Key Postmortem

Key was the culmination of my years exploring in the consumer social app space. The core premise of Key was that peoples’ experience at events was primarily determined by the people they interacted with there- both friends they already knew and someone new that they just met.

I recruited a few of my brightest peers: Tucker Haas, Jiren Zhu, and Yuetong Wang to join me on this endeavor. We rented a house in Palo Alto (785 Newell!) that was both our office and living space. This was an exciting summer filled with long, hard work and very rewarding payoffs when we saw the joy we created for our users.

A major part of our strategy was to host events in San Francisco as a “sandbox” for testing new iterations of our app. We planned our Launch Party before we had even built the first version; however, this acted as a forcing function on our productivity. We had 3 weeks to build out our first version:

In hindsight, this version was very rough and had a lot of elements going on. The knowledge we learned from talking to users at our Launch Party was informative to help us understand the nuance of behavior from 18–40 year-olds in SF at events.

Pictures from the launch party: The Key Experience

Our more polished next iteration aggregated users in the different parts of the city to allow people to break the ice with anyone in their area:

It was difficult to kickstart the chicken-and-egg problem of getting users and content simultaneously. We had an insight where we realized we could piggyback on Facebook events and allow people that were going to the same event to match ala Tinder before and during the event. Here are the screenshots from this iteration:

From our next event, we realized that our users were unclear as to the intent of the matching feature. Why were they swiping on people? We clarified this by adding additional nuance around your intent for specific events. Did you want to flirt? Meet people? Network? Get Drinks? Dance? Or Chill?

This version had legs! We had thousands of users in San Francisco using our app 5 or more times a week. It was a fun and engaging way to figure out which event you wanted to go- based on who you matched with ahead of time! Users on average were swiping over 100 user cards for each event that was active (more than 15% of the people attending the event on Facebook were users of Key).

The magical experience that we were trying to provide for our users was as follows. Open the app before you go to an event, swipe on a bunch of other users, get some matches, and when you go to the event see some of the people you matched with. As you might be telling yourself now, each of these steps required the confluence of many random variables. The percentage of the matches that resulted in a real-world meeting was less than 3%. This was not going to work. What was the key driver of matches not meeting at the event? The answer is very simple: people are flakes. From our analytics, people attended Facebook events that they affirmatively said they were going to less than 20% of the time. Because the RSVP’s on Facebook were not reliable, if one person in a match went to the event and the other did not, then there’s obviously no way that they would meet.

This ultimately was the downfall of Key, the user experience was fun and engaging, but the magical moment we were trying to create for our users happened at such a low frequency due to the flakiness of Facebook RSVPs. This paradigm did not work. We passively watched the usage of Key and it asymptotically went to 0 when we stopped holding the events that people were guaranteed to meet each other at. We wound the company down, returned to school, and shut the app down.

The silver lining in my experience with Key was that we brought thousands of people together: new friends, new roommates, new colleagues, new investors, new significant others (including me!) through a combination of our app and the events we hosted. This was an unforgettable experience that taught me an invaluable amount about what motivates people and how to bring them together.