In Q4 2022, we saw an opportunity: generative AI was about to explode, but the tools were trapped in Discord servers and command lines.
Our hypothesis: Consumers wanted to create personalized AI content with their friends, not just generate random images. Users didn't want another AI image generator. They wanted to see themselves in impossible scenarios. This insight came from watching early users spend 80% of their time uploading selfies rather than using text prompts.
This led to our product thesis: Personalization + Social = Retention.
My Role: Founding Product Manager
As the founding product manager of the team, I was responsible for bringing the entire experience to life. From a napkin sketch in the board room to a production ready mobile app, I was involved at every step of the product lifecycle.
Decision 1: Daily Avatar Drops
Hypothesis
Scarcity and anticipation would drive daily active use.
What we built
Daily themed avatar drops with push notifications.
Result
90% day-1 retention, but only 30% day-7 retention.
Learning
We created habit-forming mechanics without solving for long-term value. Users collected avatars for two weeks, then churned.
Decision 2: Social Feed vs. Private Creation
Hypothesis
Users would want to share AI creations publicly.
What we built
Instagram-style feed with creator profiles.
Result
Less than 5% of users ever posted publicly. 89% of generations were never shared.
Learning
We underestimated the uncanny valley effect. Users loved creating weird AI content but were embarrassed to share it publicly.
The Metric That Killed Us
400,000 downloads → 15,000 DAU = 3.75% stickiness
40,000 generations/day ÷ 15,000 DAU = 2.67 generations per user
Users were trying the product once daily and leaving.
What I Would Do Differently
Solve for the "Day 8 Problem" First
We optimized for activation (first generation) and day-1 retention. We should have focused on why users stopped finding value after the novelty wore off.
Build for Private Sharing, Not Public Broadcasting
Our data showed users wanted to share with close friends (DM, iMessage) not broadcast publicly. We built Instagram when we should have built something closer to BeReal.
The Truth
PlaiDay was fun for a few weeks, then forgotten. The 15,000 DAU who stayed were hobbyists who would have used any AI tool. But we were 6 months ahead of the market with our powerful AI engine, which eventually drove us to pivot to the team’s next venture:
Salt AI.
PlaiDay Team
Aber Whitcomb, Chris DeWolfe, Jim Benedetto, Charlie Basil, Scott Baggett, Alex Duffy, Ryan Wang, Mat Blysz