Leadership Garden
I caught myself feeling good about a dashboard last quarter.
Deploy frequency up. Cycle time down. Every tile is green. I sat in the review and felt that small clean hit of satisfaction, the kind that tells you the work is going well. Then it occurred to me that I had no idea whether any of it was true. I knew the numbers were up. I didn’t know if the engineering was good.
A writer who goes by german s. recently named the thing I was doing to myself. He calls it dopamine fracking: pumping disproportionate resources into something layered and complex to forcefully extract the purest, most concentrated hit, with no regard for anything but the hit. Like oil fracking, the short-term yield is intense, and the long-term damage is the whole point you’re ignoring. He wrote it as a piece about culture, hobbies, music, and the flattening of things people used to love. I read it as a description of my job.
This is for engineering leaders who measure teams for a living. Which is most of us. I want to make the case that the dashboards aren’t the real risk, and that the thing actually at stake is something you won’t see on any chart until it’s already gone.
Goodhart’s Law with a body
We already have a sterile name for half of this. Goodhart’s Law: when a measure becomes a target, it stops being a good measure. Every engineering blog has covered it (including me, from a different angle). Lines of code get padded. Velocity gets inflated. Coverage hits ninety percent with tests that assert nothing.
All true, all known. The DORA team themselves tell you not to use their metrics for league tables. If gaming were the whole story we’d have solved this years ago, because gaming is the easy part to catch. You can see padded commits. You can feel a sprint that hit its points and shipped nothing.
What “dopamine fracking” adds is the part Goodhart leaves out. Goodhart describes the mechanism. It doesn’t describe what the mechanism does to you after years of standing next to it.
The part everyone skips
german’s strawberry example is the one that stuck with me, so I’ll borrow it.
A real strawberry is absurdly complex. Hundreds of compounds, thousands of strains, every single one slightly different. You can isolate the one molecule that smells most like strawberry, synthesize it, and put it in everything from cheap candy to expensive dessert. Cheaper, more concentrated, and most people can’t tell. That’s the food industry. It’s also your metrics program.
The move that matters here isn’t the gaming. It’s what the synthetic version does to your palate over time.
First you stop noticing the difference. Then you start to prefer the concentrated hit because it’s cleaner and louder than the real thing. Then, eventually, you can’t find the real thing at all, because nobody grew it. The market filled up with the cheap version, and the actual strawberries quietly went extinct.
Run that forward in an engineering org. You optimize for the legible signal long enough, and you lose the ability to tell good engineering apart from the performance of good engineering. A green board and a healthy team produce the same dopamine, so you stop distinguishing them. You promote the people who generate clean numbers over the people doing the quiet work that never charts, because the numbers are what you can see. After a few cycles of that, you’ve selected an entire organization for how well it grows in synthetic soil. The people who knew how to grow the real thing left, or never got hired, or learned to stop bothering.
Nobody games their way there. They taste their way there, one reasonable quarter at a time.
Your judgment is the most frackable thing you own
This is where I think the existing writing stops too early. It aims all of this at engineers: how we measure them, how we accidentally train them to pad and inflate.
Turn the rig around. Point it at yourself.
The least measurable substance in the entire building is leadership judgment. It has no dashboard. There’s no tile for “made a hard call that looks wrong for two quarters and right in year three,” or “spent six months developing someone’s taste,” or “killed a project everyone wanted.” So those are the things that get fracked hardest, precisely because they’re illegible. They can’t defend themselves with a number.
What you reach for instead is the concentrated hit. The status that goes green. The reorg you can announce. The “we shipped” post in the company channel. Each one is a real, clean shot of looking effective, and each one is far easier to feel good about than the actual work of leadership, which is slow and quiet and rarely legible in the moment.
I’m not above this. I reach for the hit constantly. The Slack post feels better than the conversation. The dashboard feels better than the doubt.
Running the rig and standing under it
The trap is that you’re doing both jobs at once.
You’re the operator. Every metric you choose to elevate is a drill you’re sinking into your own teams, training them toward the signal you reward. You’re also standing under someone else’s rig, reporting up, getting fracked by the dashboards your own leadership stares at. You optimize your teams for legibility because you’re being optimized for legibility. It runs all the way up.
Some people will say the answer is better metrics. SPACE instead of velocity. DORA done properly, three dimensions, qualitative balanced against quantitative. That’s true, and I use all of it. But it isn’t the interesting part, and on its own it won’t save you. A better proxy is still a proxy. It still trains your palate toward what’s countable. You can build the most balanced scorecard in the industry and still slowly forget what a real strawberry tastes like, because the whole apparatus is pointed at the legible by design.
The fix isn’t a cleaner dashboard. It’s keeping your palate alive on purpose, against the constant pull of the convenient version.
What I actually do now
Awareness is most of it, and I want to be honest that I haven’t got this solved. But a few things have helped.
I write down what I believe before I look at the numbers. Ahead of a review, I jot what I actually think is true about a team. Then I open the dashboard. The gap between the two is the real signal, and it’s the one nobody charts. When the numbers and my gut agree, fine. When they disagree, that’s the most useful thing I’ll learn all week.
I name the illegible work out loud. In promotion cases, in calibration, in 1:1s, I force myself to articulate the mentoring, the glue work, and the well-judged no, even though none of it shows up on a board. If I won’t say it, the org won’t value it, and the people doing it will stop.
I treat the clean hit as suspect. This is the habit I took straight from german’s piece. When a number gives me a pure, frictionless jolt of satisfaction, I now read that jolt as a warning rather than a reward. Real engineering health is rarely that smooth or that loud. The frictionless hit usually means I’m tasting the synthetic version.
And I make myself eat real strawberries. I read the actual code sometimes. I sit in the actual incident. I talk to the actual customer instead of the summary of the customer. Not to micromanage, but to keep my own palate calibrated, because a leader who only ever consumes the dashboard is the first person in the building to lose the taste.
The synthetic version always wins on cost and convenience. That’s the whole problem. So if you want the real thing to survive in your org, you have to protect it on purpose, and you have to start while you can still tell the two apart. That’s the part that gets me. The window doesn’t slam shut. It closes quietly, one good quarter at a time, and one day you go looking for a real strawberry and there just aren’t any left.