It doesn't take many bits to encode a customer. As it stands today, 33 bits is enough to represent every person on this planet. Bump it up to 64 bits and you could represent not just the purchaser, but the purchase itself, for quite some time to come. Now 8 bytes wouldn't be hard to hide in an ebook, several times over and encoded in obscure ways that would be difficult to detect. An extra space which is never even rendered by the ereader. Tricks involving UTF which are undetectable when the document is rendered. A quirk in the markup or stylesheet which, again, isn't visibly detectable. Yet the most profitable place to hide such data would be in the images (most ebooks have at least a cover, many have graphical embellishments, some use images fundamental to the text). Done properly, the watermark can be implemented in such a way that it is impossible to remove without visibly altering the book itself. That's particularly true when it is embedded in the images.

So yeah, have fun.