Formula is a bioinformatics startup that uses data to tailor nootropic brain supplements to each users’ unique brain chemistry and daily activities.
Formula conducted the largest trial ever done on the efficacy of cognitive enhancing supplements (aka. nootropics) in order to train a recommendation engine that accurately predicts the best nootropics for each individual user. They tested 100+ ingredients on 2500+ participants, gathering over three million data points.
The Formula team wanted to see if they removed the controls and stringent oversight of the widely accepted clinical trial model, increased participant numbers by orders of magnitude, would they be able to build a data set large enough to train an algorithm to accurately predict user-specific supplement efficacy.
Specifically, Formula asked:
Formula conducted a semi-blind four-way crossover trial where participants were given pre-measured, daily pill packs for thirty days. They were told the ingredients inside the box, but they did not know which ingredients they were taking each day. Formula sent two SMS texts each day each asking eight questions to collect data.
After a 24 month period, 2500 people had gone through the trial. 127 ingredients had been tested. Three million data points had been collected. For those of you familiar with large datasets, three million may not be that much. However, the largest trials done on nootropics prior to Formula included between 10 and 100 participants.
What they learned was that the most popular nootropic blends on the market work for only 48% of the population. After testing over two thousand combinations (aka stacks), there were 10 that worked for 85% of the population, each with their own optimal use case.
They also showed that Phenylpiracetam, out of all the 19 derivatives of Racetams, had the highest efficacy and highest reorder velocity. Second highest efficacy was Noopept.
In terms of adverse reactions, they found that Aniracetam and Noopept both have a side effect occurrence lower than 5% of the population.
While there’s no clear winning combination, as even genetically similar people like siblings can respond differently to different nootropics, there are specific stacks effective for particular mental tasks. For example, Aniracetam and Pramiracetam mixed are optimal for creative tasks and social fluidity. It scored the highest in creative writing tasks and helped people who normally struggle in social situations to lubricate their social interactions.
They wouldn’t have been able to build a bootstrapped business with an unlimited amount of skews i.e. if each combination only works with a few people. However, being able to identify the top ten that worked for 85% of the population allowed them to create a pared down system that we could scale quickly without large amounts of cash being poured into the business, while still maintaining really high efficacy. Which is the real reason that people come back. The key to not only finding the best nutrients and exact dosages, but also finding what they needed to focus on to build a business to scale quickly.
If they had to make a new supplement combination for each user by hand, they wouldn’t be able to handle more than a hundred customers a month. Understanding the core ten that have 85% to work allows us to work with thousands and still provide value.
With improvements in price and convenience of at-home genetic testing and wearables becoming more ubiquitous, they’ll use this data alongside their own user observations to expand into growing categories such as sleep, probiotics and daily vitamins. Eventually, they’ll be able create a 3-D printed, auto optimizing daily supplement that knows all of your genetic dispositions, nutrient deficiencies, activities and can adjust on a weekly basis.
There are years and years with geniuses upon geniuses that worked to develop the scientific method and clinical trial methods. For academics and scientists, the goal was always to create insights that are published and reviewed by peers.
Formula took a similar process and changed its goal. Rather than focus on a publishable study, the goal was more altruistic, it was less politics, it was very simply to train a recommender engine. To collect enough data to personalize nootropics.
Try this for your project: Take a rigorous process and redefine why you’re doing it. Loosen regulations that over-constrain your outcomes and leverage the open-ness to collect more information than you normally could have.