Comfama: From Operational Efficiency to Enrollment Growth at Scale

Among its many services, Comfama runs a continuing education program called Educación para la Vida, offering short courses in sports, cooking, arts, technology, and personal development across Antioquia, Colombia.

With over 800 course types, 450 instructors, and more than 13,000 individual course events per quarter across dozens of physical locations and a virtual platform, the operation is large by any measure. The growth challenge was proportionally complex.

When they came to us, enrollment was the core problem. Comfama had demand, instructors, and facilities, but no reliable system for predicting which courses would fill and which would not, no scalable way to communicate offerings to prospective students, and no shared language around growth across the people managing their physical locations.

• Instructors were promoting courses by sending screenshots of Excel files over WhatsApp
• Leaders at each location had a Power BI dashboard they did not understand or use
• No demand prediction — course offerings were based on intuition, not data
• No shared growth language — the organization was thinking about efficiency, nobody was thinking about growth

We started by doing something most partners skip: we spent the first phase learning the business in depth, talking to the people running each physical location, understanding how enrollment decisions got made, and mapping where the real constraints were. That diagnosis shaped everything we built after it.

One of the clearest opportunities we found was the instructor network. Comfama's 450 instructors already had direct relationships with their students, and many were already trying to promote courses informally. The problem was they had nothing useful to share.

We built a system that takes Comfama's course catalog, with all its codes, schedules, locations, and photos from their professional image library, and generates branded visual assets for each course automatically. An instructor can share a polished, course-specific card directly to WhatsApp with a QR code linking to enrollment, instead of forwarding a spreadsheet screenshot.

The system uses Cloudinary to handle image manipulation at scale, applying AI-assisted cropping and composition logic so faces stay visible and text placement adjusts per image. What used to require a designer working manually now runs as a batch process across thousands of courses at once.

On the data side, we worked with their team to build a machine learning model that uses three to four years of historical enrollment data to predict demand for upcoming courses by physical location and time of year. Leaders at each location now use that model to make decisions about which courses to offer each quarter, matching supply to predicted demand instead of relying on intuition.

We also rebuilt how performance data reaches the people who need to act on it, replacing dashboards nobody understood with reporting designed around how location leaders actually think about their work.

The results reflect both the operational changes and the shift in how the organization approaches growth:

1. Spending the first phase learning the business — not building — shaped everything that came after and avoided solving the wrong problems.
2. Instructors were already trying to promote courses informally. Giving them professional, shareable assets turned an existing behavior into a scalable channel.
3. ML-based demand prediction replaced intuition for quarterly course planning, aligning supply with actual demand by location and time of year.
4. Replacing dashboards nobody understood with reporting designed around how leaders actually work changed the organization's relationship with growth data.