directions at once, which is much harder to do by accident.
• Once a quarter, ask the proxy question out loud. What was this number originally standing in for, and would moving it still move that? Sometimes the answer is yes and you carry on. Sometimes there’s a long pause in the room, and the pause is the finding.

Targets aren’t the villain here. Teams need direction, and “make it good” is not direction. The discipline is remembering that the number was always a shadow on the wall, useful exactly as long as you stayed interested in the thing casting it.

The open rate, for what it’s worth, eventually came back down. Somebody stopped asking about it, and the emails got better almost immediately.