Could finding the best matchmaking algorithm be an important utilitarian cause?

It would certainly create a lot of utility.

I have no experience with dating sites (so all the following information is second-hand), but a few people told me there was still an opportunity on the market to make a good one. On the existing dating sites it was impossible to make a search query they wanted. The sites collected only a few data that the site makers considered important, and only allowed you to make a search query about them. So you could e.g. search for a "non-smoker with a university degree", but not for a "science-fiction fan with a degree in natural science". I don't remember the exact criteria they wanted (only that some of them also seemed very important to me; something like whether the person is single), but the idea was that you enter the criteria the system allows you, you get thousands of results, and you can't refine them automatically, so you must click on them individually to read the user profile, you usually don't find your answer anyway, so you have to contact each person to ask them personally.

So a reasonable system would have some smart way to enter data about people. Preferably any data; there should be a way to enter a (searchable) plain text description, or custom key-value pair if everything else fails. (Of course the site admins should have some statistics about frequestly used custom data in descriptions and searches, so they could add them to the system.) Some geographical data, so that you could search for people "at most X miles from ...".

Unfortunately, there are strong perverse incentives for dating sites. Create a happy couple -- lose two customers! The most effective money-making strategy for a dating site would be to feed unrealistic expectations (so that all couples fail, but customers return to the site believing their next choice would be better) and lure people to infidelity. Actually, some dating sites promote themselves on Facebook exactly like this.

So it seems to me that a matchmaking algorithm done right could create a lot of utility, but would be very difficult to sell.

EDIT: Another problem: Imagine that there are is a trait which makes many people unattractive. A dating site that allows to search by this criteria will make searching people (who dislike this trait) happy, but people with this trait unhappy. If your goal is to make more money, to which group should you listen? Well, that depends on the size of the groups, and on their likelihood to leave if the algorithm goes against their wishes.