Chatbot Guide to Your City

While iterating through different concepts at Ten Percent Happier, we spent a lot of time reading books about behavior change and psychology. One concept stuck with me: happiness is derived from the experiences we live and shared with others. After I left Ten Percent Happier, I spent a few months experimenting with a new idea inspired by this simple truth.

To test the concept, I posted a Google Voice phone number in the Boston Subreddit. Instead of building something, I experimented with different conversation flows manually.

Prototyping a New Concept

As I reflected on my own lifestyle, I thought about moments in my life that felt adventurous. I characterized adventure as something new and possibly out of my comfort zone. This definition encompassed long trips abroad, but also small local experiences that I hadn’t pursued before.

A pattern emerged in conversations with others about this topic. Trying something new can be nerve-racking. Think about the first time you tried improv comedy or went skydiving. A pit forms in your stomach, your heart rate picks up, and your palms may start to sweat. As we try more things, the uncomfortable become more comfortable.

Some of us are fortunate to have an adventurous friend; someone who seeks out new experiences and brings us with them. I wondered if I might be able to distill the leadership this friend offers us and turn it into a service.

One Saturday morning, I launched a “see what happens” prototype. I wrote a post on Reddit/r/Boston about this promise of helping people break from their routines to explore more of what Boston has to offer. In the post, I included a prompt to text a Google Voice phone number to start.

As I started to field more conversations, I built out a simple Rails + Bootstrap application to host information about events that I recommended to users.

Early Successes and Failures

Within a month, a few hundred people had texted the phone number. I built out a simple Ruby on Rails application to save events and experiences that I would share as recommendations to users.

My friend Paul joined the project and began leveling up the Rails application to scrape event data from Facebook and other websites. Paul also began working to automate some of the conversation flow that I had refined through manual experimentation.

The numbers looked good. We polled users and recorded a 70% NPS score. Retention (measured as users who returned once per month) was at 30% and the service had grown organically to over 1,000 conversations per month. But we hadn’t reached product/market fit yet.

It was painfully obvious that we didn’t know who our target customer was. We were catering to Boston locals and folks traveling to the city. We helped plan date nights for couples, family experiences for parents visiting their kids in college, and weekend escape from the city for groups of friends.

We also didn’t have a clear path to sustainable revenue. There was little appetite for a paid service on the user side. Some local venues offered us a small fee for promoting their events, but this was insignificant. I was concerned we were building a vitamin, not a pill; we were a helpful service, but users could live without it.

The user side of the service consisted only of a text message conversation that would occassionally link out to an event card on my website.

Shutting Down Ferris

After a few months of running the service, we shut it down. There are a few things I learned from this experience.

First, good experiments have well-defined goals and constraints. There was a very positive response to the simple prototype that I launched. Had I limited that prototype to a three week experiment, I would have spent more time properly reflecting on learnings from that prototype. Instead, I was too wrapped up in making small adjustments to step back and learn from the experiment.

Second, I relearned that I’m mission-driven. This is a lesson I often forget. Building a company takes time, and this simply wasn’t a project I was willing to spend years creating.

Third, time spent outside of work is time well spent. I hit burnout while trying to build Ferris. Finding time to exercise, cook a meal at home, get eight hours of sleep, and spending weekends with friends and family — these are all things that make me more effective when I’m at work.

Paul and I also started to build out some internal tools to help improve the service. This is an incomplete dashboard concept that would allow me to view recommendations previously made to a user while looking for new recommendations to make to that user.