10 August 2010

Can we please put the whole term “Best Practices” to rest now? Apparently, according to this link (forwarded to me by John Dietz, thanks!), the very place where it originated (or was best popularized, depending on your interpretation of history) has now seen the whole concept basically debunked:

For example, Jim Collins’ blockbuster business book Good to Great, published in 2001, featured 11 supposedly great companies. All of them did extraordinarily well on the stock market for 10-20 years. But by 2008, when Steven Levitt posted Good to Great to Below Average on Freakonomics, two of them had died.
(Read more: http://timberry.bplans.com/2010/07/the-sad-truth-about-best-practices.html#ixzz0wBOxDrkh)

The point is, best practices just don’t exist. They are an attempt to take a solution to a problem out of the context and apply them across the entire spectrum, and that essentially invalidates the entire thing. A solution is only useful when considered in context.

Don’t believe me? Consider this challenge: when is it not a best practice to breathe? When you’re under water, of course.

(Unless you’re a fish. Or a frog. Or maybe a prince turned into a frog.)

Point is… context matters, folks.

Blind application of best practices don’t work, as Tim Berry’s article quotes from Jim Collins’ book (my emphasis):

Nine of the eleven companies remain more or less intact. Of these, Nucor is the only one that has dramatically outperformed the stock market since the book came out. Abbott Labs and Wells Fargo have done okay. Overall, a portfolio of the “good to great” companies looks like it would have underperformed the S&P 500.

Still think that the “best practices” idea might work? Prove it to yourself: do a detailed study of CEO performance across any given CEO’s career and across a variety of different companies who changed CEOs. In short windows, yes, you can find scenarios where a CEO had a stellar performance with a particular company for a particular period of time. But when you pan back and look across the CEO’s entire career (during which he/she practiced the same “best practices” at different firms), almost none of them have repeatable successes at a variety of different firms in any consistent manner (despite the millions handed to them).

In other words, for a good many of them, their success was nothing but blind luck. A monkey could have done just as well. The “superstar CEO” is generally a product of the five or so years in which his one firm was wildly successful.  Attempts to repeat the success at other firms (that is, in a different context) typically have failed or generated mediocre results. (Anybody know what Lee Iacocca is up to these days, he of “I will rescue Chrysler” fame? Come to think of it, can anybody name any of the 70’s or 80’s “superstar CEOs” that is still wildly successful today? Just one?)

How did this “best practices” thing get to be such a common meme? Because “best practices” mean, essentially, that the questioner doesn’t want to have to think. And it’s a seductive premise—if I just push the right buttons, type the right keywords, call the right methods and/or use the right classes, I can get something that “just works” without having to think about all those nasty little details that seem to trip people up: performance, scalability, security, blah blah blah.

Here’s the dirty little secret of our industry: Software development is hard.

Computer science is about tradeoffs and hard choices. Optimizing the code or design or architecture one way means taking hits another way. Trading static typing for dynamic typing means losing a set of already-written unit tests in exchange for a degree of flexibility in certain parts of the design. Using inheritance instead of parametric polymorphism offers some benefits but also adds some restrictions, and vice versa. Choosing an agile approach gains you greater feedback and closer connection to your customer (which typically means you’re closer to budget and critical features being completed on time), but requires more work and expertise to pull it off over other, more waterfall-ish, processes. And so on, and so on, and so on.

Tim says this well:

Don’t ever just blindly follow. You always think about it, consider the options, how it might be different in your case, and then, if it still sounds good, try it. Carefully.

If I ever give you any advice, I want you to please never take it without thinking first, analyzing, and deciding for yourself whether or not, and how, and to what extent what I say fits your situation.

(Read more: http://timberry.bplans.com/2010/07/the-sad-truth-about-best-practices.html#ixzz0wBQU3yCB)

But it’s not just the questioners’ fault: speakers, in their zeal to prove that they were smarter than everybody else, bought into the idea, too, because if I’m the one holding the “best practices”, then clearly I’m the one you want to come to with the development or consulting work. After all, who better to hire than the guy/gal with all the answers? In essence, we fed their addiction by tossing off “best practices” in pithy one-line answers (like “every class a service”) that turned out to be utter B/S pronounced by people who had, in some cases, never actually put that practice into practice for anything other than a demo or an example.

It’s time to say “no” to “best practices”.

Speakers: The next time somebody asks you for the “best practices” on a technology, respond with “The best practice you could possibly employ is to hire me.” That, or else with “There are no such thing. You cannot answer a question about a problem outside of its context.”

Attendees: The next time a speaker starts talking about “best practices”, walk out, because clearly the speaker is trying to feed you easy answers when in fact there are only hard choices.

It’s time for our industry to break the habit of taking hits off the crack best practices pipe, and start facing the fact that software development is hard.

Tags: philosophy   psychology   industry  

Last modified 10 August 2010