Online dating and a formula for love

Editor’s Note: Adapted from “Data, A Love Story: How I Gamed Online Dating To Meet My Match,” to be published January 31 by Dutton.

Story highlights

Amy Webb took online dating algorithms into her own hands

72 characteristics helped define Webb's ideal mate

Webb posed as a male online to gather information for her project

Many women Webb studied lied about their height in online dating descriptions

This wasn’t a part of the plan. At age 30, I was still single and had no exciting prospects.

Rebounding from a terrible breakup with my longtime boyfriend, I turned to online dating to find a suitable mate. I joined Match.com, eHarmony and JDate, a website for Jewish singles, and decided I would go out with whomever asked, as long as he didn’t seem terrifying.

The first two dates were comically bad.

Date Number One asked me out to a really nice restaurant, then didn’t offer to pay for (or even split!) the bill. He ordered a very expensive bottle of wine and two appetizers – neither of which I got to enjoy – before moving on to a three-course meal. While walking back to our cars, he suddenly diverted to a public park bench where he asked if I wanted a smoke. He then lit the shaggy end of a large marijuana joint right in front of an assortment of passersby. He mentioned something about his weed habit and impotence, but by then I was already running toward my car.

Date Number Two claimed to be an orthopedic surgeon, but about halfway into our cappuccinos I mentioned an elbow surgery I’d had and he said that his brother was an [sic] anesthologist. After struggling to say “anesthesiologist” two more times, my eyes drifted down to his forearms, where I noticed what looked like sawdust. As he got deep into the minutiae of mitering wood, it occurred to me that I was actually out on a date with a carpenter. And a lying one at that.

Listen to Amy Webb’s story at CNN Profiles

In less than a month of online dating, I came to understand that the algorithms used by dating sites are ineffective, in large part because they rely on user-generated data. Most of us tend to answer profile questions about ourselves that are either aspirational or, in my case, fast and minimal. Bad data in means bad data out, effectively crippling even the best algorithms.

I discovered that it was because of my online profile that I was going on bad dates. I didn’t feel like answering question after question about myself, so instead I copied and pasted from my résumé:

About me: I’m the CEO of a digital strategy agency that solves strategic and operational problems related to emerging technology. I lead a brilliant team that advises a worldwide client base of Fortune 100 and Global 1000 companies, government agencies, media organizations and foundations.

There were too many other fields to complete. What were my favorite books? Best places I’ve visited? What I like to do for fun? I scrolled down to the “specialties” area of my résumé, copied all of the bullet points, and, skipping all the other sections, pasted them into the “Things I could never live without”:

• Future of technology
• Emerging platforms
• Content management systems
• Monetization
• Fluency in Japanese
• Conversational ability in Mandarin
• Fluency in HTML, CSS, JavaScript, and other web languages

Of course it’s obvious now how ridiculous it was for me to just slap together my online dating profile. I hadn’t stopped to consider how badly I was representing myself during that critically important first-impression stage, where my digital self would be judged, without a filter or explanation, by potentially hundreds of men.

Yes, my online profile was bad, but I needed context. And if I were being honest with myself, I’d admit that I hadn’t thought enough about my audience. For whom was I really searching?

To find out, I launched a short-term experiment.

First I created a giant list of 72 “ideal husband” characteristics. It included everything from “likes jazz, but only jazz from the 1920s to the late 1940s” to “must weigh 20 pounds more than me at all times” to “likes selected Broadway musicals: ‘Chess,’ ‘Evita.’ But not ‘Cats.’ Must not like ‘Cats’!” Then, in order to prioritize my list, I applied weighted scores to the list of 72 characteristics. I also built a system to evaluate each and every man who I met. Unless he scored a minimum of 700 points, I’d refuse to go out with him, even once.

I also wanted to learn everything I could about my competition. So I created profiles of 10 male archetypes and spent a month as these men, interacting with 96 women, researching their methods and scraping data from their profiles.

What I discovered about successful online daters was astonishing, and it’s emblematic of things I see people doing elsewhere on the Internet.

Among the highlights: Popular profiles used aspirational language, kept descriptions short and general, and lied about certain physical characteristics (though not the ones you’re thinking).

