While it may seem an impossible task to take on huge platforms that have network effect flywheels on their side, our approach had an advantage. A platform like TikTok has massive amounts of content, but it needs to have a lot of content, because only a small portion of it will ever be any good, or relevant to the audience. TikTok relies on massive amounts of data for their algorithm to function.
By comparison, if we could ensure that all our content would be informative and entertaining, as well as meeting precise format requirements that TikTok cannot similarly guarantee, we wouldn't need all the rest.
Another benefit of having control over the content is avoiding a race to the bottom. Decentralized creators will tend toward trying to compete for sales, pitching the same product. This will result in increasingly loud and obnoxious videos, as can be seen with the ubiquitous "influencer voice".
While the long-term plan was to open the platform to the public to create videos, the founder effect is very important in social media. This means that the initial culture of the first creators will define norms on the platform for its entire lifecycle. This meant we would want to encourage collaboration between the creators rather than pure competition.
We could accomplish this by going beyond incentivizing only direct sales. Rewarding creators for making contributions to the process of selling products would encourage a diversity of content types, less redundancy, and a more enjoyable watching experience.
Gamefied metrics that encourage experimentation
We created a simple dashboard for the creators which was designed to avoid the optimization trap. So early on, we did not know what the ideal content format would be, Instead, we wanted to encourage creators to experiment within the constraints we had already provided.
We only provided three explicit metrics: views, orders correlated with their videos, and followers. We did not let the creator know exactly how many products their video had sold directly.
The chart in the bottom left is a T-SNE, which compares the various creators to each other by how similar their content and audience behavior are. This enables creators to either seek to differentiate or align with other creators without knowing precisely how to do so.
The bars at right track a creator's unique strong suits, showing where they do better than the average. If a creator is particularly adapt at getting users to add products to the wishlist, this visual will reflect that. Because everyone shares in the commissions, it means people will want to find what they are good at rather than trying to only optimize for sales.