Online Dating Insider delivers cutting-edge insight and commentary on all aspects of the online dating industry. Topics include industry news, site reviews, emerging trends, analysis of dating site features, discussion about safety safety, finance and other issues important to the online dating market. Don't miss our Startups directory, useful to anyone running dating or social networking sites. Subscribe to the RSS feed (you can subscribe via email as well). Your comments and suggestions for stories are welcomed.
We offer consulting services to dating sites and social networks as well.
I wasn’t going to mention this until I read the first paragraph of the press release. 200 million questions is a lot, but more from the perspective of “who ARE these people?” than gawking at the sheer amount of data OkCupid is able to gather from members, which is for another post, because if you don’t learn something about your audience after they answer 200 mullion questions…
What got my attention was that OkCupid.com applies a mathematical algorithm to every question answered to determine a compatibility score for potential couples.
The last week or so there has been an increase in the volume of discussion surrounding scientific matching. I continue to read whitepapers, both academic and marketing, trying to make sense of whole new lexicon pertaining to the science of matchmaking. I can barely keep up with Fernando talking about Big-5/Big-7 based personality traits tests and multiple_regression relationship_satisfaction_equations, whew!
I hope Sam at OkCupid will chime in and tell us more about this equation. Does it match animal lovers, or tennis players, or mine a deeper vein of compatibility?
In the past, any time I see the word “scientific” mentioned in the same sentence with “online dating” I tended to attribute the turn of phrase to an overzealous marketing department bent on betting the farm on an indefensible market differentiator. I’m slowly coming around, but it takes a lot of time to understand how and more importantly, why these profiling sytems can, and in some cases, do work. I know that eHarmony has been very successful touting it’s scientific matching system, but I attribute most of that success to $120 million yearly marketing budgets driving pervasive radio and television advertising campaigns.
OkCupid’s press release goes on to say:
Unlike eHarmony, which requires its users to respond to demeaning statements like “I feel unable to deal with things� or “I am easily discouraged,� OkCupid.com gives its members control by letting them suggest the questions to ask, and answer only the questions that matter to them.
Let’s talk about this for a minute.
Consumating (what up, Ben?), and to some extent Tickle, have been letting members ask and answer questions for a long time now. Obviously OkCupid sees this as an important part of their feature mix, and I agree 100%. I was talking about adding questions and blogs to dating sites three (four?) years ago at SITRAS. Anything to move past static billboard-style profiles. A few sites have added blogs, mainly Fastcupid (nee SpringStreet Networks, whatever happened to that site? It’s like it dropped off the face of the earth shortly after the integration with Friendfinder (who bought and rebranded SpringStreet as Fascupid, not smart IMHO. )
It’s interesting that Yahoo Personals and Match are #1 and #2, yet one makes heavy use of personality testing and the other doesn’t. Match featured personality testing from WeAttract years ago. Speaking of Weattract, where are they now? They were supposed to launch a low-priced, easy-to-implement a la Userplane personality test for smaller sites that can’t afford six-figure licensing deals, but that never came to fruition. Anyway, Match used the WeAttract system for a while, dropped it, and Yahoo picked it up and ran with the ball, integrating the answers with the search function, something Match never got around to doing for a variety of reasons. I’ve written about this before, sometimes its good to remind people of these things, especially you new blog readers.
Back to OkCupid. Who says that self-identification of less than positive personality traits is demeaning? Will people answering only the questions they feel like affect how the answers are used to match them with this supposedly scientific matching system? Who knows? Heck, maybe OkCupid’s $6 million is worth enough advertising dollars to catapult them past PlentyOfFish? The dating space is just to squirrelly to make guesses about what’s coming next, except that revenue will grow and site will get bought. I thought we would be far ahead of where of where we are today when I started blogging about online dating in 2002.
OkCupid now has a TON of marketing data on their members. Can Sam and his team convert what they learn from the data into features and functionality heads-above the rest of the industry, or will they settle for a killer advertising campaign, hopefully without going downmarket into True.com boobs-buns?
{ 1 trackback }
{ 5 comments… read them below or add one }
1) The selection of the questions
2) The algorithm used to turn responses into compatibility scores
As to the selection of questions, we called out a couple of eHarmony’s questions (”I feel unable to deal with thingsâ€? or “I am easily discouragedâ€?) not because they are negative personality traits but because we doubt their ability to predict compatibility.
