My Experience Building Chicory
I came up with the idea for Chicory two years before I started building it. I felt unprepared to pursue it because of the fatigue incurred by my previous venture. This changed when I was invited to MHacks by a close friend in February 2013. I was inspired by the drive and excitement of the people I met there, so I started building the application. I was able to build a proof-of-concept by the end of the competition. Several months later, I decided to dedicate all of my free time to Chicory in addition to being a full-time college student.
The initial prototype was a shell script that parsed two text files together to produce a parallel text inside of a static HTML document. I synchronized the audio with the text by using timestamps. I didn’t record any audio yet, but I did have a German audiobook laying around that I used as a sample. I timestamped the first chapter by hand.
At this point I had to flesh out the user interface, so I added a CSS framework to avoid cross-browser layout inconsistencies and began building the interaction functionality. The flow of user interaction is as follows:
- The user clicks on a paragraph
- The paragraph is highlighted and the corresponding audio plays
- The next paragraph scrolls into view and step 2 is repeated
It took only a few weeks to go from concept to working prototype.
Testing the Water
I decided to show my project to an online community of language hobbyists. I described the problems that I was trying to solve with the listening-reading method and many people replied with enthusiasm and were anxious to try it. I told them I would have a demo out within a few weeks.
As I kept them waiting, they grew anxious and began watching the news page very closely, reporting the most minor changes to the community. This surprised me; They actually cared about it. One of the language enthusiasts replied with a story about how they had trouble keeping track of where they were in the text, and how they wanted an easier way to replay audio. Several others replied with similar stories. Chicory solved this problem.
I created a mailing list to keep them updated and continued to interact with them on the message board. At this point, I had verified that the problem I was solving exists for this group of people and that they were interested in my solution.
I did not release the demo on time and the language hobbyists became anxious – some were even angry. My decision to delay the demo was foolish. I felt the need to enhance the user interface, and along the way decided that adding a datastore and user accounts was a good idea, but the best thing I could have done at that point was to release the product immediately and begin receiving feedback. The lesson here is to release the product as early as possible to start the customer development feedback loop.
In the beginning, the user interface was just two columns of text with an audio player at the top. This was perfect for my minimum viable product. I didn’t have a dashboard or an interface to filter through the available books, but I convinced myself that this was “necessary” and began a month-long expedition to redesign the interface. I spent days playing with new designs and scrapping them after realizing they were terrible. I experimented with an “ebook reader” interface that featured a “bookshelf” as part of the dashboard. These were all far too complicated and added nothing useful.
Instead of focusing on features, I should have directed my attention toward how the user benefits from my product. Features mean nothing outside of the context of the user. If it’s not crucial to the user experience, then it doesn’t belong in the product. Identify the key aspects that make the product great and build those.
While building the prototype I started looking for voice actors. I began by emailing every teaching assistant in the foreign language department (approximately 150) at Purdue. I drafted a long email to fully explain the project, which resulted in an abysmal response rate. This was obviously a poor approach, but I did meet with those who responded to see how we could both benefit from the relationship. I met with several candidates over the course of a month and set up more meetings for when the others returned from their summer break.
This didn’t work out well because I was not specific enough about what they would be doing and was undecided on which languages I would record first. My message lacked clarity. Once I had decided to focus on one language, finding candidates became incredibly easy. I wrote a new email with a clear and concise message and sent it to the Chair of the Chinese department at Purdue. I stated that I had built an application for learning Chinese and was seeking native voice actors. By the next day I had set up meetings with two candidates.
I met the candidates at a busy cafe where I showed them the prototype and had them briefly demonstrate how they would voice act by giving them a section of a Chinese story to read aloud. These went well and I set up another meeting for each to teach them microphone techniques and coach them on voice acting. At subsequent meetings where we actually recorded the material, I continued to coach them on how they could improve. Neither of them had any experience voice acting, so it was a rough start, but over time they both grew tremendously into quality voice actors.
The Recording Studio
When I began, I knew nothing about microphones or recording. I spent several days reading articles and watching youtube videos to learn about microphones, recording techniques, voice acting, and how to set up a proper recording environment. I settled on a USB condenser microphone so I could record directly on my computer. I later added a microphone stand, shock mount, and pop filter to improve the quality of the recordings. The hardest part was finding a place to set up the studio.
I searched for weeks to find a suitable location to set up my studio. I tested empty classrooms after-hours, the city library’s private rooms, and even my own dorm room. None of them were good enough, either because of echo or background noise. I finally settled on using a tiny conference room made of concrete in the basement of the new campus co-working space. Outside noise wasn’t a problem here, but the echo was still pretty bad. I bought several moving blankets which I suspended from pipes on the ceiling using hooks and wires to form an enclosure. This dampened the echo significantly and – although not perfect – was good enough.
