Building a better way to find roommates

Finding roommates outside of your friend group is a lengthy and frustrating experience. People turn to Facebook Groups and Craigslist only to spend weeks or months in search of compatible roommates.

oWow is a San Francisco-based startup aiming to solve this problem by offering spatially-efficient housing at half of market rent prices while offering a solution to help match roommates for oWow apartment buildings. In August 2017, I led the team’s efforts to design the first version of their roommate matching experience.

My Role

I worked closely with one engineer (the Co-founder and CTO). As the only designer on the project I owned the process from end-to-end including design research, prototyping, UX design, UI design, interaction design.

Timeline

6 weeks, August — October 2017. The project has since been put on hold.

Constraints

Match four roommates together in a single lease. We had six weeks to arrive at our MVP solution and hand over to the development team. All of these constraints were given by the CEO and CTO.

Problem

Finding roommates is a lengthy and frustrating experience that requires sifting through housing listings, asking for and sending strangers social media profiles, and initiating conversations with people that only lead to a dead-ends.

Solution

How we got there

We focused much of our time in the early stages understanding and defining the problems with finding housing and roommates. We sought to uncover the decision-making models people use to assess potential roommates. We prototyped early (within one week of starting the project) and got feedback often. Sometimes we were wrong; but we didn’t fret — they were only prototypes. As we grew more confident in our solution, we spent more time making it pixel perfect and ready for development.

Understanding the current experience

We sought to understand what the existing experience of finding a roommate was like. We talked to users, interviewed people in coffee shops, and talked to internal team members with knowledge matching roommates. We also reflected on our own experiences finding roommates — a perk of solving a problem you’re so close to.

While asking for roommates at the apartment building you’d like to live in seems like the logical thing to do, unfortunately it won’t get you very far. Nearly 100% of these apartment communities distance themselves from roommate-matching so Facebook Groups and Craigslist have become the most popular way to find roommates. Alternative solutions include apps and website such as RoomieMatch, Room8, and Roomster and peripheral solutions include Homeshare and Common.

We found Facebook Groups and Craigslist to be crowded with new posts in the 100’s per-day in the Bay Area alone. With no way to narrow down location within Facebook Groups, users have the tedious and time-intensive task of sift through dozens or hundreds of posts from people looking for roommates.

Craigslist offered a way to search by location but offered little in the way of assuring safety or showing personality fit. Craigslist posts offer zero photos of the roommate and minimal text about him/her, but ask repliers to send links to their Facebook and/or LinkedIn profiles. This kind of information-sharing asymmetry was uncomfortable to most of the people we spoke with.

Top-left: a Craigslist poster asks for social media links from responders. Top-right: RoomieMatch bombards my inbox with potential matches. Bottom: Two of the many Facebook Groups with 30,000+ looking for roommates.

It’s also worth noting that all of these platforms seem to give users a sense that they are wasting time — something users looked at as a cost of doing business when searching for roommates.

Uncovering the people problems

Searching for roommates is a lengthy process that requires dozens of hours sifting through listings and initiating conversations with strangers.
  1. Why the demand for roommates? High rent prices in desirable areas, modest incomes, a desire to save money, and a desire for social connection fuel the need for roommates.
  2. Why not room with friends? Many are unable to fulfill their roommate needs by looking only within their friend groups. Start and end dates for the lease and location needs (e.g. proximity to work) not aligning are two common reasons friends don’t end up being roommates.
  3. What is the existing roommate search experience? Outside of friend groups, the existing experience finding roommates takes weeks or months. Many fear housing scams and creepy experiences meeting potential roommates online. The process is filled with no-replies and dead-ends which ultimately leads to a long and frustrating experience.

How do we know they are problems?

Qualitative feedback from users told us people spend weeks — sometimes months — in search of the right roommates. Multiple Facebook Groups aimed solely at Bay Area housing have 40,000+ members. The oWow team had experience matching roommates in their existing buildings and attested to the problem. And I, freshly relocated to San Francisco, whole-heartedly empathized with our users.

How will we know once we’ve solved these problems?

People will sign a lease with oWow within 5 days. Our ultimate goal is to get people to sign a lease for our buildings. “Within 5 days” comes from our belief that time is our enemy — that is, the longer it takes a user to find good roommates, the less likely they are to sign with us. We want to minimize the time between “Yeah, I think I’d like to live here” and “Great! Where do I sign?”

Starting at perfect and working backwards

We thought through what a “perfect” experience searching for roommates would be like and worked backwards until we were comfortable building a solution that was within reach. Storyboarding the experience made our thoughts more tangible and allowed us to get better feedback from team members.

