[Product Teardown] Love, Sex, and Operant Conditioning - How Tinder Makes You Stick

Sep 2021 6-minute read

Following my first product teardown on my journey learning product management, here’s another teardown on the psychology tactics used to drive retention in Tinder.

(And since learning in public is all about making mistakes and sharing them, here’s mine - I’m running out of products to analyse for the “psychology tactics used to drive retention” theme. #oops 😅 I’ll probably include writing PRDs and competitive analysis from here on.)

For now, let’s get swiping!

What is Tinder and how do you use it?

From their website, “Tinder is a global online dating platform where you can meet new people, expand your social network, or meet locals in 190+ countries.”

It “connects you with profiles using location-based technology based on the gender, distance, and orientation filters you set.” Most notably, Tinder uses the concept of swipes to denote Like or Nope, and a match happens when both you and the other person swiped Like on each other.

How does Tinder make you stick?

Based on my observation, retention tactics on Tinder boil down to this formula:

Retention = Operant conditioning + Catering to people’s needs

I’ll explain why.

[Variable Ratio Operant Conditioning] You don’t get a match every time

First described by behaviorist B.F. Skinner, operant conditioning relies on a fairly simple premise: Actions that are followed by reinforcement (reward) will be strengthened and more likely to occur again in the future.

Reinforcements can be given at a variable or fixed ratio schedule.

Tinder adopts a variable ratio operant conditioning model. Rob Haisfield explained this beautifully,

Every time you swipe right and get a match, you get excited about the possibility they represent and you feel good about yourself, which rewards you for swiping and encourages you to do it some more. However, you recognize that you won’t get a match every time you swipe right, so you don’t get discouraged if you don’t get rewarded right away and you just keep swiping.

To dive into how Tinder does this, I viewed 120 profiles of both men and women in San Francisco. I broke the profiles into sets of 20 profiles, tracked how many people I Like, and noted how many matches I get immediately.

I looked through 5 sets of profiles (100 profiles in total) in one day and another set of 20 profiles two days (6th set) after to see if my usage of the app changed anything.

There’s an average of 14% chance of finding a profile I Like and a 1.7% chance of being matched with someone I Liked immediately. There was no significant difference when I used the app again a couple of days later.

This is quite… sad. I got curious to dig deeper so I swiped through another 120 profiles, this time using Tinder like a man liking all the profiles shown to me. I also did the 6th set of profiles 2 days later.

The odds of matches improved. It is now 33% as opposed to the previous 14%, more than 2X the chances of being matched. While there’s no scientific rigor here, a 33% chance of getting matched is still sad. I was puzzled at why Tinder offered such poor chances of matching, then I realised - it’s the business model.

Enter the money

Tinder offers various subscription tiers to allow you to look through people who have liked you and to get your profile seen by people you liked.

The free tier just needs to give you a tiny taste of what it could be, which itself is a big draw because

[Maslow’s Hierarchy of Needs] You can find love & sex on Tinder

Maslow’s hierarchy of needs is a motivational theory in psychology comprising a five-tier model of human needs, often depicted as hierarchical levels within a pyramid.

All that operant conditioning wouldn’t work if it does not cater to one or more people’s needs. I mean, if Tinder was an app that shows you photos of marbles to like or pass and give you marbles based on that, it wouldn’t be an app with 66 million monthly active users, would it?

What makes Tinder successful is that it caters to people’s innate needs, particularly around love, relationships, and sex.

Another example of variable-ratio operant conditioning working really well is slot machines because it caters to people’s desire for money. Although money is not listed on Maslow’s hierarchy of needs, money can buy a lot of it.

How would I improve Tinder?

Tinder is perhaps one of the most popular swipe-based dating apps out there, but it’s not an app I’d use every day. In fact, I find it almost a waste of time. To better drive retention, I’d suggest a better play of the psychology concepts discussed here:

Learn what users like better and quicker:

14% chance of finding someone I may Like is sad. I wished Tinder could learn from my past likes, based on facial and profile similarities, and better surface profiles I may Like. It would make it desirable to pay for Tinder Platinum.

Offer random free Tinder Gold/Platinum days to users:

I can understand why the free tier is designed to match poorly (1.7% chance) to upsell the paid tiers - there’s only so much a product can do independent of broader ecosystem stakeholders such as revenue needs, jurisdiction, partnerships, or competition.

To mitigate that, perhaps Tinder can extend their play on variable ratio operant conditioning a level higher by randomly offering people free use of their paid subscription plans for a day. This is akins to slot machines where you don’t know when you’ll win, so you keep performing the same action until you do. This should drive app-opens retention.

Improve on safety features:

Tinder murderers are real. As a woman, my safety is as important as finding love, more so as I live in a country where I could be murdered because a man was having a bad day. Tinder focuses quite a lot on safety, but I feel that there’s a huge unmet gap between informative guides, reporting features, and emergency call button….. yes that’s right, I’m talking about stalking and awkward first dates.

Nailing (lol) the safety features around that could make Tinder a more compelling product for women, which they certainly could need more of (27% of users identified as women on Tinder). Hint: PRD coming soon.

How was this teardown?

I am very swiped out working on this teardown, how did I do? Any feedback on my suggestions or how I could improve? Let me know!