It's not a talent shortage, it's a hiring problem

People who would work out perfectly in the R&D division of SAS Institute are not necessarily great first hires at consumer web start-ups (forgetting the potential skill impedance mismatches). There are lots of implicit ideas about "how things get done" that are part of any group. These ideas are the culture.

e.g. Does the team practice TDD? Do they focus on unit tests or integration tests? How is work assigned? Does the team prototype ideas or discuss them in the abstract? If someone says something you think is wrong in a meeting, do you confront them publicly or privately? 12-hour days? Do you pick tools that are cutting edge or proven? Duct tape together two open source libs for one feature or write your own? And of course, the classic: Is it better to a) ship early with known non-critical bugs, b) ship on time with no known bugs, but inadequate testing, or c) when all/most stakeholders feel confident the project is bug-free?

These questions aren't nearly as binary as I've presented them (and there are many more). If you and your potential new team don't agree on these issues that's a poor cultural fit. Like a poor technical fit, it doesn't necessarily mean you shouldn't be hired, but it does merit significant consideration.

There is a whole field of study, called industrial and organizational psychology, that researches issues like what hiring processes are most likely to obtain employees who do good work for organizations. Studies of many different hiring practices across many different industries and job categories have shown that the two most valid predictors of work performance and success in an occupation are

a) work-sample tests, if one can possibly be devised for a particular job,

and

b) general cognitive ability tests (IQ tests or IQ-like tests such as SAT scores).

Both work-sample tests and general cognitive ability tests have about the same predictive ability (predicting less than half of the variability in actual work performance, but predicting more than any other kind of hiring screen) as each other, and considerably more validity than many more common kinds of hiring procedures.

Expecting job-seekers to have

1) X years of work experience in Y technology,

2) a smooth manner in a face-to-face interview,

3) a degree in field Z from an elite university,

or any of many other hiring criteria are all less likely to identify job-seekers who will be successful on the job than a work-sample test. If the work at your workplace is so complicated and varied that it is difficult to design a representative work-sample test for job-seekers, test the job-seekers' general cognitive ability with one of the validated hiring tests used for that purpose. That can reduce the expense of your hiring process and improve its results.

> give A LOT of people a simple test (which SELECT statement is an inner join...)?

That is like asking, how do you implement a reduce function in CouchDB? The answer is easy, and would only take a couple of minutes with Google to answer, but I bet even many top programmers could not answer that question off the top of their head.

If you are looking for someone who writes SQL queries all day long, maybe that is a good question. But if you are writing software that does complex visualizations, occasionally pulling some information from an SQL database, should a recite-from-memory knowledge of SQL really make or break the candidate? INNER JOIN syntax is something you can easily query Google for.

What works for me is:

a) Making the candidate interested - I want them to be really interested in being hired, to ensure he will give his best shot (this is much easier when hiring with a specific spot in mind, which is not always the case);

b) Using a coding test that the candidate can do remotely at the time most convenient for him (http://www.codility.com/ currently - they are really good for this) as a first filter. This establishes basic programming skills, and allows me to...

c) Call the candidate for an in-person conversation and spend much more time discussing higher level questions, validating problem solving skills, and understanding the candidate past history. Unless it is remote position, in which case this will be a phone screening - notice I say an phone and not some kind of crappy can-barely-hear with-echo-or-lag-or-some-other-crap VOIP solution.

Finally, keeping this in mind has helped a lot: http://news.ycombinator.com/item?id=3690544

I cannot comprehend how someone could have so "excellent hygiene" that I would want to hire them...