Best Practices Guarantee Average

Picture two engineering organizations. Same standup format, same sprint cadence, same code review gates, same DORA dashboard on a screen nobody looks at. Both teams run the practices correctly, and I mean genuinely correctly, not the cargo-cult version. Neither team is remarkable. Neither is bad. Both ship at roughly the rate the industry ships, with roughly the industry’s defect rate, and their retros produce roughly the same action items.

Now multiply that across the whole field. The best practices won. Adoption is near total, the tooling assumes them, and the conference talks have moved on to refinements of refinements. If those practices produced excellence, following them should be an edge, and an edge is by definition rare. Instead, following them is the single most common configuration in the industry, and the results cluster in a band so tight you could lose your own team’s metrics in a lineup.

Something in the promise does not survive contact with the arithmetic. That gap is worth investigating, because what sits underneath it changes what you should do with every standard you have ever inherited.

The pattern shows up everywhere you look

Software is the easy case, so start wider.

Hiring. Over the past decade nearly every serious company converged on the same structured loops, the same rubrics, the same behavioral questions, most of it borrowed from the same five famous employers. The instrument was designed to reduce variance, and it does. The same leaders who installed it then complain that they cannot seem to find exceptional people, using a process built specifically to sand off the tails.

Content. The hooks that “perform” are documented in a hundred threads, the posting cadence is settled science, and the result is a feed of interchangeable competence that readers scroll past at highway speed. The advice all works, in the sense that it reliably produces what everyone else is producing.

Then the case on my own desk. When coding agents landed, the prompt guides converged within months: same patterns, same scaffolding advice, same context tricks. I read them, and they were fine, and they got me to fine. Every result I actually care about came later, from hours of working the tools against my own repositories until something held up that no guide had suggested, including a few things the guides mildly warned against. The guides were the on-ramp. The interesting road started where they stopped.

And across all of these domains, listen for one tell. Notice how rarely anyone defends a best practice with recent local results. The defense is adoption itself, some version of “this is industry standard,” delivered as the end of the argument. Which, as it turns out, is a clue to what these things actually are.

What a best practice actually is

Strip the branding and look at the mechanism. A best practice is an approach that worked in enough other contexts to survive averaging. It gets written down only after the consensus has already formed, because that is what makes it citable. And to travel at all, it has to be sanded of every contextual edge: the team it assumed, the constraints it answered, the taste of the person who made it work. What ships is the part that generalizes.

By construction, then, a best practice is the trailing average of other people’s contexts, a floor formalized rather than a summit. Follow it correctly and it will deliver you to the average of everyone else following it, which is real value. Floors matter. If your incident response or your payroll runs below industry baseline, adopting the standard is the fastest improvement available to you, and pretending otherwise is vanity. The overpromise lives entirely in the name. “Best” was never an honest word for “what we could all agree on after it stopped being novel.”

The part that gets stripped in the writing-down has a name, and it is old. Aristotle called it phronesis, practical wisdom: the capacity to perceive what this particular situation requires, as opposed to knowing what situations in general require. He insisted it cannot be learned from a textbook, and his reasoning maps onto this problem with no translation needed. Documents traffic in universals, while situations arrive as particulars, one at a time, each carrying context the general rule never saw. The judgment that made the original practice work in its original shop was somebody’s phronesis, and it is exactly the part that does not survive being turned into a checklist. The best practitioners cannot even hand that layer to you in conversation, let alone through a PDF. What they wrote down is what could be written down.

There is a second function at work here too, and it is the quieter reason these things spread. Adopting the standard is defensible. When the standard fails, the method absorbs the blame, and the practitioner walks away clean. Also smaller, because what got surrendered in the trade was authorship, the only part of the outcome that was ever fully theirs. Deviating means the outcome carries your name. A great deal of best-practice adoption is insurance rather than a bet on results, and the premium gets paid out of the upside.

What that means if you want to be better than fine

Three implications fall out, and none of them are comfortable.

First, the floor is where the competition is thickest. If every team in your market holds the same baseline, then working harder at the shared playbook buys you nothing but parity maintained at higher cost. You are sprinting to stand still, alongside everyone else who read the same books.

Second, differentiated results cannot come from shared method. This is close to a tautology and still routinely ignored. Whatever your team ends up known for, it will be something no one handed you. Even a playbook imported with perfect fidelity can only reproduce what its authors already have, and the harder version of that truth is that a successful import buys you their baseline, not their edge. The tempting reply is to run the shared playbook better than everyone else, and the reply defeats itself. Ask what “better” consists of at the same standup with the same rubric, and the answer is judgment in the moments the document does not cover. The difference was never in the shared part.

