Yeah, there is always the danger that the data is just too little or too noisy. But the technical issues you mention (normalization, subtracting bias, using temporal data) all came up in the netflix movie recommendation competition as well so you can always look at how people handled it there.

Some useful information can be found in the report of the winners: http://www.netflixprize.com/assets/ProgressPrize2007_KorBell...

There's a lot of fancy stuff, which would be overkill in a real system, but lot of practical info also.

Also, there was an article of the winners "collaborative filtering with temporal dynamics" which might be useful and I think it is freely available.