This is a great idea -- it's really exciting to see publishers trying to get actual data about the market, rather than simply condemning all copying as piracy and hoping that the Internet just goes away.
However, I think that Harper Collins got this one wrong. They've put the text of American Gods up in a wrapper that loads pictures of the pages from the printed book, one page at a time, with no facility for offline reading. The whole thing runs incredibly slowly and is unbelievably painful to use. I think we can be pretty sure that no one will read this version instead of buying the printed book -- but that's only because practically no one is going to read this version, period.
The fact is that the full text of American Gods has been online for years, and can be located with a single Google query. I managed to download the entire text of the book in less time than it took me to get the Harper Collins edition to load the first page of Chapter One (literally!). The "security" that Harper Collins has bought with its clunky, kudgey experiment is nonexistent: pirates will just go get the pirate edition.
Unfortunately, the "security" has also undermined the experiment's value as a tool for getting better intelligence about the market. This isn't going to cost Neil any sales, but it's also not going to buy him any. We take our books home and read them in a thousand ways, in whatever posture, room, and conditions we care to. No one chains our books to our desks and shows us a single page at a time. This experiment simulates a situation that's completely divorced from the reality of reading for pleasure. As an experiment, this will prove nothing about ebooks either way.