Prototyping a dating app without daters
We designed a dating app that, reflecting reality, bridged users' professional and romantic lives. We then built a web-based control panel to remotely manipulate the interface, allowing us to user-test the app without a user base.
Avery had previously pitched a concept for another class: a dating app that, reflecting reality, bridged users' professional and romantic lives. As we thought about possible projects, this one presented a unique design challenge:
How could we do realistic user testing of a dating app without requiring a bunch of users to sign up first? We could, of course, just do a click-through mockup, but without presenting "matches" that seemed relevant and appealing to a tester, it would be hard to get faithful feedback about the app
We were testing two main components to our concept. The first was the tie-in to work and professional interests, to see whether this appealed to users, and whether they would have feedback on the execution. The second was a new feature we added which required users to earn each of their own matches by choosing a match (from among three pre-selected potential suitors) for another user. In other words, can we make playing matchmaker to strangers fun and rewarding?
The user creates a dating profile for themselves, describing themselves personally and professionally.
At each of the profile creating screens, their profile information is sent to the control panel in real-time. The researcher using this panel can use this information to create custom, fake matches for the user.
Rather than relying on an algorithm, we wanted to explore the potential of users match-making for other random users. Users are presented a profile and 3 potential matches, and have to decide whether or not they think the pairs are a good match. Users need to match-make for 3 users before unlocking their own matches for the day.
How do we get users to engage with a dating app without personalized profiles to view, like, and chat with? Referencing the users profile information, the researcher can use the control panel to combine peices of pre-created profile data to create fake profiles that have a high likelihood of piquing the users interest!
All user actions are conveyed to the researcher via notifications in the control panel. Profile likes, visits and dismissals are all conveyed in real-time to help the researcher both understand usage in real-time and respond accordingly to drive engagement to gather useful insight.
If the user likes, say, James, the researcher can chose to create the illusion that James likes the user back, thereby creating a match. The user, or our researcher can then choose to break the ice and start a conversation.
During the Open House, we got interesting feedback on the "play matchmaker" concept. Those we talked to generally seemed to find it interesting and potentially engaging, if well designed. They also had concerns about “competition” between themselves and those they were matching for (if I'm matching for Abby, and I see Ken, and I really like Ken's profile, maybe I don't want to recommend him for Abby - maybe I want to keep him for myself!). This is great feedback, and indicates that we might need to present matching scenarios that weren't related to the user doing the matching, perhaps by having them all be from another city or something like that.
We also got questions about what happens once the system receives a "match recommendation" from a user: does it immediately go through, or must the match be recommended multiple times before being pushed to the two users being matched? If one user is better or worse at matching, will his/her recommendations count more/less? These type of questions would help us delve into what expectations users would have for how a system like this works.
Given more time, we would like to hone our procedure a bit and test with a number of users. We might also tweak the presentation to include more gamification based on success with matching (maybe a Match Score, etc).
Overall, this approach did what we set out to do: it allowed us to test a matching service in a realistic way, allowing us to get feedback without too much handwaving and suspension of disbelief. Given a little more time to hone the prototype, this method would definitely be useful in future cases where we were trying to work with social features, communities, or recommendations.