Storyboarding the user experience and getting feedback from team members outside of Product/Design/Eng

Interesting discoveries from the field

“Relatively normal” is the baseline. Early-on we placed a lot of emphasis on identifying and listing the things people care about in a roommate. We thought those would likely be interests, hobbies, and places visited. But our user research taught us that most people just want to make sure their roommates are relatively normal people. Finding someone that won’t end in a roommate horror story is a must-have. Finding someone that shares a love for cryptocurrencies is only a nice-to-have.

Pictures > all else. Pictures were consistently the single most valuable piece of information when assessing a potential roommate’s profile. They show information about someones lifestyle, interests, mood, character, friends, hobbies, style, and so much more. I guess a picture really does say a thousand words.

People want to understand social context. A common millennial term, “Facebook stalking” means looking someone up on Facebook or Instagram to find out more information on someone. And while we know pictures play a large role in that behavior, we also found users want to understand a potential roommates online behavior and see the social context behind photos and life events.

People want to see that their potential roommate is relatively normal and that other people like them. Most do this by looking at pictures, social media profiles, and meeting the person face-to-face.

Outside of dating, dating app interfaces come with a stigma. We tested roommate profiles that resembled dating apps (it’s hard to not resemble a dating app when matching people). While this formed an easily digestible mental model, people told us dating app-esque experiences for anything other than dating were “weird”.

Reaching out to strangers often leads to a non-starter. Even when reaching out to a stranger is the only path to finding a roommate, users find it uncomfortable. “What do I say?” “How do I know they’re still looking?” are common hesitations.

Guessing less and failing fast

One of our major early bets was “Do not be another messaging platform.” We felt strongly that messaging platforms are aplenty and there are a few clear winners in the space. Adding another platform for messaging potential roommates would only add to the user’s pain.

Based on this bet, we tested creating profiles through a responsive website and being introduced to potential roommates through SMS texts.

But users were confused by our method of using Twilio to protect their phone numbers. We believed most communication would be taken outside of our control and found it difficult to explain why the number a user would be texting and calling was not the actual roommates phone number.

Left: wireframing “Paired matching”. Right: Defining flow for SMS introductions.

At the same time, we were testing another major early bet — something we called “paired matching”. Paired matching was the idea that, when filling a four bedroom apartment, if I can just fine one person I’m very confident I’ll like living with, I’m not as concerned about the other two. From the business perspective, we believed matching 2 people first (as a pair) and then matching that pair with one other pair would make decision making easier and speed up the overall process.

Simple illustration showing our rationale behind our early bet “paired matching” to fill the four bedroom apartments

While we didn’t quite fail on this bet, we believe we’ve found a better way and have since reevaluated the way we design around the idea of finding at lease one roommate you like. More on that shortly.

Getting feedback on prototypes along the way

Check out a later prototype which tested our chatbot interface.

Forming and designing our big bets

Bet 1: Take out the guesswork; don’t overwhelm.

A key value-add for our platform is simplicity. RoommieMatch, Craigslist, and Facebook Groups all require users to sift through dozens or hundreds of listings and potential roommates to find one suitable for them. By only recommending users download the app after they tour and are interested in the building, we create a pool of people with high intent and shared interest. We also only introduce you to one or two groups at a time — a distinctly better experience than being bombarded with emails or sifting through hundreds of housing posts.

Bet 2: Help ensure safety.

Every person’s Covalent profile starts by signing in with Facebook. We explain this in onboarding to show how seriously Covalent takes safety. Additionally, each profile is reviewed by a human

Bet 3: Be a catalyst for action.

A clear theme in our research was that many conversations led to non-starters. We didn’t want that to happen on our platform so we incorporate Heather into every conversation with suggestions and call-to-actions along the way.

Bet 4: Introduce people in groups of four.

Introducing people in groups of four maximizes the chance any given roommate finds someone he or she likes in a group. We’re still betting that you’re better off finding one person you really think you want to live with — but by introducing you in groups of 4, we’re tripling the chances that you find that person.

Iterations & designing details

A few challenges in the later-stage design stages along with iterations of how we solved them and why.

Challenge 1: How to make an intuitive interaction when replying to Heather, our chatbot? Solution: Make quick reply buttons resemble “sent” messages.

Challenge 2: How to make names visible and profiles accessible when in a group chat with potential roommates? Solution: minimal information on message screen + swipe down or tap top-right button to expand.

Challenge 3: How to design a rating interaction that is fun and intuitive? Solution: Three emojis — “yes”, “no”, “not sure”.

Challenge 4: How to seamlessly navigate and search through the app? Solution: gesture-based navigation throughout app (similar to Instagram) with Search button always available on menu.

Next steps, reflection, and learnings

As project moves into development, we will focus on designing more details like interactions and enhancing visuals. I will also support the engineering team (try to make their lives easier) and help define what we need to measure so we can make data-informed decisions post-launch.

Overall, I’m very satisfied with what the team accomplished in six weeks. At the top of my list of learnings: the importance of prototyping rapidly by ignoring my pixel-perfect tendencies.