Let’s look to the tried-and-true world of offline dating for an analogy. Imagine that a friend of yours wanted to set you up on a blind date. You’re not quite sure you trust your buddy enough to agree, so he offers to answer a few questions about her to help you make up your mind. Are you going to ask “Does she feel able to deal with things?” or “Is she easily discouraged?”? I think not. In the real world, people ask questions like: “Does she smoke?” “Is she in to hip-hop music?” “Does she want to settle down and have kids?” It’s not that any of these are positive or negative — it’s these questions, and their answers, have predictive value.
So, whereas a site like eHarmony relies on its psychologists to generate questions they believe are important, we turn to our users — the daters — to tell us what questions they would ask potential dates before accepting a date. And they’ve replied by suggesting over ten thousand questions. After removing redundent, ambiguous, or otherwise poorly designed questions, we have an active database of about 3,000 questions.
Each of these 3,000 questions (to which we have 200,000,000 answers) takes roughly the same form:
- Question Stem
- Your answer to the question
- The answer(s) you’d most like your ideal match to have selected
- Your weighting how important this question is
So, consider the following question:
Do you smoke?
a) Yes, regularly
b) Yes, but only socially
c) No
Now, I might be a smoker trying to quit, in which I would answer (a) but might seek someone who answered (b) or (c). I might be a non-smoker who is completely intolerant of smokers, so I would answer (c) and seek (c). Or I might be an open-minded non-smoker and answer (c) but seek any of the above. So it’s not enough to simply know someone’s answer.
Dave, you asked whether we match tennis players or animal lovers, but that’s a question you’d have to ask a site like eHarmony which has a “special sauce” to determine that two animal lovers make a good pair but that two tennis players won’t. We empower users to express their own preferences.
The third part of the question is the weighting, another element missing from the eHarmony model. It’s exceedingly obvious that some people care a lot about religion while others care more about smoking. I, for one, am allergic to cats, so that’s a bigger deal-breaker than either God or cigarettes. The OkCupid system allows all of those subtleties to come out.
On to the algorithm for determining compatibility. eHarmony relies on psychologists to know whether someone who “feels unable to deal with things” should be matched with someone who “is easily discouraged.” OkCupid takes a much simpler approach. Users tell us tons of information about themselves and tons of information about what they’re looking for.
We then run unbiased, impartial, hands-off mathemative algorithms to determine which pairs of people have the highest cross-compatibility. That is, how well did person A satisfy what person B is looking for and how well did person B satisfy what person A is looking for? See, we’re not promising soulmates, which intuitively seems impossible to do, anyway.
We’re simply saying that we’re going to rank order the database by how compatible the other members are with you. By no means to we suggest that your #1 match on OkCupid is your soulmate. Instead we suggest that pool of people with, say, 70%+ matches is going to be much more fruitful for you than the pool of people below 30%.
If you really want to get a feel for our matching algorithm, answer about 50 questions (the average user answers about 200) and then, instead of sorting by match percentage, sort by ENEMY percentage. Yes, this is a light-hearted tool, but people tend to more easily identify things/people they dislike than those they like. My guess is that you’ll be very happy not to come across those profiles.
Fundamentally, there is a philosophical question that differentiates OkCupid from eHarmony. Why do people use online dating sites? Is it (a) they don’t know what they’re looking for and they want someone to tell them or (b) they have a pretty good idea of what they’re looking for and they want an environment where they can look through a much larger group of people?
eHarmony clearly believes (A). OkCupid believes (B).
Some people like scientific tests, others like quizzes, there is room in the market for both.
I don’t see OkCupid as serious Eharmony competition, especially given the demographics of the site, but all dating sites don’t have to be for serious singles.
Tickle was an early test-taking site, and was snapped up by Monster for more than $30 million. Obviously people like tests and companies see the value of the test-taking community.
I like that the OkCupid questions are generated by users, sites with a dynamic component are much more compelling. The dynamic features will no doubt change over time.
It seems that games from companies like Bunchball are the next interactive feature for dating sites. Agree or disagree?
Thank you for this comment. I must admit that I was quite impressed
by the OKCupid model when I was doing research for my book. I much
prefer your philosophy to the “we’re psychologists and we know best�
approach.
I bet you could find statistical significance in just about any type of matching with a big enough sample size. It’s probably at least as significant as the weather on the first date. If it were me, I’d try my luck with photos on plenty of fish, or something I could have fun with like OkCupid.