When the Star Player Leaves

The Day Jamie Disappeared #

Tribal knowledge as a concept has always bothered me, even as I respect it as a force to be repeatedly reckoned with, like gravity. It's everywhere, all around us. In scandals about expertise on a national level ("do your own research" → okay, but what if I can't really?), and confusion sitting right at the desk next to you. You encounter it all the time, and the answer is almost always some version of "document it."

It's a good response to a degree, and probably there's nothing else to do, but what bothers me is that whether it's already been documented or not, half the time the document only makes sense to the person who wrote it in the first place, or the person maintaining it, or the person who worked at the place long enough to know where the organizational bodies are buried.

"Check Edge Case 3"

Me: "Great, what the hell is Edge Case 3?"

It's what tribal knowledge is to me. It's something between the steps to follow to do something and how to apply judgment.

I had a colleague who was hit by a bus once. This is real, it happened, and I couldn't believe the proverbial reason to actually document something actually happened in real life. They were banged up but fine. It was unbelievable in multiple dimensions.

That person had been there since the early days and had absorbed every exception and workaround, every "just this once" decision that quietly became the process. It was a natural accumulation, then boom: bus.

Out for a while. Obviously.

The first few days passed without incident. A few business days later, some of the warning lights were going off. Old Slack threads were reopened and read like artifacts. Decisions made a year earlier had to be reconstructed from fragments of conversations between people who no longer worked there.

I don't really have an answer to the right or best way to do this, because there were notes, there was capital-D Documentation. Despite their precision, they were written inside the system of one human mind, or maybe they were written without being able to truly convey the reasoning behind the steps. "Check Edge Case 3" first is technically documentation. But it isn't enough for your encounter with Edge Case 3, which was clear to exactly one person and opaque the moment that bus did what it did.

Edge Case 3 #

The other thing I learned from the actual bus was that the famous "bus factor" sounds much more dramatic than it looked or how I'd been describing it before the event. It was one of my favorite things to say: "Well, what if so-and-so gets hit by a bus?" What if this really, absurdly horrible thing happens? It wouldn't actually happen. I expected a cliff. Instead we got a slow decline into friction.

Things got done. But they took longer, as we Rosetta Stoned our way through notes and how-to's. A question that used to get an immediate answer became one to three people trying to piece together why something worked the way it did from an old message, half a story told at lunch, an email with a question mark. Eventually the thing would move again.

Which is somehow more annoying than the system simply breaking. A broken thing announces itself with alerts. But a lot of stuff can limp around looking healthy for a long time until surprise: it isn't working anymore.

Eventually the person hit by a bus healed, came back. And the red and yellow warning lights shifted back to green. We could stop diving through old Slack threads like work archaeologists. “Check Edge Case 3” once again meant something because the person who knew what Edge Case 3 was was sitting there again.

Nothing much about the system had really changed while they were out. We got the person back, which was comforting. And, in retrospect, probably the problem.^[1]