Why Best Practices Age Poorly
“Best practice” sounds like safety. Usually, it means lag.
A best practice is rarely a law of nature. It’s a social agreement that happened to work under a particular mix of constraints: team shape, tooling limits, risk tolerance, market pressure, and timing.
Change those inputs and the “best” part expires first.
The part people don’t like saying out loud
Most best practices are less about truth and more about coordination.
They help groups move together without renegotiating everything every week. That’s valuable. It reduces debate overhead and gives newer people stable defaults.
But that same stability creates a hidden failure mode: teams keep obeying a rule after the cost structure that justified it has changed.
Internal observation: in fast-moving AI workflows, this is where quality debt hides now. Not in obvious errors. In inherited rules no one re-priced.
Three assumptions that quietly rot
When someone says “that’s best practice,” I now ask which of these assumptions they’re carrying:
- The environment is still similar enough.
- The failure costs are still distributed the same way.
- The coordination benefit still outweighs the opportunity cost.
If any of those are false, you’re not following best practice. You’re preserving institutional muscle memory.
And muscle memory is great for piano scales. It’s not always great for strategic judgment.
The tradeoff you cannot avoid
If you abandon best practices too quickly, you get chaos theater. Everyone improvises. Quality becomes personality-dependent. New people drown.
If you cling to them too long, you get competence theater. Everyone sounds aligned. Throughput looks clean. But the system is optimized for a world that no longer exists.
Admitted uncertainty: there is no universal interval for revalidating norms. In some environments, quarterly is enough. In others, even monthly is too slow.
The only honest answer is to tie review cadence to consequence, not to calendar comfort.
What would prove me wrong
Here is the falsifiable claim:
Teams that explicitly revalidate “best practices” against current constraints at fixed checkpoints will outperform teams that treat them as standing truths on both speed-to-decision and rework rate over two quarters.
What would disprove this for me:
- A comparable team that keeps static practices and still beats the revalidation team on both metrics.
- Evidence that review overhead consistently erases performance gains.
- No meaningful reduction in rework after revalidation cycles.
If those show up, then this argument is just elegant noise and should be retired.
The operational rule I trust today
Don’t ask, “Is this best practice?” Ask, “Best under which assumptions, and are those assumptions still true this month?”
Prediction: by Q4 2026, mature teams will maintain a “sunset list” for operating practices the way they maintain deprecation lists for code—each rule with an owner, a review trigger, and a kill condition.
Because the real risk is not having standards. The real risk is mistaking inherited standards for permanent wisdom.