Third, the recipe’s edge cases mark the excellence zone. Every place a practice says “it depends” is a place where a practitioner with judgment beats a practitioner with the document. And be precise about what excellence means here: the activity itself, done at the level this particular case demands, judgment included. Being known for it arrives later, as a byproduct. The standard maps the settled territory. The results that build reputations happen off the map, and the map is honest about where it ends if you read it closely.

What to do with the standards you have

Keep the floor. Most of your operation should sit on it, on purpose. Payroll, security hygiene, on-call handoffs, the plumbing that only gets noticed when it fails: run the standard, be average there, and spend nothing extra. Deviating everywhere is chaos with a manifesto.

For everything you keep, pass one test: can you explain why it works? A rule whose mechanism you understand is a rule you will know when to break. Rules have premises, and premises expire. The practices you cannot explain have stopped being decisions at all; they are habits with a citation attached.

Then choose your deviation deliberately. One or two places, the ones that decide what you are known for, where you go non-standard with your eyes open and a name on the outcome. Yours, ideally. That named ownership is the exact cost the standard was shielding you from, which is how you know it is the price of the upside.

Be honest about the aim while you are out there, because it decides whether any of this holds up. The reason to work past the standard is the work itself, done the way this particular situation demands. Being preferred is what the market pays you for that when it notices, and the standard you hold cannot depend on it noticing. If the applause stopped tomorrow, the obligation to do the work at the level it deserves would not.

And build the judgment that deviation runs on, because there is only one source. Reps. Decisions made where the recipe runs out, outcomes examined honestly, over years. That is how phronesis forms, in Aristotle’s account and in every shop I have ever watched produce something worth copying. The unwritten layer does not live in a better book, because there is no better book.

One last test for daily use: the next time a practice ends an argument with “this is industry standard,” ask what result it produced here, lately. Keep the ones that can answer, and put the ones that can only cite their own adoption back on probation.

Final thoughts

The best practice was written to get a stranger to competent in an unfamiliar context, and it does that well. Respect it for what it is. A field’s consensus, executed harder, has never once made anyone the thing their field talks about. The organizations that did become famous for their standards, Toyota with its production system, McDonald’s with its line, are the ones that wrote them. The fame belongs to authorship. By the time the practice reached everyone else, it carried somebody else’s signature.

Follow the recipe wherever average is acceptable, which is more places than pride wants to admit. Where average is not acceptable, the recipe was never going to be enough, and the sooner your calendar reflects that, the sooner the work that could distinguish you gets the hours it needs.

FAQ

Why do best practices produce average results?

Because of what a best practice is mechanically: an approach that worked in enough other contexts to survive averaging, written down after the consensus formed, with the contextual edges sanded off so it can travel. It encodes a field’s trailing consensus, which makes it a reliable floor and nothing more. Following it correctly delivers the average of everyone else following it. Nothing that everyone executes can separate anyone.

Should teams stop following best practices?

No. The floor is worth having, and most of any operation should sit on it deliberately. Run the standards wherever average is acceptable: payroll, security hygiene, incident handoffs, the plumbing. The discipline is choosing the one or two places that decide what you are known for and going non-standard there on purpose, with a named owner for the outcome. Deviating everywhere is chaos with a manifesto.

What is phronesis and why can’t it be written into a best practice?

phronesis is Aristotle’s practical wisdom, the capacity to perceive what a particular situation requires and act on it. He argued it cannot come from a textbook, because documents deal in universals while situations arrive as particulars carrying context the general rule never saw. A best practice is universals on paper. The judgment that made the original practice work is the part the writing-down strips out, and it only rebuilds through reps: real decisions, honest review of outcomes, years of both.

How do you go beyond best practices without being reckless?

Three tests keep deviation honest. Explain why the standard works before you leave it, because you have to understand a rule’s mechanism to know when it stops applying. Deviate in one or two chosen places rather than everywhere, with a name attached to the outcome. Judge the deviation by results in your context, the same standard the practice itself was supposed to meet and often can no longer. Recklessness is deviation that nobody signs.

If you are doing the reps that build judgment no playbook carries, and you want to do them alongside people holding the same bar, that is what MasteryLab is for.