Very early on, I’d used qualitative and quantitative analysis to evaluate language. I could clearly see that the best-performing profiles were those that read as easygoing, youthful and spontaneous.

Short profiles that express just enough information to pique someone’s interest performed best. In my case, I’d written close to 900 words—a dissertation. That put me in the bottom 8% of all profiles I looked at.

What shocked me was how many women seemed to be lying about their height. All of the 96 women I interacted with listed their height as 5’1” - 5’3”, even though the average height of an American woman is 5’4”. Though it’s not impossible that 100% of these women were below the average, it’s statistically improbable.

Popular female daters were friendly and assertive, reaching out to my profiles with casual messages that would open with “Hey” or “Hi there” and follow with “I like that you [detail from profile]. I’m interested in [detail] too.”

Shortly after I concluded my experiment on JDate, I logged back in with the super profile I’d created for myself. It was me, only optimized to attract the widest possible swath of men. More than 60 contacted me initially, but none crossed the necessary 700-point threshold.

Finally, after widening my geographic search to 100 miles, I found Brian. We chatted online and on the phone for three weeks and finally agreed to meet in person. Our first date lasted 14 hours, and for every date after I continued to marvel as his total point score increased.

I eventually showed him my 72-point list, explained the experiment and even revealed to him my scoring framework. He smiled and said that he’d expect nothing less of me. During our wedding vows, he promised to continue to score as high as he could for the rest of our lives together.

Amy’s rules for online dating:

1. Use aspirational language.

Keep language aspirational, positive and optimistic. Talk in generalities about your hopes, dreams and passions, as long as those things are not controversial. Keep your tone conversational and light.

2. Write succinctly.

Keep your profile short but pithy. Aim for between 90 and 100 words, which works out to about three sentences. Choose your words carefully. If you’re not a good writer, figure out the keywords and points you need to make, then ask a friend to help you out.

3. Use amazing photos.

Photos should focus on your waist up, unless you have amazing legs. Then it’s OK to include one or two full-body shots in your gallery. The majority of your photos should be closer up, highlighting your face. Make eye contact with the camera. Don’t stage a smile. Instead, try to laugh just before the shot is taken.

4. Don’t try to be funny.

If you want to use humor, write whatever you’re planning to say down and show it to some friends or co-workers. Have them read it aloud. With your tone of voice and inflection, it may be hilarious—but out of your friend’s mouth it may fall flat or even be offensive.

5. Don’t use specifics.

Avoid mentioning specific comedians, shows, books, musicians or movies unless those are top-tier attributes on your list. It’s possible to be generic about what you like while still being specific enough to sound interesting. Just because you like Louis C.K. or Larry David doesn’t mean that a potential suitor does. Unless that comedian is one of your deal-breakers, leave him or her off your profile.

6. Avoid controversy.

If you think there’s something about what you’ve done in life that may be controversial or open for interpretation in a way that disadvantages you, then leave it off. I’m talking about political or activist work here, not things like jail time.

7. Downplay your accomplishments (but only initially).

Women: if you’ve won a Pulitzer or climbed Mount Everest or for some reason own a jet, this is wonderful news—just don’t share it online. These are the types of details to work into a conversation on your first or second date. If someone introduced himself to you at a party, would the next thing out of your mouth be items off your CV? Of course not, so don’t act that way online. Most people don’t want to see a list of what you’ve done unless they’re hiring you for a job.

8. Flirt wisely.

Be careful while flirting online, since it’s easy to sound too aggressive too soon. The best way to flirt is to care deeply about whatever your date is saying and to focus all of your attention on him or her. We’re flattered when people throw attention our way. So ask thoughtful questions. Take a keen interest in the conversation. Be enthusiastic. If you wouldn’t use your cheesy line in person, don’t use it digitally.

9. Use the 20 hour rule.

If someone instant messages you while you’re online, go ahead and IM back if you want. Otherwise, wait 20 to 23 hours between e-mail contacts for the first few messages. I found that successful daters waited that amount of time and as a result still seemed eager without coming off as desperate.

10. Don’t act like a stalker.

Avoid sending any messages while most people are sleeping, even if you’re wide awake. Shoot for business hours or just after dinnertime.

Watch a dramatization of Amy’s online dating adventures documented in her book.