And1Analytics → Monster Roster / Consumer product
The prediction worked. The product needed to change.
I recognized that distributing an identical lineup could limit value when customers entered the same contest. I led the move toward an algorithm-guided core with meaningful user choice, then used observed behavior to shape personal recommendation tracking.
Everyone’s roster.
Your decisions.
One good answer could become a shared limitation.
And1Analytics began in 2015 as a paid recommendation service for daily fantasy sports. Customers assembled a roster of real players for a contest; the players’ performances determined the roster’s score. We helped with the research by selling an algorithm-generated lineup through social media and a website.
The distribution model was simple: multiple customers received the same lineup. In an early tournament, customers entered that recommendation in the same contest and shared the top prize. They were pleased with the result. I saw a product problem inside the success: a strong prediction could still give customers an identical outcome when they were competing with one another.
The issue was specific to how the recommendation met the contest. It did not mean every new subscriber reduced every other customer’s value. It meant the product needed to consider what happened when many customers used the same answer in the same setting.
Decision 01 — Design around the recommendation
Keep the guidance. Give the customer real choices.
I led the move toward an Interactive Lineup Optimizer: a recommended core combined with model-selected options for the remaining positions. Customers could participate in the final roster without starting from a blank page or doing all the underlying research themselves.
The important interaction was bounded choice. Player-selection cards, relevant detail, and a final roster view connected the analytical recommendation to the customer’s own decisions. I owned the product direction and experience design; engineers and quantitative collaborators implemented the software and recommendation logic. The later product also benefited from support through the Philadelphia 76ers Innovation Lab.
The tradeoff was more interaction and implementation complexity in exchange for meaningful participation. The guided approach still allowed overlap, including shared core players. Differentiation was the design objective, not a guarantee that no two users would create the same lineup.
The resulting optimizer moved the venture beyond selling one identical answer. Product artifacts and contemporary coverage support that evolution. I do not attach a retention lift or duplicate-lineup reduction to it: those measurements are not available. The product decision stands on the problem we identified and the experience we built in response.
Decision 02 — Follow the behavior beyond the feature request
A filter request revealed a personal-history task.
As the venture evolved into Monster Roster and expanded its recommendation products, customers asked for filters such as team, rating, and date. We shipped them. I then noticed people repeatedly filtering recommendations from previous days. They were using a selection feature to reconstruct something over time.
I interpreted that behavior as a need for a personal view of selected recommendations and their outcomes. I designed tracking that let users choose recommendations to follow and inspect their historical results over a chosen period. The team implemented the feature.
Its scope mattered. The tracker followed recommendations selected inside Monster Roster. It did not independently reconcile sportsbook accounts or prove which wagers a customer placed. A personal recommendation history solved a narrower task with the information our product actually held.
That experience changed how I read feature requests. A request tells me where a user encounters friction; repeated behavior can reveal the job underneath it. Shipping the requested control is a starting point for learning how people use it.
Different disciplines needed different feedback.
I managed a small team spanning engineering, data science, and design. Early on, I treated their work too similarly. I changed the structure so design and customer-experience work could use mockups and beta feedback, while engineering and data work had a distinct iteration and testing loop.
My responsibility was connecting those efforts around the product. Managing specialists meant understanding what each needed to make progress, as well as setting priorities. That learning sits alongside the consumer-product work: the experience customers saw depended on how clearly I organized the work behind it.
Real paid demand, across one venture’s evolution.
Across And1Analytics and Monster Roster, the venture acquired more than 1,000 paying customers and was the first company selected for the Philadelphia 76ers Innovation Lab. We developed the original paid recommendation service into subscription web and mobile products. Those are milestones of the whole venture, rather than outcomes attributable to one feature or a count of concurrent subscriptions.
My titles reflect successive phases of that same business: Co-Founder & President of And1Analytics from February 2015 through June 2016, then Founder & CEO of Monster Roster from July 2016 through December 2019. I returned to Georgetown in 2019 and completed my BA in Psychology in 2022.
Evaluate what happens around the model.
The lasting lesson is to evaluate the whole customer outcome. A recommendation can be strong while its distribution limits value. A useful filter can expose a deeper need. My role was to notice those gaps, decide what the product should do about them, and lead the team through the change. That perspective continues to shape how I build products around predictive systems and AI.