Thursday, October 28, 2010

Results of the "What Really Makes a Good Programmer" Survey

Survey Response CountsBack in September I wrote a blog post called What Really Makes a Good Programmer? My goal was to ask various members of the development community what traits they thought contributed to the quality of a programmer. If you haven't taken the survey yet, I'd recommend you do that before reading the results. Wouldn't want to bias your opinions, right?

If you have taken the survey, but haven't told everyone you know to take it too, go ahead and do that; I'll wait.

Now that we've covered that, let's get on to the analysis. As you recall from taking the survey, there were 17 traits and you rated them on a scale of Very Important, Important, A little Important, Negligible, and Completely Unimportant. In order to do any analysis, I had to take these ratings and convert them to numbers.

I decided that A Little Important was the baseline and basically represented a lack of opinion on the topic. These traits are the ones most people feel are nice to have but aren't requirements. Basically, a good programmer often has these traits but not having them doesn't mean you're not a good programmer. I recoded these values with 0 points.

Very Important and Completely Unimportant are the ratings that people used when they felt strongly about the item. That means that, respectively, the trait is either absolutely necessary to being a good programmer or the trait has no bearing on the quality of programmer you are whatsoever (or perhaps a little bit of a negative indicator). These values were recoded as 10 and -10 respectively.

The central values of Important and Negligible are what I call the non-committal values. You feel that they make a difference but they're not quite important enough to bank on them. I gave these a 3 and -3 respectively.

For example, "Has good problem solving skills" was rated Very Important and "Cheap" was rated Completely Unimportant. One could say, "someone with poor problem solving skills would make a poor programmer." On the other hand, you cannot say, "someone who is not cheap is a poor programmer." Of course, the contrapositive could also be stated that, "someone who is a good programmer is also a good problem solver." The contrapositive "someone who is a good programmer is also cheap" is considered by the community to be a largely invalid assertion.

"Fast" and "Co-located" appear nearest to the baseline. This may be due to the fact that many of the respondents didn't know what I meant by co-located. In any case, one might say, "a good programmer is a good programmer whether or not she's in the same building," or that "just because you're fast doesn't mean you're good and just because you're slow doesn't mean you're bad."

"Communicates effectively" and "Interested in helping teammates" are moderately rated in favor of contributing to being a good programmer while the two "college degree" items are rated moderately against contributing to being a good programmer. For example, you might know a lot of programmers who are great developers but who lack the social skills or interest to become good communicators or team players. As a result, you might say that good programmers often communicate effectively and help teammates, but some good programmers cannot.

You might also say, "Many people without college degrees, let alone a computer science degree, are great programmers. Thus, while having a degree is helpful, you can learn to be a great programmer without it."

Now, I feel it is worth noting that "unimportant" does not mean "negative." Just because most of the development community feels that having a CS degree or other certifications is unimportant to being a good programmer, it doesn't mean that the degree itself is unimportant. It just means that having the degree won't make you a good programmer; you still have a lot of learning to do.

With notes on the analysis out of the way, what do you say we take a look at some data? Currently, the new Google charts won't let me force the display of the categories, so you'll have to hover over the dots you're interested in. Here's a legacy chart you can look at if the javascript version below is unsatisfactory. The legacy chart has current data; it's just not as fancy as this one:
I'm sure you noticed there are two series in this graph. All Responses represents the average responses for all respondents; however, as you noticed in the respondent histogram that I have a dramatically unbalanced sample with programmers outnumbering all other groups combined by almost 3 to 1. To get a more accurate representation of the "general feel" of the community, I included the Group Average measure. This is the average of the averages. It's sort of the electoral college of informal research.

This chart demonstrates the average responses for each group:

No comments:

Post a Comment