Scientifically based dating service

Dating agencies are generally based on the idea that users may increase their chances of a successful relationship by arranging meetings with a series of compatible people. Let us suppose that this is true, because it seems reasonable.

The problem for the agencies then becomes to determine what exactly makes people compatible. I believe that most of them are questionaire based, asking people what they think they want, and what they are.
I assume that this works to some extent, but it totally overlooks the biology of the situation. I've covered part of this before (see 'Marianne's moor dating service' link), but I now propose to incorporate part of that concept in a much wider scheme.

First of all, of course we should do the compatibility questionaire thing. We should also try to hone this scientifically for what actually works based on success rate. If people are generally shown to disregard their stated preference in any category, it can be junked or weighed down, as otherwise it is just wasting time and discounting potentially good matches.

Secondly, we test the major histocompatibility complex loci of each applicant. There's no point arranging meetings between people who share alleles - they're not biologically compatible, however shared their interests may be. Female applicants would of course have to forgo birth control pills while playing for this to work.

Thirdly, it turns out that I may have been on to something before. Apparently, when in an intense situation people are not able to determine the source of their arousal. An experiment put (presumably single) men in a vertigous situation, or not, then a young female researcher gave a contact number on some pretext. Subsequently, there were more calls to that number from the men on the cliff edge, as they could not determine the basis for their arousal. (I'd link to where I read this if only I could remember, sorry)
Now, the dating service would need to be careful here, as there's a risk of leading people into relationships which fail after a short period. I therefore propose that this be used only to cement already developing relationships; a couple's second date should be something adventurous.

If there's anything else which is scientifically shown to increase the success of the ensuing long-term relationships, that should be included. For example, men may benefit from being reminded 2 weeks ahead of their fifth wedding anniversary, (perhaps with a list of appropriate gifts) and so on.

(Incidentally, before anyone asks, this should work for homosexual customers. The forms may be different, and the MHC thing may or may not work, but analysis of long-term couples should help determine this.)