Quantified Swipes

A month of online dating, in data and charts.


Introduction

When I thrusted myself back into the dating game after a long relationship, I quickly found out that the scene had changed — a lot. Hell, it was unrecognizable from even two years ago, when I was last on the market.

Most striking was the plethora of technology at my disposal, which wasn’t available before. This is especially true in San Francisco, a city that’s both perpetually single and obsessed with software. So it seemed natural that people here made big use of online dating — OkCupid, Tinder, Zoosk, Hinge, Grouper, Coffee Meets Bagel, and countless others — without the stigma it carries elsewhere.

Everyone has an opinion on them, too. There are countless blogs, columns and forums, packed with (often contradictory) advice on Love 2.0. It can be overwhelming to take in.

So instead of opinions, I decided to turn to data for advice — my own data. I conducted an experiment.

The Experiment

For one month, I meticulously logged every interaction I made on three online dating sites, which I then asked a friend to analyze.

I’ll walk you through my entire process: what I was testing for, how I set up the experiment, and what the results were. Yes, I also made charts!

Just keep in mind — the results may surprise you.

Hypotheses

This experiment attempted to prove two general dating hypotheses, and three messaging-related ones. The two general hypotheses were:

  1. Tinder is just a hookup app: Tinder’s focus on appearance over personality, and filtering mechanisms (you can only message people who like you back) has lent it a dubious reputation as being a “hookup” app. But how true is this?
  2. It’s easier to meet people on paid sites: people say it’s because there’s a much smaller dating pool, and paying for access or functionality means that these sites’ users take dating more seriously. But again — is this really true?

Finally, since so much in online dating hinges on message content, I also tested three common messaging hypotheses:

  1. Demonstrating you’ve read her profile works: as the theory goes, since women on dating sites receive a barrage of boring, noncommittal or straight up creepy messages, demonstrating some thoughtfulness makes you stand out. Sounds intuitive, but we’ll see what the data says.
  2. Pickup lines don’t work: this one’s a mixed bag, as almost every woman says she doesn’t respond to pickup lines, yet a lot of men claim to have a few cheesy, yet “proven” sizzlers. So which is it?
  3. “Hey” doesn’t work: related to the first hypothesis, the reasoning goes that “hey” just doesn’t stand out amidst the flood of messages women get on online dating sites. Is there some truth to this? Read on, and find out.

Methodology

I’ve taken several dating sites for a spin, but for this experiment, I limited my tracking to just three: Tinder, OkCupid, and HowAboutWe.

  • Tinder: It’s all the rage right now. Over the past year, Tinder has taken the dating world by storm, through its simplicity, its deviously addicting “swipe left/swipe right” gestures… and its previously mentioned reputation as a “hookup” app.
  • OkCupid: Tried and true. OkCupid pioneered the matching algorithm, and is still the most widely-used free dating site. Though it does offer paid subscription plans with certain perks, these aren’t necessary to browse profiles and message people, so I refrained from paying.
  • HowAboutWe: A bit of an outlier, HowAboutWe takes a different tack. You post date ideas to your profile, or use them to ask others out. HowAboutWe is my control network: out of the three, it’s the least well-known, and it’s also the only paid service I used.

The Profile(s)

Maintaining consistency across each service is hard, as each one has designed the profile to serve a different purpose.

Tinder, for example, puts your picture in the spotlight, while OkCupid has you fill out a detailed profile and answer a bunch of questions. Meanwhile, HowAboutWe emphasizes your date ideas much more than your bio.

Still, I achieved some consistency by picking the same pictures, and using a similar voice when filling out taglines and profiles. Samples below:

Tinder:

OkCupid:

HowAboutWe:

Matching / Messaging Protocol

Once my profiles were streamlined, it was now time to start browsing profiles and messaging people. I established the following protocol for each site:

  • Tinder: Swipe right on everyone. Send messages to every match.
  • OkCupid: Message every new profile that shows up under “Browse Matches” and is over a 75% match.
  • HowAboutWe: Message all profiles that show up under “Daily Matches,” as well as those under “Dates For You” which match at least two profile details.

Recording Results

While logging my messages in a spreadsheet, I assigned a number to each message, and did the same for each outcome. This helped me categorize both messages and outcomes for later analysis .