I met with each voice actor once a week for two to three hours, unless they cancelled because of other plans. The first few sessions were spent practicing with them by going over techniques. I provided them with feedback as we recorded to improve their technique and get them more comfortable with voice acting. Although I recorded two books simultaneously, a different one for each voice actor, it took almost two months to record what amounted to less than two hours of edited audio. Most of The True Story of Ah Q (our first Chinese book released) had to be recorded twice.
I fixed a lot of mistakes during editing, including the removal of breathing and ambient noise. I fixed the pacing by adding or removing silence where appropriate, which improved the quality significantly. It took two to three hours to produce just ten minutes of audio, not including the time to record.
I had fun while producing the audio, including the equipment set-up, coaching, recording, and editing. It felt good to hand-craft an experience for my users, even if I was impatient about how long it was taking to get anything done.
Despite my belief that doing my own audio production was a worthwhile experience, with regard to the business it would have been better to outsource the audio production to a professional voice actor. The monetary costs would end up being equivalent, but using a professional would have taken far less time. I viewed the equipment and training as a long-term investment, but failed to realize that I needed to validate the product before making such an investment.
From the outset, I realized that I may run into legal problems if I wasn’t careful about checking that each book was in the public domain. This meant I spent a lot of time researching books. There is a lot of ambiguity in copyright law, so this was frustrating. Even if the English version of the book was in the public domain in the U.S., I often had to make an assumption about the Chinese translation based on the publication date. I was also searching for books in many other languages at the time, including Arabic, Japanese, and German.
This made me more aware of covering my liabilities elsewhere in the business – specifically with regard to who owned the audiobooks that I was recording. I found several voice acting work-for-hire contracts filled with legal jargon and attempted to modifiy them to suit my needs. This was useless, so I abandoned the idea and instead scheduled an appointment with a lawyer at the university. That meeting sealed my decision to just do it and ignore all of the legal rituals. Instead of asking for permission, I decided I would ask for forgiveness.
I never ran into any problems and am satisfied with my decision to forgo the bureaucracy. Had I pursued all of the legal avenues recommended to me by the lawyer, I would have been set back several more months and several thousand dollars (which I did not have). The lawyer was just doing the job of a lawyer, which is to advise their client on how to prepare for the worst possible scenario. This did not apply to me, however, since I had no assets to lose.
Business Plan Competition and Market Pivot
Purdue holds an annual business plan competition where the winning team of the undergraduate division can recieve up to $20k. Even though I didn’t need the money, I thought it would at least help me think through the execution of my business model. Instead, it hindered my thought process, because I wrote my business plan in an attempt to win the competition, not to create a successful business.
My focus for the competition was on proving that there is a large and growing market for my product. Starting a new company is an experiment to determine if a product will be successful with a particular market, meaning that you are searching for product-market fit. The problem is that the competition judges want proof that you have already achieved product-market fit.
A business plan is a detailed execution plan for a proven market, so I needed to find a “proven market”. I targeted college students with the naive assumption that they actually want to learn a foreign language. It’s one of those things that everyone says they want to do, like exercise more, but how many people actually do it? Why was I targeting college students when I already had a proven market of language hobbyists who were dying to get their hands on my product? The language hobbyist market is not large or growing. This meant it was not suitable for the business plan competition. This set me down a long road to nowhere.
Since the listening-reading method does not appeal to casual users, I had to dumb down the concept considerably to make it more palatable for sporadic use (ten to thirty minutes rather than two to six hours) on a mobile device. This also solidified my focus on the mobile platform and directed my future design decisions. From then onward, my foremost thought was always, “How can I make this work for students?”
After realizing that few college students would willingly expend time and effort learning something that was not required for their degree, I focused my effort on college students enrolled in language courses. I thought I may be able to sell it to the teachers as a homework assignment or lab exercise. This fundamentally shifted my market from students to teachers! This was where it all came unravelled. I was building a product for people who don’t care about it, and someone else is paying for it.
Although I made it to the semi-finals, I did not get into the final round. As soon as the competition concluded, I set my focus back on selling directly to college students enrolled in language courses, rather than selling to the teachers. This was obviously still wrong, but not as wrong. Unfortunately, that doesn’t matter in business. I ignored the language hobbyists, who were anxious to get their hands on my product, to focus on college students, who didn’t care about it at all. Writing a business plan will skew your thought process and encourage you to focus on the wrong things. I won’t write another business plan. Ever.
