Using Machine Learning to Rapidly Discover & Scale Profitable Opportunities
The Business Challenge
A mid-sized email marketing client struggled with a tangled process for identifying customers to target. What began as simple rules around recency, source, and offer had evolved into a dense web of conditional statements. Any rule change risked unintended consequences, making the process slow, opaque, and impossible to measure for value. This complexity prevented the client from pursuing new opportunities and optimizing campaigns.
The Komodo Solution
Komodo partnered with the client to clarify objectives: they needed a system to identify the most profitable email addresses and maximize campaign returns. Their existing process could only handle a few data types and one email platform, yet the company had years of rich engagement data waiting to be used.
The goal was to build a model that ingested large volumes of data and generated daily recommendation scores and user status details for 2 million active users. These outputs could be consumed by any downstream platform, laying the foundation for data-driven decision making.
Within three months, Komodo developed a prototype model that processed far more data and produced usable outputs for the client’s current platform. Early versions emphasized rapid iteration to validate prediction quality, keeping infrastructure lean until the right model and data scale were identified. Once validated, data engineering and DevOps scaled the model with confidence.
Figure 1: Early versions of the model emphasized rapid iteration to assess quality of predictions. Infrastructure was simple and cheap. Once we identified the right model to build and the right magnitude of data to process, data engineering and devops were able to scale a model with confidence.
The model successfully identified users who engaged at 1.5x the average rate. But deeper analysis revealed a challenge: the most profitable users were already being targeted heavily. Unlocking this insight required a fundamental shift in mailing strategy and infrastructure investment, delaying ROI.
Recognizing this, Komodo pivoted. Instead of focusing on heavily targeted active users, we turned our model toward inactive users—98% of the client’s database. This represented a massive untapped opportunity. We re-engineered pipelines to score not only 2 million active users but also 100 million inactive ones.
To enrich this dataset, we evaluated more than a dozen third-party vendors using a profitability framework built for the client. By creating a portfolio of vendors, we secured competitive pricing and diversified data sources.
With the augmented data, the refined model achieved excellent predictive power. Evaluation analyses showed which inputs contributed most to engagement, guiding smarter business and infrastructure decisions—such as whether to invest in tracking new data types.
Figure 2: The model in its maturity had excellent predictive power, and evaluation analyses allowed the client to see what inputs were contributing the most to higher user engagement. This allowed the client to make better business and data infrastructure decisions, such as whether or not to invest in tracking new data types.
We tested small batches of previously excluded users and discovered tens of millions who were just as profitable as the current active pool. Suddenly, the client had a vast new audience they could target using existing platforms, without waiting for a costly infrastructure overhaul. With proven results, Komodo scaled the model into a production-ready system.
The Results
In just six months, Komodo delivered a production system that transformed the client’s targeting strategy. The newly identified users generated $115,000 in additional revenue in the first month alone, with recurring gains continuing well beyond the project’s conclusion.
Importantly, the solution was designed as “set it and forget it,” delivering ongoing value with minimal maintenance. A thorough handoff ensured the client’s internal team could extend the model for future projects and channels.
Komodo’s predictive modeling not only created a profitable new revenue stream but also sparked innovation, proving the power of machine learning to unlock opportunities hidden in plain sight.
