Like RGB colors Y'CbCr uses 3 channels. However, there are 2 color channels and one luminance channel.
The trick is, that you have to save the luminance (Y') in one channel and in a second channel you save Cb and Cr alternately, like that: (one [xx] = one pixel) [Cb][Cr][Cb][Cr] [Cr][Cb][Cr][Cb] [Cb][Cr][Cb][Cr] [Cr][Cb][Cr][Cb]
If you read them again either Cb or Cr is missing (of course) but you can just take the missing information from a neighbour pixel because they are likely to have the same or a similar color.
So actually you loose 50% of the color information (hue & saturation) but you still have 100% of the luminance (the more important part). And although there are one third of the information missing, you don't really see any difference.
edit: I hope you understand what I'm trying to say
