What if I'm Wrong

December 28, 2022

I've spent almost all my career working at startups (so far) and we are always venturing into the unknown. Everything is an evergreen deployment. Company processes are being built from bottom up. The product is constantly changing trajectory. We're doing whatever we can to hit to arbrarily set milestones for the company. Everything is effectively being bulit under a handful of opinions from a select set of people.

And this trickles down as the company grows. Engineering leaders make their best judgement for technical direction. Product leaders make their best judgement for feature development. Ops leaders makes their best judgement for company wide operations. But what if they're wrong? No human is correct 100% percent of the time. What can leaders do to help mitigate the cost of being wrong?

What if we remove the leadership context here and just focus on how we might navigate uncharted territory in general. Let's start from the most naive strategy. We try a thing, see how it did, recalibrate and try again. Maybe we can give ourselves a better starting position because of the domain knowledge we've collected over the years. And sometimes when we recalibrate, we might be backtracking -- and that should be an accepted strategy of charting the unknown. Imaging trying to navigate a maze, but we're not allowed to backtrack. That's effectively an all-or-nothing strategy. It's all-in on the first, initiall decision -- and if it's wrong, we fail.

Now if we step back into the context of leadership, any sign of backtracking is seen as a wrongdoing. I don't believe that any amount of signal is going to give us a perfect insight -- we work with probabilities instead (as we do with just about everything in life). What is difficult is that a leader makes decisions on behave of many people, and its become absolute blasphemy to be wrong, ever. Part of the issue is that any decision made by a leader is levered by other folks. If a founder's founding thesis is wrong, backtracking means putting everyone at the company out of a job. But on the flipside, it should never be a massive surprise. Company progress should ba constantly tracked. Large initatives should be broken down. To go from "everything is fine" to "nevermind" may point to a never-backtracking mindset.

There are also just some things in life that aren't reversible or shouldn't be and that's okay too. No analogy is perfect. Look at Zuck going all-in on VR. It's a thesis that may take a long time to pan out. It's a large gamble and it's hard to imagine as a bystander of how exactly Zuck should've broken down the VR execution. But one of the advantages of software is that its incredibly easy to build and tear down. Companies should lean into that trait of software and harness it to allow the company to also be enabled by fast feedback cycles.

Effectively, I believe that the most effective method of executing (particularly in a startup sense) is to optimze for quick iterations (I'm sure this will look very different based on context), and allow backtracking (i.e canceling projects*, reverting changes, going back to status quo). Everything I listed for backtracking hurts, but it's better to go in reverse if we're headed for a cliff.

* Google might be an exception here because they don't seem to learn at all from all their canceled projects. There is never a silver bullet!