you should read Start Small Stay Small by Rob Walling, it has great advice on this stuff:

http://startupbook.net/

Also Rob has covered outsourcing in a couple recent episodes of "Startups for the Rest of Us" -

http://www.startupsfortherestofus.com/episodes/episode-64-hi...

http://www.startupsfortherestofus.com/episodes/episode-68-ho...

1) Make sure you get people with English of 4.5+

2) The people who are actively responding to your posts are likely to be among the least talented people in the system, seek out people with high ratings and invite them to your project.

3) You have to recognize and play the numbers hiring on odesk, say 40% of the people are able to do what they say they can do. Hire 2-3 people, let them know they're in a runoff, after 2 weeks one will clearly be the most talented (or said another way will actually be able to do what they claim they can do), end the contract with the others.

This 'run-off' idea can get expensive very fast, maybe not for a company, but it would be expensive for an individual person without significant cash flow

It's very possible but initially a bit time consuming to get a relationship to the point where you can just give work to someone and get results. What I've done to get there:

Post a smallish job with a key element or two you will need (as a test). Require that you only want to hear from folks who have delivered very similar work and can provide links to it, and ask a pertinent question they should be able to answer in their response.

Ignore the folks that are not responsive to your request as dozens of individuals and outsource groups will respond yes to everything and reply only with a generic reply of work done.

Award the simple project to multiple folks who look like they've tackled things similar to what you want, done in a style you like. Set very specific milestones and structured release of payments for work against that and pay promptly. After completion, pick from them who to work with going forward on your project

BTW, I disagree with 'run-off' method. I think results of test are biased if they feel know are trying to win bigger job. Much better I think to assess completed small project, then offer a larger job only to those that you feel you would want to continue to work with on a larger or more extended basis.

I tried this a little bit back before I made much programming experience (but I had some), basically I realized it can lead to wasting a lot of money, it can work for small things but its really hard to manage unless you can get someone in a similar time zone, otherwise just too much time goes by between communications and its hard to stay on the same page, unless you stay up all night to be online with them.

You have to be careful too as some will start working on a project for a day or two and seem to know what they're doing but then all of a sudden productivity will drop to nearly zero, for whatever reason (they found a better paying gig, busy with other stuff, etc)

EDIT: Now that I think about it more, you can find some people who are definitely good but they are not going to be dirt cheap, the few good people I did work with I usually ended up hiring just for small tasks as it was something I needed done but I couldn't have afforded to employ them on a regular basis, usually it seems the minimum rate is $25-30 per hour, which is cheap for a company but it can be hard to swing that if you're just paying out of your own finances.

Also, these people tend to like to work on specific tasks to where they can estimate the hours and they would usually have a few clients at once, it would be hard to get them to give their full focus to you for an ongoing project, I think, unless you could demonstrate stability to them where they felt comfortable turning down other gigs to keep working with you.

They also don't usually want to invest any time in learning your project structure, setup, etc so if you want things done a certain way it can be hard to get them to do it unless you are going to pay them for time spent reading documentation, etc. You can find newbie dev's who will attempt to follow your structure, etc, but they tend to not be of the best quality.

My experience posted before

http://news.ycombinator.com/item?id=3314170

I agree with hiring multiple people, some will be bad, you want to find good ones. Also VERY CLEARLY defining what you want done is important. They will do exactly what you ask for. Start with small projects and incrementally increase complexity. Don't be afraid to ask for updates.

I've done this. I found it hard and had to cleanup a lot of code. Maybe my standards are too high, but I think outsourcing is best for things that you're not good at personally.

I worked with some really good .NET MVC developers on my last personal project. They worked USA Pacific time zone (9 to 5) and were available via skype and google chat. Let me know if you need their profile and I can shoot it to you in email.

I'd be interested in this. Would you mind pinging me on Twitter?

sure will do.