Update: you can find a Spanish translation of this blog post here.
Last night, I tweeted the following:
I'm confused by the claim that "10x" or "rockstar developers" are a myth. Are star athletes, artists, writers, and, uh, rock stars, a myth?
— Yevgeniy Brikman (@brikis98) September 29, 2013
I got tons of replies and questions, but Twitter is an awful medium for discussion, so I’m writing this blog post as a follow-up.
There have been a bunch of articles that claim that 10x developer doesn’t exist. The arguments against it generally fall into 3 buckets:
- The original 10x number came from a single study (Sackman, Erikson, and Grant (1968)) that was flawed.
- Productivity is a fuzzy thing that’s very hard to measure, so we can’t make any claims like 10x.
- There is a distribution of talent, but there is no way a single engineer could do the work of 10.
I disagree with all of these. Let’s go through the arguments one by one.
It’s not one study
Although armchair scientists on Twitter and Hacker News love to shoot down peer-reviewed studies, the evidence in this case is fairly compelling and not limited to a single study. Allow me to quote the top reply from this related discussion on StackOverflow:
...The original study that found huge variations in individual programming productivity was conducted in the late 1960s by Sackman, Erikson, and Grant (1968). They studied professional programmers with an average of 7 years' experience and found that the ratio of initial coding time between the best and worst programmers was about 20 to 1; the ratio of debugging times over 25 to 1; of program size 5 to 1; and of program execution speed about 10 to 1. They found no relationship between a programmer's amount of experience and code quality or productivity.
Detailed examination of Sackman, Erickson, and Grant's findings shows some flaws in their methodology... However, even after accounting for the flaws, their data still shows more than a 10-fold difference between the best programmers and the worst.
In years since the original study, the general finding that "There are order-of-magnitude differences among programmers" has been confirmed by many other studies of professional programmers (Curtis 1981, Mills 1983, DeMarco and Lister 1985, Curtis et al. 1986, Card 1987, Boehm and Papaccio 1988, Valett and McGarry 1989, Boehm et al 2000)...
If you can’t measure it, you can still reason about it
Even if you ignore the studies above and declare that “programming productivity” is hard to measure—which it is—we can still have a discussion about 10x programmers. Just because something is hard to measure doesn’t mean we can’t reason about it.
For example, how did you pick the programming language for your most recent project? Did you look up a study that “proved” the language was more effective than other alternatives? Personally, I don’t need an experiment to prove that Ruby will be an order of magnitude more productive choice for building a website than, say, C. You could throw together some rough metrics (library availability, community support, documentation), but the reality is that most people make this sort of language decision based on intuitive reasoning and not a double blind study. And despite the lack of hard data, I’d bet that picking Ruby over C for website development turns out to be the right decision most of the time.
Of course, this isn’t unique to programming: what “metric” could tell you one writer, artist, teacher, or philosopher is better than another? Merely from observing them, I can’t give you a “productivity metric” that suggests Shakespeare, Nabokov, or Orwell were an order of magnitude better than the average writer, but the vast majority of people would agree that they are.
Programming is not manual labor
The biggest problem with the pushback against a 10x programmer is that some people think of programming as manual labor and programmers as assembly line workers. Some programmers are a bit better than others, but surely, a single programmer could not consistently close 10 times as many tickets as another! And a team of 10 will always outperform a single coder! Nine women can’t produce a baby in 1 month!
The logic above makes it sound like programming productivity is all about typing speed; as if the 10x programmer is simply the one that produces 10 times as much code as the average guy. This line of reasoning ignores that programming is a creative profession and not manual labor: there are many, many ways of solving the same problem. Instead of the baby analogy, think more of a crime solving analogy: 10 average detectives versus one Sherlock Holmes. Who will solve the crime faster?
A 10x developer will have insights and find solutions that would never occur to an average programmer; they will avoid entire categories of problems that eat up enormous amounts of time amongst average programmers. 10 engineers writing the wrong code could definitely be out performed by a single engineer writing the right code.
Programming is about choices
Consider how many decisions go into building a single software product, such as a website: what language do you use? What web framework(s)? What do you use for data storage? What do you use for caching? Where do you host the site? How do you monitor it? How do you push new changes? How do you store the code? What kind of automated testing do you setup?
10 average programmers will make “average” quality decisions at each step and the costs or benefits of these decisions will multiply. Imagine traffic increases exponentially, and this average team setup an average website, with a data storage engine that’s hard to shard, hosting that doesn’t have enough redundancy, version control without proper backup, no CI environment, and no monitoring. How productive will those 10 coders be if they are spending all their time putting out fires?
A single programmer could outperform this team of 10 if the programmer can model the problem in a way where there is an order of magnitude less work to do. From years of experience, a great programmer will know that errors are much more costly to fix later. By making good decisions up front, a 10x programmer may avoid months of work down the line.
It’s not about writing more code; it’s about writing the right code. You become a 10x programmer not by doing an order of magnitude more work, but by making better decisions an order of magnitude more often.
This isn’t to say 10x programmers make no mistakes at all; but programmers make a huge number of choices every single day and great programmers make the right choices far more often than average programmers.
And this isn’t unique to programming. Would you rather have 10 average scientists or 1 Isaac Newton? 10 average scientists did not come up the laws of motion, theory of gravity, binomial series, calculus, etc; a single Isaac Newton did. Would you rather have Michael Jordan on your team or 10 average players (note: Jordan got paid ~10x the average NBA salary)? Would you rather let Steve Jobs or Elon Musk run a company or hand over the keys to 10 average entrepreneurs?
10x programmers are rare
It’s important to put things into perspective. Star programmers, athletes, writers, and scientists are exceedingly rare. I wouldn’t recommend building a hiring strategy around solely hiring “rock stars”; you’ll end up looking foolish and lonely. Don’t let perfect be the enemy of good: hire the best engineers you can get and give them ample opportunity to develop and get even better.
However, don’t fall into the fallacy that all programmers are created equal. There is a vast spectrum of ability in any creative profession. On one end are the type of hires that can sink an organization, actively increasing tech debt with every line of code they write. On the other, there are people who can write code that changes what is possible and have an impact that is an order of magnitude greater than the average.
Yevgeniy Brikman
If you enjoyed this post, you may also like my books. If you need help with DevOps, reach out to me at Gruntwork.