Message Categories

I classified messages in ascending order of “cheesiness.” So a safe, non-cheesy greeting like “hey” is ranked a 1, while an opener clearly meant as a pickup line is a 6.

If all else fails, quote Anchorman...
  1. “Hey” — Self-explanatory.
  2. “How’s it going?” — In the same vein as “hey,” but slightly more involved.
  3. Question about Profile — Questions relating to any written content in a person’s tagline, bio, questions, etc.
  4. Question about Picture — Questions relating to a person’s profile picture(s). Generally considered riskier than profile questions.
  5. Joke / One-liner — Quips, jokes or wisecracks that are primarily meant to be funny instead of thoughtful (ie, “Wanna hear the worst joke in the world?”).
  6. Pickup Line — Quips, one-liners and other lines that are both very cheesy and meant as direct dating propositions (ie, “Are you my Tinderella?”)

Some messages fit into multiple categories — where this was the case, I filed it under the category it most fitted under.

For example, a question about someone’s profile that was clearly meant as a joke, or an attempt to be funny, would be classified as a 5.

Outcome Categories

I also classified outcomes in ascending order. I did not count non-responses as outcomes; instead, I started with message responses at 1, all the way up to a successful date at 3 (as well as its subcategories).

…You never know, it just might work!
  1. Response — A reply to an opening message I sent.
  2. Unsuccessful Date — A date that didn’t end in a hookup, or lead to a repeat.
  3. Successful Date — See below.

Successful Dates are further classified into two subcategories:

  1. Date ending in hookup — Since this is an incredibly subjective term, I defined “hookups” as broadly as I could. In this experiment, a hookup can include everything beyond heavy kissing, up to and including the sexy times.
  2. Repeat date — A first date that led to subsequent dates down the line. This doesn’t include the actual followup dates. Rather, if a first date led to a repeat down the line, I counted it as successful.

Note that these two subcategories can overlap. If a date both ends in a hookup, and leads to a repeat, it’s logged as one successful date. In short, one message = one outcome.

The Analysis

I asked Colleen, an old friend and professional statistician, to run a statistical analysis on my raw data. She described her methods as such:

I ran a series of ordinal regression models on your message content hypotheses. The first tests the usefulness of “Hey,” the second tests reading her profile against the remaining message content options, and the third tests pick-up lines against the remaining message content options.

She also ran a similar analysis on my general hypotheses. So, what do the numbers say?

Results

Overview

To start, here are two charts, illustrating what went down in 30 days of online dating:

Throughout one month, I…

  • Viewed 3,709 Profiles
  • Sent 215 Messages
  • Engaged in 43 Conversations (w/o additional results) [20% of messages]
  • Went on 4 Unsuccessful Dates [1.86% of messages]
  • Hooked up 3 Times [1.4% of messages]
  • Went on Repeat Dates with 3 Girls [1.4% of messages]

Fascinating, but what else does this data tell me? Besides the fact that online dating takes time and effort? Do my hypotheses still stand?

Is Tinder Just a Hookup App?

Does Tinder deserve its reputation as a hookup app? The answer surprised me — no, it doesn’t! I had the same amount of hookups and repeats on Tinder as I did on OkCupid and HowAboutWe:

So as you can see, Tinder is no more of a “hookup” app than the other two sites. Furthermore, the response-to-hookup ratio is higher for both OkCupid and HowAboutWe, suggesting you have a much better chance to hook up on those sites.

Are paid sites better than free sites?

In terms of raw numbers, Tinder (a free site) provided better outcomes than both OkCupid and HowAboutWe. So at first glance, Tinder seems to be the way to go. However, there is a caveat:

On Tinder, I had to swipe A LOT in order to get matches and actually start messaging people — to the tune of thousands of swipes throughout the month. With OkCupid and HowAboutWe, you can message anyone, at anytime, unless they block you. So I got a much smaller message and response pool for Tinder, which could have affected these results.

So what about the paid sites? There doesn’t seem to be evidence that using paid sites like HowAboutWe results in better outcomes — in fact, the analysis still shows that Tinder provides a better predicted outcome of a date than the other two sites. However, the data does suggest a significantly better return on responses on HowAboutWe than on Tinder.

