![[Dozy Project Update.png]] ## Welcome In 2019, I founded a startup, Dozy. Dozy's mission was to make insomnia treatment accessible and affordable to everyone. Towards that end, I designed prototypes, built them, tested them, raised money, hired people. In general, the full startup path as a solo founder. In the end, it didn't work out. The following is a summary for anyone else looking at a similar path to decide if it's what they should do. ## Why read this? This story is relevant for people considering solo-founding, looking at the insomnia treatment space, looking at the mental health tech space, or people who are just curious about me as a person. ## Dozy's purpose As I mentioned above, my goal with Dozy was to make insomnia treatment accessible and affordable to everyone. I got started on this for two reasons. First, I had pretty severe insomnia myself at the time, and these techniques helped me fix it, which resulted in a massive improvement to quality of life. Second, I wanted to build something that helped people using skills they already had. Quoting one of my early assessments: > Estimates on the prevalence of insomnia vary widely, according to definition and methodology. According to one 2008 paper, numbers range from 10% to 40%, and "Given all the information available, the prevalence of insomnia symptoms may be estimated at 30% and specific insomnia disorders at 5–10%". > > A 2018 paper estimated the QALY burden of insomnia at 5.6 million QALYs per year. This estimate only counts quality of life effects, not length of life, so the real number may well be larger than this. Its methods are not extremely robust but they end up with an insomnia prevalence of 28%, which is close to the number from the previous paper. > > If we assume the QALY burden of insomnia is similar among global populations (may or may not be true), then we can expand the US number to ballpark the global annual burden of insomnia at 130,000,000 QALYs lost per year. *For the full research writeup, see https://forum.effectivealtruism.org/posts/ztrYNRJ66NQeMpS64/insomnia-with-an-ea-lens-bigger-than-malaria ![[Pasted image 20240611183528.png]] *Insomnia prevalence chart from [The epidemiology of insomnia: Associations with physical and mental health.: The HUNT-2 study](https://www.sciencedirect.com/science/article/abs/pii/S0022399909001809). More conservative estimates than my initial analysis.* Cognitive Behavioral Therapy for Insomnia (CBT-i) has been recommended as the first-line treatment for insomnia care for quite a while now. Unfortunately, there are nowhere near enough providers for this in the United States or anywhere. This has resulted in a situation where most people don't even know that this treatment exists, despite it being probably more recommended than sleeping pills. And, similarly, because we live in the United States, it is prohibitively expensive, and often not covered by insurance. However, the treatment itself is not actually that complicated. Some aspects are tricky, but the overall structure is more like a flowchart, with a defined structure. I believed I could use my development and design skills to make a version of this treatment that anyone could access and afford. ## Journey The story starts in Korea, where my insomnia was at its worst. It was taking me two to four hours to fall asleep every night, for over a month. This is pretty miserable, and I would not wish the experience upon anyone. That pain led me to do extensive research on sleep and insomnia, and eventually discover the CBT-I techniques, which helped me fix my sleep problems. These days, I fall asleep in a mere 5-10 minutes. ![[sam_tired_selfie.png]] *Me around peak insomnia in 2018, taking a selfie for government documentation. Look at the sheer vitality in those eyes 👀 * I was looking for something that would be impactful for anyone seeking effective treatment. At the same time, I was getting more involved in effective altruism. I was looking for something that could be highly leveraged to improve people's lives. This led to initial research on the problem space, where I discovered how widespread insomnia is, how far-reaching its health impacts are, and how effective this treatment can be. After that initial research, I sought advisors who could help me evaluate and develop this idea further. I found a practitioner of CBT-I in Arizona who offered to help. Simultaneously, I found a researcher on this based in the Netherlands. With their assistance, I developed a more operationalized version of CBT-I that could be implemented into an app. After this, I created some user interface prototypes, which I tested on various people I knew with insomnia. The broad design guidelines I kept in mind were - Insomnia is most common among older adults, especially women. Any good solution can't require strong technological familiarity. Similarly, any app interface elements should be large, and ideally should support accessibility settings text enlargements. - CBT-I is hard to actually do. A good solution should provide as much behavioral support as possible to make sure people can complete the treatment and get the value. - Most use of the app should happen during the day, not just before bed, since we don't want to encourage phone use before sleep. - No one wants to read a textbook. The app should show us a little information, as is helpful, and present it in a clear and simple way. (in contrast to existing solutions here) - It is far more convenient to consume content and enter information on your phone versus your computer. Any good solution here that wants to help a general audience should be mobile first. ![[Pasted image 20240611184702.png]] *Early Figma wireframes of the app* After iterating on the initial Figma prototypes, I created a no-code alpha prototype in a documents tool called Coda. I ran three people through the Coda version. Got pretty substantial results - one user started getting an additional hour of sleep per night. After that came the React Native version and closed beta testing. I built the React Native app, got people to use it, and slowly started to develop an audience and a small web presence. Around this time, I gained a bit of awareness among the Effective Altruism community, and that's how I accidentally fundraised. ![[Pasted image 20240611185704.png]] *Screenshot from the Slate Star Codex post mentioning me. [Link](https://www.astralcodexten.com/p/highlights-from-the-comments-on-cbt)* Scott Alexander, the tech blogger, wrote an article about one of my main competitors in the space who is known for charging an insane