Remember a few months back, when Amy Webb’s TED Talk about how she hacked online dating went viral? Well I sure do; I still get tons of clients from my analysis of her strategy. And now we’ve got this Wired article about a mathematician dude who “hacked” OkCupid over the course of several years. You can’t blame me for analyzing the heck out of his story, too! This is just what I do, people.
(Psst—wanna skip to the takeaways? I won’t tell. This post is looong.)
Now, there’s a lot of interesting content in Chris McKinley’s approach. However, it’s FAR from scaleable or approachable for the average online dater—let’s make that part very clear. Check out this quote from right before the “high” point in his experiment (emphasis mine in this and all other quotes):
McKinlay’s dissertation was relegated to a side project as he dove into the data. He was already sleeping in his cubicle most nights. Now he gave up his apartment entirely and moved into the dingy beige cell, laying a thin mattress across his desk when it was time to sleep.
Sounds like an awesome quality of life, right? Totally balanced, rational, measured, and emotionally and physically healthy approach to finding love. 😉 (Obviously this kind of intense task submersion works for some—heck, I can get a tad like this when I’m writing a really intense newsletter or blog post. But even I don’t sleep in a cell when I’m in a groove.)
When he finally lumped women into enough categories to be able to figure out which one appealed to him the most, he was “ecstatic.”
This was the golden cluster. The haystack in which he’d find his needle. Somewhere within, he’d find true love.
My initial reaction to reading that sentence is to think that reducing your target dating population to these data chunks is a potentially misguided use of your time and energy. But again, different strokes; math professors like McKinley are going to enjoy different strategies than I would or than most singles frankly would. And he actually decided to broaden his approach a bit by targeting a second dataset.
He text-mined the two clusters to learn what interested them; teaching turned out to be a popular topic, so he wrote a bio that emphasized his work as a math professor.
Now here’s where I really need to put on my online dating coach hat. Let’s suspend reality for a sec and take a little hypothetical jaunt. If McKinley had booked me for an online dating session, I could have told him that emphasizing his teaching work would appeal to his target dating demographic. I would have spent an hour talking to him and asking him strategic questions to open him up and learn more about his personality, history, and interests; I think we would have concluded much the same about what to surface to his target dating population. And there are probably a few other things about him that would attract the right mind, but wouldn’t be gleanable from text mining. (After all, a huge part of how appealing your profile is comes down to how well you’ve written about yourself in a natural voice, emphasizing the kinds of quirks and human details that most people don’t think to include.)
I don’t mean to suggest my approach is superior—if nothing else, his is CLEARLY more comprehensive and technical—but softie stuff like “true love” often doesn’t cater to being molded by algorithms. (We’ll circle back to that point later.) He got to this conclusion via his method, but I want to stress that there are more intuitive and/or emotionally intelligent methods to determining how you can present yourself in a way that will attract and excite the people you most want to connect with romantically, and those methods tend to be WAY more approachable for singles who aren’t of the type to get featured as “math geniuses” in a Wired article, ya know? OK, < / self promotion hypothetical >—back to the article.
Next up we have the part that most frustrates me, as someone who approaches these things as a non-mathematician as well as an online dating coach. Consider this quote (emphasis mine):
The important part, though, would be the survey. He picked out the 500 questions that were most popular with both clusters. He’d already decided he would fill out his answers honestly—he didn’t want to build his future relationship on a foundation of computer-generated lies. But he’d let his computer figure out how much importance to assign each question, using a machine-learning algorithm called adaptive boosting to derive the best weightings.
If we’re parsing things with hardcore cold logic, well, I fail to see how leaving the “importance ranking” part of these answers up to your algorithm is consistent with your purported goal about answering things honestly, haha. I think OkCupid would argue that it’s super important that you be human and honest about your importance designations as much as your answers—heck, if you fill out too many Match questions without designating importance, the UI yells at you to cooperate better. (That’s the sign of a great user experience, BTW—when the software berates you for using it like a fickle human instead of a predictable machine.)
This is around where McKinley’s approach, data-driven though it is, starts to win me over a bit. It’s commendable by my standards because it upends online dating norms. McKinley viewed his target profiles and in doing so actually generated incoming messages from women, based on a mere view. The article states that he was now getting 20 messages a day. Let that sink in for a minute. 🙂
Most straight guys who aren’t yet middle aged are well-versed with the imbalanced message behavior in online dating—they generally find that they have to be the pursuers, and that they have a fairly low response rate for most messages. (Straight guys who put time and effort into their messaging strategy in the same way McKinley has put it into other aspects of his search for love are likely to receive higher response rates, but that’s another topic.) So I have to hand it to McKinley in the sense that his targeting, which I found so off-putting at first glance, is demonstrably effective at getting messages/attention.
