Extra Strawberries
And the folly of manufactured certainty
Last week, we talked about strawberry farming and about the trickiest challenge of running a strawberry farm, which is to avoid having anyone pick the fruit for the first time. The way that the strawberry farm addresses this problem is to take really good care of its seasonal workers so that they return year after year—and we asked ourselves if this strategy could be valuable in organizations with a less precise understanding of individual productivity than the farm.
But today I want to look harder at the challenge of measuring productivity in non-strawberry-farm organizations. Reading about the details of polytunnels and kilograms per picker per hour, someone from the world of knowledge work might feel something approaching envy: the farm has at its fingertips wonderful clarity of effort and result. In knowledge work, however, this level of informational clarity is much harder to come by, and much rarer as a result.
Strawberry farms know exactly what productivity looks like, and most of the rest of us don’t.
That’s not anyone’s fault—but it does create a problem. And the obvious solution is to look for measurable things which appear close to productivity. So we look at billable hours, say, or utilization. Or performance ratings. Or office attendance. Or keystrokes. Or goal attainment. Or AI token use. Or any of a number of things that are not productivity itself, but are one step removed from it—that are productivity proxies, if you like. The prevalence of these inside an organization is a consequence of that organization’s inability to measure individual productivity directly—there is no knowledge-work equivalent to the person at the end of the polytunnel grading and weighing each tray of strawberries and allocating the result to the specific picker standing there waiting for his or her score.
But any organization that uses these productivity proxies has a choice to make, and while this choice might initially seem to be about management information, it is in fact a choice about motivation, decision quality, and trust.
The choice is simply whether or not to hold people accountable for these proxies.
If you don’t, but you simply treat them as useful information—as evidence that tells you how things are going—you’re in the clear.
But if you use the scores to make decisions about individual employees—if, say, they are used in determining bonuses, or promotions—then three things happen in quick succession.
First, people optimize the proxy rather than the underlying outcome the proxy is standing in for. And because the metric and the outcome are different, this leads to unintended consequences. This is the phenomenon known as Goodhart’s Law, after the British economist Charles Goodhart, who observed that if you judge people by numbers they can fudge, then sooner or later they will.
Second, once this happens, the proxy stops telling you what you hoped it would tell you. Office attendance no longer tells you about collaboration—just that some people have figured out coffee badging. Goal attainment doesn’t tell you about work outcomes as much as it tells you about who managed to get away with writing goals which were more easily achieved.
And third, people resent it. They sense that the organization is no longer trying to figure out what’s true, but is instead choosing winners and losers based on a stand-in. Rather than trusting them to do high quality work, it is yoking them to a metric which is meaningfully different from the work itself. Instead of trusting people to judge whether remote work or in-person work is more effective on a day-to-day basis, the organization mandates a number of days in the office. Instead of trusting people to use AI when it’s useful and not when it’s not, the company rewards those who use the most tokens. Trust erodes at every turn.
The result of this choice, then, is that you end up with lower quality information, with lots of people trying to optimize their scores, and with lots more people annoyed at you for holding them accountable for something they know in their hearts is one step removed from the quality of their work.
So, what to do?
The strawberry farm knows the answer to the question of how are we doing? at an individual level—they know how much fruit each picker has picked each hour, and how much they need to pick for the farm to make a profit. For knowledge work, this degree of individual precision is impossible to achieve.
So we should stop looking for it. The proxies, when applied to individuals, are an attempt to answer the unanswerable, and they are making things worse. In a complex and dynamic world, where teams and technologies and context change constantly, and where it’s practically impossible to parcel out credit or blame along the way, we should attach accountability to outcomes, not to echoes of outcomes; and if there are no individual outcomes to measure, then we should attach accountability only to collective outcomes. This is the real world of work we’re familiar with; it’s time for our measures to catch up.
And then, we should realize that just behind the question of how things are going is a different one: how can we get better? This is arguably much more useful—and it is also readily answerable: ask the people who are doing the work, and listen to what they have to say.
Once you do this, a lot of pseudo-rigor and proxy-complexity melts away. Trust returns, accompanied by deeper insight into what the business needs. Most people want to do a good job—it’s fun to get better!—so let them tell you what they need in order to do their work more effectively.
The irony is that today, we don’t get to the how can we get better? bit at all, because we are so caught up with the tortured attempt to know, for sure, which teams are good and which are bad and whether we have the right people in the right jobs. We envy the strawberry farm because we envy its certainty, and we devote considerable effort to creating our own.
But the right response to uncertainty in one part of the system is not to invent a proxy for it, but rather to understand where else certainty—or at least helpful information—may be had. Some worlds permit precise measurement; some don’t; and to wish either of these to be the other is folly. The ultimate leadership lesson of the strawberry farm is not to force certainty where none exists, but rather to learn how to make good use of the information that is already at hand.
A quick programming note: I’m off for a couple of weeks’ vacation, so The Second Circle will be back sometime next month.