I would choose a very small niche. The chicken-and-egg problem disappears almost complete if you choose the niche small enough. This can be a movie category or a region, maybe the movie category is better in your case.
Because, if the niche is small enough, you will always be able to handle the short coming of the other (supply/demand) side in a realistic way, be it by cold calling or adwords or whatever.
I would stick with this niche for a while. You will have some first successful transactions, some success stories, some data you can use for statistics and pr, generate modest earnings, build up some ranking positions on google, etc... and you start understanding the numbers of your business, the conversion rate, etc..
And you will provide traction for your current users. Promoters that have signed up for the newsletter will regular receive some opportunities, and this is what counts to make your useres happy in a marketplace.
When your first niche is doing well, you can risk the next step, expand two other segements (categories). And repeat.
From my point of view, the critical thing is that you will loose a lot of ressources everytime you start acquiring new customers/segments. There will always be a mismatch of the demand-side and the supply-side. To be successfull, you needed to cover one segment by segment, otherwise you may run out of time and money before you get significant traction.
I made pretty good experience with a VA from India (no agency), currently working with the same guy for over 1 year now.
Regarding success: If you picked a reliable person, I think one of the most important things is how well you describe a task (unambiguous, measurable results). And you should have done the task before by yourself, so that you can estimate the effort and the difficulties of the task.
Usually I describe the task to the VA in form of sequential steps like a use case. When I am not sure how much time it needs, I tell him to invest 1-2 hours and then get back to me so that I can see how far he got. By this approach, I avoid to waste time and money.
For the quality assurance, I always tell him to do first 2-3 examples of the task and then get back to me. Then I check it, and if everything is fine, he proceeds with the task. This also saves me from wasting time and money.
I found the VA through mturk. After I completed my tasks there, he mailed me and asked whether I got some more tasks.
In retrospecitve, I think this was a pretty good quality filter since only the really motivated guys write you an email and asking for more. Of course, this is not a scalabe approach but on a small level, it worked for me.
And I recommend building a working relationship and see him as partner, not as cheap assistant. This is something that you do rather indirectly, by how you communicate and express yourself in the emails, etc... From my experience, veing a nice person will have a big impact on the quality of the work. But that's actually valid in any field.