I'm guessing Goodread's data set on book collections is maybe 1/1000th as large as Amazon's. They're more likely buying them for the community of heavy readers and the team.
The collections on Goodreads are generally more comprehensive than those on Amazon. I've added over 750 books on Goodreads, but fewer than 10% of those were purchased through Amazon.
You're talking about the actual user-generated lists. Amazon also has the data on what all of their users have actually bought over time. (Hundreds of millions vs Goodreads millions.) That's what they use to create purchase recommendations, because it's higher value data than manual user-generated collections.
Yes — Amazon certainly has significantly more data, but Goodreads' data is deeper, because it's not restricted to items bought from Amazon. Goodreads has a much better picture of what I read than Amazon could ever get simply from my purchase history (even if they cross-match with bookdepository). Being able to mine the Goodreads data too should enable them to make even higher value recommendations. (Note: I'm not claiming this is why Amazon bought them!)
