Development and Enhancement of a Marketing Mix Modeling Platform

About Client

1749 is a marketing analytics consultancy that empowers businesses to adopt a data-driven approach to marketing. They utilize a suite of innovative, often open-source tools to analyze and interpret marketing-related data, helping their customers integrate those solutions into routine workflows.

One of their key areas of focus is development, maintenance, and improvement of machine learning-based models — known as marketing mix modeling (MMM) — designed to facilitate in-house marketing effectiveness measurement and future performance forecasts.

The Challenge

The client aims to build and maintain a full-fledged platform around open-source marketing mix models, focusing on maximizing their efficiency and simplicity of use.

Proceeding from the prototype of the platform made by our client via ChatGPT, our main challenge encompassed further development, enhancement, and extension of an MMM platform functionality.

Since the very beginning, our team prioritized separation of the existing monolith platform into independent frontend, backend, and machine learning (ML) parts. We also aimed to apply the best engineering and design practices and implement the recent technologies to further build and modernize the platform.

Dev Process

Robust System Architecture

As defined at the initial project planning stage, the separation of the monolith platform into independent system parts is the key to robust system architecture. Our team therefore created separate frontend, backend, and ML parts to simplify further development, enhancement, support, and modernization of the platform.

Mockups

Based on the best practices in UI/UX design, our team prepared mockups for every screen to incorporate the new features and enhance the existing interface for a more streamlined look.

Functionality Development

Together with the client, we defined and prioritized the main platform features to be added or enhanced. Overall, we contributed to the enhancement of the model management system by implementing intuitive model organization, added the enhanced visualization tools, developed an asynchronous model fitting, and enabled a simultaneous model processing.

Production

Our team selected the right instances to ensure the best possible performance of ML models. Also, we set up CI/CD pipelines to secure the deployments and adopted the blue-green strategy.

Design Process

Mockups Creation

Since this project prioritizes functionality over complex design decisions, our mission was to revamp the existing interface elements bringing simplicity and ease of use. Proceeding from our expertise in UI/UX design and understanding of user-centric design principles, we created mockups for every screen. To further implement those mockups, we took an open-source React component library with ready-to-use material design components that allowed us to simplify and speed up the process of MVP development.

User Feedback Collection

Following the creation of the first version of the platform, we conducted a round of usability testing to learn how end-users interact with our interface in real time. This process helped us identify the areas of improvement to better tailor our UI/UX to users’ demand.

Changes Implementation

Our team introduced the changes to the UI/UX to make sure the product resonates well with end-users.

Outcome

Our iterative approach — based on incorporating users’ and client’s feedback at each stage of platform development — ensured that the final version of the product not only met technical specifications but also aligned well with users’ demand.

We tailored the system architecture, particularly backend, to machine learning workflows that significantly improved platform performance. Now, end-users can easily access, store, manage, analyze, and fit multiple models while performing other tasks within the platform simultaneously.

Guided by the principles of simplicity and ease of use, our team enhanced the platform’s UI/UX making even the advanced features accessible to users of all experience levels.

Additionally, we added a wide variety of output charts with consistent colors and clean designs that enable users to better understand insights and tailor their visualization options to specific data types.