The Decision Log: Why We Record Decisions We Later Reverse

Most decision logs are highlight reels

The instinct, when you keep any kind of record of what a company decided, is to only write down the decisions that worked. A list of good calls reads like a confidence-building exercise. It also isn't useful to anyone -- including us.

The decision log we actually keep includes the decisions that were later reversed, and says so explicitly. Not as a confession, but because a decision log that only shows wins isn't a record of how decisions get made -- it's marketing.

What goes in it

Any decision that affects more than day-to-day execution gets written down: the situation at the time, what was decided, what alternatives were considered, and what it changes going forward.

That's a deliberately narrow bar. Most day-to-day engineering choices don't need a log entry -- which library, which naming convention, which component to reuse. The log is for decisions that would be expensive to get wrong silently: what stack to default to across the company, how to structure ownership between divisions, when an idea is allowed to graduate from an experiment into a real product.

Why bother

Two reasons, both practical rather than philosophical:

For us: when a decision gets revisited six months later, the log says what was actually considered at the time -- not what we'd like to remember considering. That's the difference between "we already thought about that and here's why we didn't do it" and re-litigating the same argument from scratch.

For anyone evaluating us: if a claim on this site should be backed by something concrete, the decision log is usually where it lives. "We take a limited number of engagements at a time" isn't a slogan -- it's a standing decision, recorded, that we'd have to actively reverse to walk away from.

The honest part

Some entries in our own log describe decisions we made and later undid. We don't think that's a weakness worth hiding. A process that never produces a documented reversal either isn't being tested against reality, or isn't being recorded honestly.

If you want the practical version of this idea rather than the philosophy, see How We Work for where this fits into the rest of how an engagement actually runs.