Forty Volunteers and a Dragon.

My parents were teachers, but the house ran on what we called “automation” back then.

My mother taught WordPerfect 5.1. My father built a new stream inside a Dutch teacher-training college on how people and computers get along with each other. So the Commodore 64 in the house wasn’t a toy — it was the family trade, one generation early.

I went the way you’d expect: a little programming, a lot of games. Dune II, Warcraft, SimCity, Commander Keen. Then the internet rearranged everything — Quake 2 and Diablo became the evenings, StarCraft the weekends. On metered dial-up you planned matches like they cost money, because they did. By sixteen I had a second and third set of friends I’d never have met otherwise, and a 2v2 and 3v3 Quake habit that kept me, on a good season, somewhere in the top tenth of the ladder.

World of Warcraft is where it stopped being a hobby and became an education. I joined a raid almost by accident, and then I was one of forty people learning a single boss one wipe at a time. The guild was A Few Good Men, run by a Swedish veteran who went by Hellfish. The core got tighter, the nights got longer, and over a couple of years a group of unpaid volunteers became the top guild on the realm — and then a name on the global boards.

None of us were paid. Nobody could really be fired. And it remains one of the highest-performing teams I’ve ever been part of. I’ve spent twenty years wondering why.

One — we built an economy, and what we rewarded wasn’t what you’d guess.

When a boss dies it drops a handful of items for forty people. How you hand those out quietly decides whether the guild survives the month. The common currency was DKP — Dragon Kill Points: you earn them for showing up and performing, you spend them on loot. We built our own version, and three design choices in it have stayed with me longer than anything I learned in a classroom.

• Priority, not price. Your accumulated points never made loot cheaper — everyone paid the same flat price for the same item. What a big balance bought you was priority: a multiplier on your standing that decided, when two people wanted the same drop, who was first in line. Loyalty moved you up the queue; it didn’t let you extract more for less. Most companies do the opposite. They reward tenure with a better deal — and then wonder why it curdles into entitlement.
• Inflation correction. Points quietly decayed over time. A balance you’d banked three months ago didn’t let you coast; standing tracked what you’d done lately, not what you’d done once. Most organisations let reputation compound forever, and end up quietly governed by people living off a single old win.
• A starting pack. New members were handed a stipend so they could gear up and contribute fast, instead of clawing from zero behind everyone else. We invested in the newcomer’s floor on purpose — because the floor, not the ceiling, is what wins.

The throughline: a dragon doesn’t fall because one person was brilliant. It falls because forty people clear a minimum bar at the same instant. So we built the economy to keep that floor high — reward loyalty with priority and trust, let status decay so it stays honest, lift newcomers fast. Durable performance is a property of the floor. Almost every pay plan I’ve seen since optimises the ceiling.

Two — measure everything, blame no one.

Full logging became non-negotiable early: every action by every player, recorded. If a healer missed a dispel at second forty-two, the log showed it plainly — not to point a finger, but so we’d know to move someone two steps left on the next pull. That’s the whole distinction: the log existed to fix the next attempt, never to litigate the last one. A mistake had no consequence beyond the learning. You usually saw your own error before anyone else mentioned it, fixed it, and moved on. Total transparency and total safety, at the same time. Most organisations manage exactly one. They measure people and use it against them, so the numbers get gamed — or they stay kind by staying vague, so nobody improves. Our data was safe, so our data was honest, so we got better. That is the entire trick, and almost nobody runs it.

Three — the bar to stay was the brand, and it enforced itself.

The guild was attractive to join precisely because it was hard to remain in. “Having played for AFGM” was a portable credential — it opened nearly any door on the realm. But the interesting question is the one a manager always asks me here: how do you hold a high bar when you can’t fire anyone, can’t dock pay, can’t do anything except not invite them back next Tuesday?

The answer is that the bar enforced itself, through three quiet mechanisms. First, the logs made contribution public — being carried wasn’t a secret you could keep; it was visible to all forty, and that was its own slow, unbearable pressure. Second, the roster was deliberately bigger than the raid, so a seat was contested every single night; a spot you can lose is a spot you respect. And third — the part that surprised me most — almost all of the correction was self-correction. People benched themselves before anyone asked. A player going through a heavy stretch at work would message an officer and step down to the bench for a month, on their own, because they could see in the logs that they weren’t holding their end and cared too much to drag the night down. The formal conversation — always from an officer, always framed as “you’re not getting the reps right now,” never as punishment — was the rare exception, not the tool. When belonging is the reward, the cost of slipping is losing the thing you value most. You almost never have to reach for a stick that isn’t there.