Things want to work, not punish errors

February 27th, 2017

For better or worse, things want to work.

Consider driving at night on unlit, curvy mountain roads, at a speed about twice the limit, zigzagging between cars, including oncoming ones. Obviously dangerous, and yet many do this, and survive. How?

The road, the car, you, other drivers, and their cars all want this to work. So for a long while, it does, until it finally doesn't. I know 3-4 people who drive like this habitually. At least 2 of them totaled cars. All think they're excellent drivers. All have high IQs, making you wonder just what this renowned benchmark of human brains really tells us.

Now consider a terribly managed project with an insane deadline, and a team and budget too small. All too often, this too works out. How?

The fact is that by cutting features, having a few non-terminal bugs, and being somewhat late and over budget, most projects can be salvaged. In fact, when they say that "most projects fail," the PMI 1 defines "failure" as being a bit late or over budget. If "failure" is defined as outright cancellation, I conjecture that most projects "succeed."

Which projects are least likely to be canceled? In other words, where is being late, over budget and off the original spec most tolerable? Obviously, when the overall delivered value is the highest, both in absolute terms and relatively to the cost. In other words, reality punishes bad management the least in the most impactful cases.

What is the biggest problem with bad management? Same as crazy driving: risk. The problem in both cases is you risk high-cost, low-probability events. It's terrible things that tend not to happen. And people are pretty bad at learning from mistakes they never had to pay for.

Wannabe racecar drivers fail to learn from driving into risky situations which their own eyes tell them are risky. For managers, learning is harder – the risks accumulated through bad management are abstract, instead of viscerally scary. In fact, a lot of the risks are never understood by management, or even fully reported. There's just too much risk to sweep under various rugs to make it all ingrained in institutional memory.

In fact, it's even worse, because risk-taking is actually rewarding as long as the downside doesn't materialize. The crazy driver gets there 10 minutes earlier. Similarly, non-obviously hazardous management often delivers at an obviously small cost. And while driving is not actually competitive, except in the inflamed minds of the zigzagging few, most projects are delivered in very competitive environments indeed. And competition can make even small rewards for risk decisive – as it can with any other smallish factor large enough to make a difference between victory and defeat.

Things want to work more than they want to punish us for our errors. The punishment may be very cruel and unusual alright, but it's rare. It seems that the universe, at least The Universe of Deliverables, is Beckerian. It delivers optimal punishment for rational agents correctly estimating probabilities. Sadly, humans are bad at probability.

And thus crazy drivers and bad managers alike (often the same people, BTW) march from one insane adventure to the next, gaining more and more confidence in their brilliance.


  1. PMI (The Project Management Institute) is a con, where they sell you "PMBOK" (Project Management Body of Knowledge, a thick book you can use as a monitor stand) and "PMP" (Project Management Professional, a certification required by PMI's conscious or unwitting accomplices in dark corners of the industry.) A variety of more elaborate cons targeted at narrower audiences incorporate PMI's core body of cargo cult practices.↩︎