He started noticing surprising trends about the women he was actually going on dates with. This chart cracks me up—I’ve long maintained that for many singles on OkCupid, avoiding Match questions entirely can be as effective as diving down that rabbit hole. But even I have to delight in the quirk and detail of his Match question data analysis—this Wired chart sums it up nicely.
Despite the Samantha Axiom and other math-based conclusions, the guy just wasn’t finding love yet.
As summer drew to a close, he’d been on more than 55 dates, each one dutifully logged in a lab notebook. Only three had led to second dates; only one had led to a third.
Most unsuccessful daters confront self-esteem issues. For McKinlay it was worse. He had to question his calculations.
Some 33 dates later, he met the gal who became his significant other. Upon telling her about this whole hacking approach, she found it “dark and cynical,” haha. I concur, and I think it’s worth noting this—she was into it. Math nerds, she was into it. So if you get in this spiral where you feel like no women are into you because you’re a math nerd, it’s entirely possible that you’re targeting the wrong kinds of women. (And I promise that you don’t need to sleep in your cubicle algorithm-tweaking to figure out how to target women who are a better fit; just call me.)
McKinley himself is the right kind of mathy-modest about his whole approach (emphasis mine):
“I think that what I did is just a slightly more algorithmic, large-scale, and machine-learning-based version of what everyone does on the site,” McKinlay says. Everyone tries to create an optimal profile—he just had the data to engineer one.
As his girlfriend Tien Wang puts it (with more of my emphasis),
all the math and coding is merely prologue to their story together. The real hacking in a relationship comes after you meet. “People are much more complicated than their profiles,” she says. “So the way we met was kind of superficial, but everything that happened after is not superficial at all. It’s been cultivated through a lot of work.”
“It’s not like, we matched and therefore we have a great relationship,” McKinlay agrees. “It was just a mechanism to put us in the same room. I was able to use OkCupid to find someone.”
(She then points out that, in fact, she found him. He concurs.)
What can you take away from McKinley’s experiment?
-Much like with Amy Webb’s approach, the essence of this story is that you can use dating sites more effectively and innovatively than they might lead you to believe; the core similarity is that these two put in a heck of a lot of time and effort in their different paths to finding love.
-You, too, can put in time and effort (by reading up on advice, hiring a coach like me, or some kind of self-directed path like Webb or McKinley). You don’t have to be a hardcore data scientist to think outside the box, try harder than your competition, and do so in ways that set you apart from other more predictable singles.
-Despite any innovation and “hacking,” there’s a certain amount of just boots to the ground, going on dates, paying attention to how that process is going, and tweaking your behavior so that you become a better date and a better dater over time. If you worry that you’re not self aware or emotionally tuned in enough to make those adjustments, you can hire someone or bug friends to help you calibrate.
-Love is more fickle than just a compatibility ranking provided by a computer. And that ranking can be pretty easily gamed by the fickle people who feed it data. So can other algorithms that surface your profile to potential matches (to say nothing of the features OkCupid offers that let you PAY to have your profile surfaced more/better than other users).
-Online dating success (i.e. meeting someone awesome with whom you are a good fit) is possible for anyone, even geeks who are socially “different” than what most of non-geeky society considers a great catch.
-Maybe don’t drink so much if you black out after your online dates, eh? I didn’t quote that part of the article, but it’s a sound takeaway! 🙂
-I will occasionally compose excessively long blog posts on a beautiful sunny busy Saturday if the topic is fascinating enough. [icon-heart]
If you find this helpful and you’re hungry for more online dating advice, you should sign up for my newsletter, peruse my free advice archives, or perhaps just hire me.
@askvirginia I completely got hung up on the fact that after all of the programming it took him 88 dates to get the relationship. Ugh.
@askvirginia “softie stuff like “true love” often doesn’t cater to being molded by algorithms.”
Do you have data to support that?
Totally get that. For what it’s worth, have you ever tried using an RSS reader, and/or a “read it later” service? I like Instapaper best, but there are lots of choices!
Oh, and my RSS reader of choice is Feedly. For me having those tools helps me come back to longform stuff when I have more time.
This is your occasional reminder that my wife @askvirginia is very good at what she does. Maybe you should hire her. http://t.co/QlKay6NbxO
RT @gkr: This is your occasional reminder that my wife @askvirginia is very good at what she does. Maybe you should hire her. http://t.co/Q…
@phillylauren @askvirginia it’s an optimization problem. He optimized for the first blocking condition, attracting a first date.
@phillylauren @askvirginia he could have better optimized for finding a relationship, but it might have taken as much time as the dates 😉
Nice. It’s gorgeous. 🙂 But I liked it better on iPad, and I gave my iPad to my mom!
Turns out data does not equal quality dates despite the similarity in letters.
Takeaways, headlines, 140 characters characters. C’est moi.
@gkr @askvirginia If I hadn’t already found my sweetie…do you do any other types of coaching? 🙂