I know I’m a research geek with lots of academic background and even a stint as a research scientist at an R&D lab, but still, this kind of data really fascinates me to no end: Online dating site OK Cupid has released the result of some semantic analysis that it did of the aggregate emails that people on the site sent each other, comparing what they wrote against whether or not they actually went on a date with the other person.
Now anyone with a research background will immediately glom onto the fact that they have committed the cardinal sin of assuming causality from correlational data, but it’s so darn interesting that I’ll let it slide anyway 🙂 In a nutshell, the problem can be explained thusly: poor people have above-ground pools, while rich people have in-ground pools, therefore getting an in-ground pool must make you rich.” See the problem?
Anyway, we’re not interested in pools or money, we’re interested in picking up that sexy possibility on an online dating service like Match or eHarmony or… etc. That’s what they looked at, and if we can even just look at the correlational data and if we can make the assumption that the people that use OK Cupid are sufficiently similar to those that use the more mainstream sites, well, let’s look at what they found…
The first data point is one that I found the most interesting: on average, only 32% of first contact email messages sent on the dating site had a response of any sort. That means that approximately only one of three initial email messages garners any sort of answer. One reason why an email a week just doesn’t cut it on dating sites if you’re serious.
First things first: if you’re going to email someone on a dating site, don’t use the same abbreviations you’d use if you were sending a text message to them.
The graphical version of this — comparing occurrences of a specific word against the likelihood that a response would be received — looks thusly:
(read this by seeing what word they analyzed and, on the left side, the percentage of messages containing that particular word that generated a response. In other words, if you use “ur” instead of “your”, you should see about 7% response, or not even one response for every ten messages u snd.
Since most people using these dating sites aren’t 15 (just a wild guess on my part, but still, do you disagree?) it’s obviously smart to take the time to use good grammar and spelling in your message. Skip the “ur” and “r” and “luv”. Oh, and skip those in your text messages too, for that matter, but that’s another story.
Maybe instead of lame abbreviations you should use a compliment or two? Well, maybe not:
The conclusion here is that you’re not likely to get a response if you talk about the other person being “sexy” or “hot”, which makes sense given that it’s the very first message you’re sending the other person. How could you know that they’re hot and sexy? And if you’re talking about yourself, well, yow, that’s a scary concept and I’d autodelete those messages too!
One more interesting graph: should you talk about your religion, and if you do, does it increase or decrease the chance of getting a response?
I find this particular result a bit inexplicable, though perhaps if we had a sense of the frequency of these words occurring and whether they’re included in profiles where the sender or receiver has already identified themselves as being a part of a specific religious belief, it’d be more understandable.
There’s more data presented in the original research, but I’ll just say that to some extent everything that they found makes complete sense when you think about the fact that OK Cupid tells customers “Tell us about yourself and we’ll help you find your perfect match. With over a million quality singles from all around the world…”
So if you had a community of “quality singles” where the goal was to find your perfect match, you can easily figure that messages like “hey hottie, wanna get together and split a bottle of Bacardi?” are less likely to be successful than “I see from your profile that you work at a brokerage account? I did too…”
Nonetheless, it is interesting to try and correlate occurrences of unusual words against likelihood that a message containing those words will produce a response from the other person. Go through the research, think about it, and then answer the question: would messages containing these words work with you, or not?