Background

OkCupid profiles have evolved from a web-only experience over the last 15 years. Being a cross-platform product the mobile apps have had to account for content that was created with a web-based approach. We were noticing diverging patterns on profiles that were created on the website and those created on an app — e.g. profiles created on web included links to videos, memes, images, and other media, typically had longer prompt responses and photos that weren’t of themselves. On the other hand, profiles ones created on apps had better photos but more selfies, shorter prompt responses with the use of emojis, and fewer questions answered.

Despite all these issues, OkC still has and collects more information than any other dating app, but parsing that info on a small screen is hard. We support several types of content including but not limited to photos, prompts, questions, details, interest topics with related prompts, and badges.

Quantitative Analysis

Members on the platform have an opportunity to send anyone they like one message that the person can see before matching with them.

Intros can currently be 

• Global (not tied to any specific content) 

• Comment on a photo

• Comment on an essay response

Currently, members can send intros to as many people as they want, however the flows on the right show that it’s not very easy to do that.

Only 5.7% votes encounter the first flow. Everyone else would have to go through their long list of likes (second flow) to send an intro.

Vision concept:
Profiles as stories

Pair content and photos to show a small number of compelling “stories” followed by progressive disclosure

Standard stories:

1. Primary photo + prompt response

2. Photo + Structured details that match viewer’s preferences and filters + overlapping interests

3. Photo + Dynamically chosen content from either overlapping prompt responses, profile’s most liked/commented on responses

4. Photo + Dynamically chosen recent or overlapping important questions

5. Instagram and other social links

6. Match score breakdown

Move away from swiping and global likes, moving towards high intent Likes for specific pieces of content, suggesting conversation starters

Scaled back

Intersperse photos and content to increase visibility of other content

• Reduce scrolling height by truncating long responses

• Ability to Like specific content and send comments as intro messages

• Leverage social media content for sparse profiles since people tend to spend a lot of time crafting their posts

• Split test with a swipe-right enabled cards, and one where members must send content-specific Likes and only left swipe for passing is enabled

Profile Creation: Prompts and responses

PROBLEM: Balancing openness and honesty with wit, more jumping off points for potential matches.

We saw in several research sessions for on-boarding that new members would be stumped when filling out a self-summary which was formerly just an open text field with a prompt to “describe yourself”.

Several participants would fill out placeholder self summaries knowing they were generic, and would say they’d come back later to fill out something more meaningful.

Running an analysis on how often the self summary changes in the first week, we found that members don’t actually go back.

The new test shuffles aims to help members write something that reflects their personality while revealing something personal about themselves

• Members write 3 prompts instead of just one

• Ability to choose the prompts and categories

• Placeholder content serves as inspiration for both content type and length

Profile Creation: Photo inspiration

PROBLEM: People stumble on which photos to add to their profile while maintaining a variety, viewers feel like they cant’t see what the person’s life looks like.

Before this change was rolled out, new members were shown a rather intimidating empty photo placeholders and only required one photo. The new test was designed to learn if we could make meaningful impact with a low lift:

• Require 6 photos instead of just one

• Provide inspiration for types of photos to upload

• Shuffle from 24+ inspiration categories to avoid homogeneous profiles

Profile Creation: Connection Types

PROBLEM: Balancing the sharing of real intentions with judgment and narrowing the pool

Our research revealed a real need for understanding and sharing intentions. Prior to this change, OkC members would have to choose between options that didn’t necessarily make sense, with higher intent options being time-boxed which further confused new members. Several participants revealed they weren’t quite sure what the difference between short and long term dating meant, or wanting to meet on a neutral ground and see where it goes.

Several options were tested to land on the ones that participants felt like they knew and understood, as well as felt comfortable sharing on their profiles.

• Removing friendship to avoid a cop-out

• Making the terms seem relatable and non-judgmental

• Adding a neutral dating ground instead of friendship

Prior to this change, new members would answer 15 questions with the option to skip answering questions such as “What are your political beliefs”. Viewers would have to sift through a long list of answered questions on someone’s profile to see how they answered. Further, these answers were not structured in a way that could be searched by, or used to dynamically surface profiles based on the viewer’s answer.

The solution was to ask for structured details while onboarding. These details were split across 5 primary categories

1. Basics

2. Background

3. Profession

4. Habits

5. Family

• Using structured details to provide members with a unified preferences system which allows them to filter and sort based on what they are looking for.

• Highlighting details that match the viewer’s preference

• Emphasize overlapping details and interests to help find common ground

• Provide conversation starters up front to reduce some of the initial anxiousness around crafting the perfect first message

• Creating a badge management system and apply it as a method to find common ground, make it feel less sponsored

Closing thoughts

• Still in development when I left

• We’re trying to influence user behavior but we’re not sure how it will affect the ecosystem as we know it today

• Unpacking and balancing impact in a double sided marketplace with conflicting motivations is a sticky problem

• Asking for more information weeds out low-intent users which can decrease upstream metrics, but is a valuable tradeoff for downstream success metrics

• Having an end state enables easier decision-making in MVPs and iterative improvements