{
  "$type": "site.standard.document",
  "canonicalUrl": "https://jacob.blog/notes/goodharts-law",
  "description": "Why velocity, coverage targets, and badly written OKRs invite gaming once they become the scoreboard.",
  "path": "/notes/goodharts-law",
  "publishedAt": "2020-04-05T00:00:00.000Z",
  "site": "at://did:plc:ckthoyuvsmkp254fyuinyzb2/site.standard.publication/3mndm6tiamb26",
  "tags": [
    "software-engineering",
    "metrics"
  ],
  "textContent": "At a summer scout camp, I was part of a fire-making competition. The goal was to burn through a taught rope above the fire pit as fast as possible. Most scout groups started collecting firewood and building robust wood pyramids. My group grabbed as many dead leaves as we could find, lit our pile, and burned through our rope before. We were done before other groups had even gotten their rope warm.\n\nOf course the idea behind the competition was to get the scouts to practice building fires. But the _measure_ that was selected as the _target_ (the time it took to burn through the rope) meant we were technically right to optimize for fast burning. Our fire was terrible by all reasonable standards. But it did technically meet the target.\n\n---\n\nLeaders reach for numbers because numbers feel legible. The trouble is that people are clever, and they’ll optimize for the target.\n\n> When a measure becomes a target, it ceases to be a good measure.\n\nThat line is the usual shorthand for Goodhart's law. A few inputs drive most outcomes until you promote one of them to the scoreboard.\n\nThe right answer here is to use measures as inputs alongside qualitative data to inform decisions.",
  "title": "Goodhart's law and engineering metrics"
}