I think this is more gimmick than anything else, like most of the services personal genomics companies provide.<p>SNPs don't tell you much about a person's perceived "beauty" and probably tell you only little about disease risks, and there are myriads of problems in the current literature of SNP-associations. Many associations (which go into this baby predictor thing) turn out to be unreproducible once you check a more diverse, or a bigger set of humans for associations. If you search Google Scholar for "failure to replicate GWAS" you'll get 1940 results for 2013 alone.<p>The only predictions that work are the phenotypes you get from your population history: eye color, hair color, skin color etc. But to predict that you don't need SNPs, you only need eyes. That's also cheaper.<p>Edit: I'm not saying that GWAS are crap all in all - they're useful in science in identifying candidate regions which may harbor genes related to favorable traits like resistance to a certain disease etc. They're just not very useful overall for private customers.