I started a weekly habit of specifically reviewing what went wrong across the ventures, mislabeled outputs, things that failed silently without anyone noticing right away, tasks that got left half-done and quietly slid off the radar. Not a general status review, a review whose entire purpose is looking for the mistakes. It's an uncomfortable habit to keep on purpose, and it's turned out to be one of the more useful ones I've built this year.
Why I wasn't doing this already
My default review rhythm, like most people's, leaned toward checking what got done and what's moving forward. That's a natural thing to want to look at, it feels good, and it's genuinely necessary to know where things stand. But a review built only around progress has a blind spot by design, it's structurally unlikely to surface the things that quietly went wrong, because going looking for what went wrong isn't the question that kind of review is asking.
Mistakes don't usually announce themselves. A mislabeled output looks fine at a glance. A silent failure, by definition, doesn't trigger anything that would draw attention to itself. A half-finished task can sit in a state that looks superficially similar to a task still in normal progress. None of these get caught by a review that's scanning for what's going well, because they're specifically the things that don't look like a problem unless you're actually looking for one.
What the weekly review actually looks like
Once a week, separate from any progress or status review, I go looking specifically for things that broke, were wrong, or got left incomplete without anyone deciding to leave them that way. Not "what's the status of everything," but "what happened this week that shouldn't have happened, and did I actually notice it when it happened or only later." That second half of the question turned out to matter more than I expected.
The pattern that surprised me
The recurring finding wasn't dramatic failures. It was small, quiet ones, an output that got labeled incorrectly and nobody caught it until it mattered elsewhere, a task that got marked as handled when it had actually just been deferred without anyone deciding to defer it on purpose. Individually, none of these were serious. Collectively, they told me something important: most of my actual failure points weren't big, visible breaks. They were small, silent ones that could sit unnoticed for a surprisingly long time before they caused a real problem downstream.
Why this changed how issues get caught, not just how they get counted
The direct value of finding these things a week later instead of never is obvious. The bigger value has been upstream of that. Knowing that these small, silent failures are the actual dominant failure mode changed what I build checks for in the first place. I stopped designing verification steps around the failures that are easy to imagine, the dramatic, obvious ones, and started designing them around the quiet ones that are actually showing up in the weekly review, mislabeling, silent incompleteness, things marked done that weren't actually done.
That's a meaningfully different design target, and it only became visible because I was regularly looking specifically for what went wrong instead of assuming I'd notice it in the normal course of checking on progress.
The discomfort is the point
I won't pretend this review is enjoyable. A weekly ritual specifically built around finding your own mistakes doesn't feel good, and there's a real pull toward softening it into something more like a general check-in that happens to occasionally surface a problem. I've tried to resist that pull, because the value of this specific habit comes entirely from it staying pointed at the uncomfortable question. The moment it drifts back toward a general review, it stops catching the things a general review was already missing, and I'm back where I started.
What I'd tell someone building a similar habit
Keep the mistake review completely separate from the progress review, even though it's tempting to combine them into one weekly session. They're answering different questions, and a combined review tends to quietly default toward the more comfortable one. Ask specifically what went wrong, what almost went wrong without you noticing, and what got left half-done. The answers won't be dramatic most weeks. That's not a sign the review isn't working. It's usually a sign you're finally looking in the place where your actual failure modes live.
