Building an AI Engine for Globe Telecom that Predicts Household Wealth Nationwide
TLDRThe speaker discusses a strategic approach to broadband planning in the Philippines, highlighting the challenges of traditional methods that involve extensive physical surveys and the risk of becoming obsolete due to rapid technological changes. They emphasize the importance of using satellite imagery, cellular data, and geospatial databases to create an algorithm that predicts economic classes and accelerates the rollout process. The success of this data-driven method is evident in the high utilization rates, exceeding both domestic and global industry standards. The speaker appreciates the problem-solving focus of their collaboration with thinking machines, which prioritizes understanding the issue before developing solutions.
Takeaways
- 📈 The traditional approach involved planning for broadband infrastructure based on economic segmentation of areas.
- 🔍 They would physically survey areas to determine the appropriate service level (Class A, B, C, etc.).
- ⏱️ The process was time-consuming, potentially taking 6-9 years to cover the entire Philippines, which led to a perpetual catch-up with technology and needs.
- 🚀 The goal was to stay ahead of the curve by using a more efficient method to assess and plan for broadband deployment.
- 🌐 They utilized satellite imagery, cellular usage data, and other geospatial databases to gather information.
- 🔍 The initial algorithm was tested against field surveys, resulting in a high level of correlation.
- 📊 The data-driven approach allowed for a nationwide rollout with greater precision and efficiency.
- 📈 The company's utilization rates exceeded industry standards, both domestically and globally.
- 🎯 The algorithms were designed to consider both socio-economic classes and their locations, leading to higher capacity utilization.
- 🚀 The emphasis was on solving the problem first, rather than starting with a preconceived algorithm.
- 🤝 Collaboration with thinking machines focused on understanding the problem, defining the desired outcomes, and working backward to design the solution.
Q & A
What was the traditional method for planning broadband infrastructure?
-The traditional method involved creating plan constructs and inferring which SKUs were suitable for different economic segments of the country. Physical surveys were conducted to categorize areas into classes A, B, C, etc.
What was the challenge with the traditional approach to broadband planning?
-The challenge was the time-consuming nature of the process, requiring a large workforce and taking six to nine years to cover the entire Philippines, which meant constantly playing catch-up with evolving technologies and changing needs.
How did the speaker's team address the challenge of broadband planning?
-They developed an algorithm using different data points, including satellite imagery, cellular usage data, and other geospatial databases, to come up with conclusions about economic classes and improve the planning process.
What was the initial validation of the new algorithm based on?
-The initial validation was based on comparing the algorithm's conclusions with actual field surveys, which showed a very high level of correlation.
How did the new approach impact the rollout of broadband infrastructure?
-The new approach allowed for a more precise design of the rollout, accelerating the process nationwide and resulting in higher utilization of capacity.
What was the speaker's observation about their utilization rates compared to the industry?
-The speaker observed that their utilization rates tended to exceed those of the industry, both domestically and globally.
What was the key to the success of the new approach, according to the speaker?
-The key was starting with the problem rather than the algorithm, focusing on the specific problem to be solved and the desired outcomes, then working backward to design the approach.
How did the speaker's collaboration with 'thinking machines' contribute to the success?
-The collaboration was valuable because it emphasized starting with the problem and working together to design the approach, which made a significant difference in solving the problem effectively.
What does the speaker mean by 'speed to market'?
-Referring to the ability to quickly bring broadband services to market, the speaker's approach allowed for not just speed but also higher utilizations of capacity, which was a concrete and positive outcome.
How did the speaker's team ensure that their broadband planning was adaptable to changing technologies and needs?
-By using a data-driven algorithm that could quickly analyze and categorize areas, they were able to stay ahead of the curve and adapt to changing technologies and evolving needs more effectively than traditional methods.
What was the role of satellite imagery and geospatial databases in the new broadband planning approach?
-Satellite imagery and geospatial databases were used as data points to supplement the team's own cellular usage data, allowing for a more comprehensive and accurate analysis of economic classes and locations for broadband infrastructure planning.
Outlines
📈 Efficient Data-Driven Planning for Broadband Deployment
The paragraph discusses the traditional method of planning broadband infrastructure, which involved physical surveys and categorizing areas by economic class. The challenge was the time-consuming process and the rapid evolution of technology, which made it difficult to keep up. The solution involved using satellite imagery, cellular usage data, and geospatial databases to create an algorithm that could predict economic classes and accelerate the rollout process. The approach was validated against field surveys, showing a high level of correlation, and led to a nationwide rollout with greater precision and higher utilization of capacity. The speaker emphasizes the importance of starting with the problem and working backward to design the solution, which has resulted in a successful and concrete outcome.
Mindmap
Keywords
💡Broadband
💡Plan Constructs
💡SKUs (Stock Keeping Units)
💡Economic Segments
💡Physical Survey
💡Satellite Imagery
💡Geospatial Databases
💡Algorithm
💡Correlation
💡Utilization
💡Socio-Economic Classes
💡Speed to Market
Highlights
Traditional method involved planning constructs for broadband and inferring appropriate SKUs for different economic segments.
Physical surveys were conducted to classify areas into different economic classes such as Class A, B, and Upper C.
The challenge was to scope the entire Philippines with a large team and a long timeline, which would result in perpetually outdated technology and needs assessment.
The goal was to stay ahead of the curve by using different data points and developing a sensible algorithm.
Satellite imagery, cellular usage data, and other geospatial databases were utilized to create a comprehensive dataset.
The first effort involved comparing the algorithm's conclusions with actual field surveys, resulting in a high level of correlation.
The approach was then scaled nationwide to accelerate the rollout, leveraging the greater precision from the assessments.
The company's utilization exceeded industry standards both domestically and globally.
The algorithms developed considered the use cases for different socio-economic classes and their locations, leading to higher utilization of capacity.
The speaker emphasizes the importance of starting with the problem rather than the algorithm to achieve a successful outcome.
The collaboration with thinking machines began with defining the problem and desired outcomes, then working backward to design the approach.
The focus was on solving a particular problem effectively, rather than just obtaining insights.
The process of using data-driven approaches has resulted in concrete and positive outcomes for the company.
The speaker appreciates the problem-first approach taken by thinking machines, which has made a significant difference in their work.