Confused? Think of it this way: if you’re looking for girls who will go out with you once you hear back from them, use HowAboutWe.

On the flip side, you’ll get more responses with Tinder, but you’ll need to work harder from there. However, Tinder gives better outcomes for dates, so the extra legwork seems to be worth it.

What’s the best thing to say in a message?

Ah, the eternal question — what to put in that opening message? Should you say “hey,” use a pickup line, read her profile, or try to be funny? Again, the answer may surprise you…

Message Content vs. Outcome (All)

In a nutshell, just about anything works better than “Hey” or “How’s it going.” I suspected “hey” to be a lousy opener, but interestingly enough, you don’t necessarily have to read someone’s profile for better dating outcomes — every opener other than “hey” or “how’s it going” worked just as well.

Colleen also controlled for each dating site in these regressions:

Message Content vs. Outcome (Tinder)

Message Content vs. Outcome (OkCupid)

Message Content vs. Outcome (HowAboutWe)

The results supports our previous findings: Tinder gives better outcomes. But regardless of what site you use, the message (no pun intended) is loud and clear: do not use “Hey” as an opener.

Conclusion

So, what’s the verdict? To summarize, here’s what the data ultimately suggests:

Your best chance at a good date is using Tinder and not using a generic message.

Of course, there are caveats. The sample size is very small. There are also too many variables at play, as human beings are complex creatures, and the variables I tested are far from the only ones at play here. I’m also dating from a position of privilege, as a young, white, straight male in San Francisco — this won’t be applicable to all of you. Finally, this experiment doesn’t include key qualitative data.

My most successful opener, which led to my most successful date so far.

For example, although my data crowned Tinder the clear victor, the most successful individual date I’ve been on was actually from HowAboutWe. After our first date, I continued to see this girl for a month afterward — but the data doesn’t reflect this (UPDATE: This is no longer the case. See addendum below).

I also didn’t capture the chemistry on each date. I’ve felt sparks with a greater number of my Tinder dates, but much deeper connections on OkCupid — but again, this was beyond my experiment’s scope. It’s too hard to quantify.

Finally, I was only logging data for a month. My experience has varied and evolved since then — and it will continue to change the longer I stay in the dating scene.

Still, this was a fun little project. I highly encourage you to take this for what it is: a fun, light-hearted look at this crazy world we date in. If nothing else, it makes for a great opener.

You can download the raw data and statistical analysis here.

ADDENDUM

I’m flattered by the response this experiment has gotten — thank you to all who’ve read, responded to and shared this little project of mine. I now have a big announcement for all of you.

I’m happy to announce that, since I published this post, I’ve entered a committed relationship and deactivated my online dating profiles. My days of conducting dating data experiments are (for the time being, at least) over.

In case you’re wondering: she wasn’t part of this experiment. But my conclusion did hold up here. I met her on Tinder, and opened with the following message:

“I’m sure you’ve seen some crazy stuff as a teacher — got any favorite stories?”

So there you have it. Swipe right, and don’t use a generic opener. There was obviously more to us going out, but now I have a real-world anecdote to go with my data.

(Oh, and when you do find someone, HowAboutWe has a separate site for couples, with some phenomenal date ideas!)

UPDATE: We tied the knot!

Acknowledgements

Thanks to Colleen Farrelly for helping me analyze this data, and generally being an awesome person. Connect with her if you ever need a statistician!

Thanks to Angélica Zetina for helping me edit. Follow her for informative tweets!

Thanks to Jenny for swiping right, and for all the fun times we’ve had so far. The best is yet to come.

Thanks to all my other Tinder, OkCupid and HowAboutWe dates. Even when it didn’t lead to anything, I enjoyed meeting each and everyone of you. You all prepared me so I was ready when I finally met the right person.

And thank you for reading. If you enjoyed this, please share with all your single friends. In the name of science, I also encourage you to conduct your own experiments and see if you get similar results. Happy dating!

Jay Rooney is a persuasive writing specialist based out of San Francisco (which may or may not have skewed his data). Be sure to follow him on Twitter (@RooneyWrites), as well as this collection, for more insights on communication and digital romance.