amount of money for access to their app. People mentioned Dozy in the comments, Scott reached out, and the next day he posted a follow-up on his blog that mentioned Dozy's existence and the fact that I was looking for co-founders and investors. The next morning I had a bunch of emails in my inbox from potential investors and has raised an annual round a week later. This was followed by many co-founder explorations, my first hires, some contractors, and the public beta launch. From there, it was a game of iterating and improving the app to get it ready for a proper marketing push. However, it was around this point that I started hitting the more fundamental struggles that would ultimately tank the project. ## What went well - The app worked, and it worked well. People who stuck to the treatment saw results that had a similar effect size as in-person CBT-I treatment. ![[Pasted image 20240611185827.png]] *Slide from the Dozy pitch deck* - Users who completed the treatment turned into super-promoters who then told all of their friends with insomnia about the app, and this led to a quite nice flywheel effect. I had a constant and steadily-growing stream of users without any active promotion. - I used a hybrid human model where the app's educational content and base reminders would all be automated, but users had a human coach (over text) who would proactively follow up with them to help them complete the treatment. This led to a big boost in user attention, and people liked the coaches. Humans-in-the-loop systems in general seem promising. - Users liked the design and structure of the app. It was friendly and approachable, while doing what it needed to do. - Users got value from the education on sleep and insomnia. A number of them mentioned those frameworks being persistently useful to them, even if they did not complete the treatment itself. ## Challenges - Ultimately, the biggest challenge was that I was a solo founder. Even though I had employees and contractors, it's not the same as having a partner in the venture. This led to burnout, exacerbated depression that I already had, and ultimately led to my iteration speed being too slow. - Getting people to stick to a treatment is hard, getting people to change their behavior is harder, getting people to change their behavior with no human intervention may be the hardest of all. Retention rates among mental health apps tend to be pretty abysmal, in the range of 1.9% for 2-week retention. Dozy was maybe double that, which isn't bad for a free app, but ultimately was too low for what I needed. - CBT-I treatment is difficult. It requires consistent control over bedtimes and wake times and initially requires a burst of sleep deprivation from sleep restriction therapy. If people could get past that, they often had no problems with the rest of the treatment, but getting over that initial hurdle is quite difficult. Especially for an app. - People's expectations of pricing for anything that's affiliated with a mobile app tends to be abysmally low. This is a known problem, it's not new, but since Dozy was a consumer app, experiments in willingness to pay were not initially promising. ## Killing blows The main two factors in the end were the fact that it took me too long to get a complete solution to market, and that more competitors with a similar angle started popping up. This meant that the counterfactual impact of what I was doing became lower, and getting investment became harder. The too-slow iteration speed was due primarily to being a solo founder and having untreated depression, which contributed to burnout. I've grown past these now, but at the time I lacked awareness of depression as a problem that needed addressing. At this same time, there were more people entering the market with a design-first, mobile-first, direct-to-consumer approach. This meant that, first, the impact of what I was doing was lower from an EA perspective, and second, the increasingly competitive market made it harder to raise money. ## Possible merge Some friends of mine were also founding a sleep company called Crescent, since acquired by BetterUp. Since our goals were similar, we spent a few months evaluating working together on the same company, but ultimately decided against it. ## Shutdown With the company merge falling through and with no more money to spend, I shut the company down in December of 2021. It was actually pretty cathartic. I think I had known it was coming for a while before it actually happened. ![[Pasted image 20240608183729.png]] And I had fun with it. I actually threw a combination startup funeral and birthday party at the time, which was broadly startup-themed. I gave people fake SAFE notes for bringing drinks, I did a ritual burning of the Articles of Incorporation, did fire spinning, etc. My favorite bit was where I wrote "Product Market Fit" on a piece of paper and hid it in the house, then told people that whoever found Product Market Fit wins a prize. ## Personal lessons I took away a few key things from this personally. First, is that working on something so large by yourself for years is, surprise, quite challenging. Next time I found a company, I'll do it with trusted partners, so I'm not alone. The second biggest learning was in dealing with the mental health challenges I already had. I have a family history of depression, but was in denial about myself having it. I just assumed that it was burnout, or I was just doing something wrong, and that was the reason I had the problems I did. It was actually when I was building a depression diagnostic questionnaire into Dozy that I realized, oh shit, I have depression. At that point I started seeking treatment, but it took a while to actually achieve remission. My broadest takeaway (and continued direction of growth) is the need for higher and more effective levels of self-regulation and healthy emotional processing. I kept seeking new techniques, new books, new systems, to try to make properly feeling feelings irrelevant, but unfortunately that does not work in the long term. The grindset was not sufficient for me, nor was value alignment. ## Dozy's future The goal with Dozy was, and is, to make effective insomnia treatment accessible and affordable. My new goal with the app is to make it entirely free, and to make it the best free resource that exists on the app stores. This is a work in progress, but I'm optimistic that it will reach people who would otherwise have trouble accessing good medical information about sleep. If this intrigues you, reach out to me! I'm available over email or through Twitter DMs.