The latest redesign of Chicory happened while I was working on my entry into the business plan competition, meaning that it was built for mobile devices. The dashboard and chapter selection screen were the major focus since the reading application only needed a typographic update. I noticed that the text never looked quite right, so I began researching typography and applied what I learned to make it beautiful and easy to read.
About a week after launching, I finally fixed the inconsistency with the audio player across browsers. Each browser had its own interface for embedded audio, and in some it failed to even load. While sitting in my local cafe playing with Chicory on my tablet, I came up with the new design and rushed home to sketch it on paper. Within two nights I had a custom audio player interface that worked across every major browser.
Although I was able to easily solve technology problems with the listening-reading method, the hardest problem to solve was finding content to use. Unabridged audiobooks in foreign languages that match the book exactly are difficult to find. It is even more difficult to find a side-by-side direct translation of a book. I had been struggling with these problems since day one, and was producing the content myself. The only viable solution was to set up a platform to allow users to produce and sell their own audiobooks to other users.
I was inspired by Valve’s model of user-generated content, by which their users create higher quality content faster than Valve’s own employees. I developed an execution strategy for this marketplace, but only implemented a small portion of it (which explains the “artist selection” feature underneath the audio player) before realizing that I didn’t have the right product for the college student market. The language hobbyist market was not yet large enough for this to work properly, so I would have to continue producing my own content – at least until I had enough traction to create a new market for language learning.
I wanted to present Chicory as a legitimate business, so I set up an email server on the getchicory.com domain. It was cheap and easy to set up. It made me feel more engaged with my users because I had a private channel to connect with them.
The mailing list I set up made it easy to announce the launch to almost 200 subscribers and track how they responded. I got feedback immediately. Some had questions about the user interface and others wanted to point out errors in the content. Every email made me happy because it meant that my users cared about my product.
Unfortunately, the emails soon stopped and user engagement dropped. I wasn’t attentive enough to their needs because I was focusing too much on getting in front of college students. Failure to consistently engage with my users was harmful. A company is not just the product – it is the people who build, sell, and support the product. Every interaction a user has with your company matters.
In my final attempt to capture the attention of the college student market, I scheduled a demonstration for two morning sections of Chinese 102 about one month after releasing the product. “Are there any questions?” Forty blank faces stared back at me. “Did they not understand my pitch, or do they simply not care?” I wondered. Only a few hands went up with questions. As the students filed out of the classroom, I realized that they were not interested. Later that day, only a few students had signed up. This was the point at which I decided to give up on this market. I had burned myself out and needed to take a break from Chicory.
I built a minimum viable product in just a few weeks, but failed to launch it immediately. This created unnecessary stress on myself to “succeed” and resulted in continued delays. I spent months adding uneccessary features (i.e. datastore, user accounts, security) and conducting aimless redesigns to please a mass audience. My advice to my future self is to focus on one market and launch as early as possible. Start the feedback loop now.
Content production was also a source of delays for the product. In retrospect, it is obvious that I should have contracted a professional to record the initial audio books, but I was unsure at the time. Bringing the content production in-house makes sense for an established organization, but Chicory was not ready for such an investment. Building a company requires experimentation, and every failure leads you closer to a working solution.
One of the traps I fell into was feeling guilty when I wasn’t working on Chicory, even if there was nothing crucial that needed to be done. Will changing the location of a button have a huge impact on the product? Probably not, but it’s something I can do to create the illusion of productivity. Putting in more hours will not result in success. This is the wrong approach. It caused me to burn out and my productivity dropped dramatically.
After taking a break from Chicory for several months, I gained insight from a product meant to entertain: Far Cry 3. In the game, you upgrade your equipment by completing objectives that require you to hunt various animals. Each time I unlocked a new item I was congratulated and the item was checked off my list. This inspired me to change the way I set goals for myself.
Instead of setting monolithic goals whose completion date is severals months in the future, I decided to set very specific goals that can be achieved in no more than two hours. Incremental progress is a powerful concept that allows me to get a lot done without feeling overwhelmed, and also lets me react quickly to changes in my environment. My productivity has gone up significantly since I began using this approach.
I am satisfied with what I learned from building Chicory and would like to continue experimenting with it, but it has since become a hobby. All of the lessons I learned are applicable to future ventures, so even though Chicory didn’t make a profit, I consider it a personal success. There are many other things I want to explore and build outside of language education applications. Tomorrow is a mystery to me until it arrives.