Speakers
Description
Implementation of state-of-the-art Marketing Mix Modeling (MMM) often requires a precarious balance between domain expertise and advanced programming skills. Google’s open-source library, Meridian, provides a robust Bayesian framework for media measurement, yet its reliance on complex Python workflows limits its accessibility to non-programming market analysts.
In this talk, we present Citrus Predict, a user interface designed to bridge this technical divide. Meridian involves high-dimensional input parameters and complex Hamiltonian Monte Carlo sampling that can be daunting for the uninitiated. Our tool abstracts these complexities through an intuitive UI, allowing users to upload data, set priors through visual guides, and interpret model outputs without writing a single line of code.
We will discuss the architectural challenges of wrapping a heavy-duty TensorFlow-based library for the web and how lowering the barrier to entry fosters better collaboration between data scientists and business stakeholders. By making research-grade software accessible, we empower a broader user base to leverage rigorous statistical methods in their daily decision-making.