Header image

Accelerated Customer Data Cleanup and Network Migration using AI for Globe Telecom

March 17, 2022 case-study geospatial customer-intelligence geospatial-analytics bigquery machine-learning telecom telecommunications

Thinking Machines helped Globe automatically validate and clean up its massive customer database by leveraging geospatial intelligence and natural language processing to manage their geodata quickly and easily.

Through the work of Thinking Machines, Globe not only validated its massive customer database, but also gained a geospatial web app to visualize their customer data. All this was built in under 2 months.

The on-ground validation returned 70% locatability—a 6x increase in the number of locatable customers, matched with available network facilities, accelerating network migration efforts.
The Problem

Globe Telecom is the Philippines’ largest telecommunications company. The telecommunications industry is fast-moving and network technology is constantly evolving from xDSL to Fibre and from 3G and 4G to 5G.

The Globe Broadband team was confronted with legacy wireless and wireline broadband subscribers that Globe aimed to upgrade to a Fiber connection. Specifically, they wanted to be able to prioritize customer migration based on the availability of facilities around the location of these customers and use this information to further inform additional network deployments.

During the assessment, some addresses were of poor quality. This made it challenging to migrate the customer base at a speed that the business wants. And a lot of time was spent on validating the actual location with facility availability.

The Broadband team was looking for a solution that could help them clean up, process, and match addresses with network facilities to make it easier to map out the desired migration strategy.

The Solution

Globe Broadband turned to Thinking Machines to develop a custom Broadband Migration Tool to help optimize their migration and prioritization activities. Specifically, the tool needed to allow Globe’s Broadband Team to process and improve the quality of addresses automatically at scale so that they can focus on the prioritization of sites and execution.

The main aim of the project was to be able to beat Globe’s current matching rate of 11% within 50m of the actual location.

Having accurate location data is essential to prioritize and plan the wireless broadband migration strategy.

Reliable analysis starts with accurate geocoding. To ensure high-quality geocoding results, we:

  1. Pre-processed addresses to ensure proper naming convention and information completeness to optimize geocoding results,
  2. Used subscribers’ tower latching information to ensure reliable geocoding results and provide confidence rankings to support prioritization, and
  3. Displayed the information in an interactive geospatial dashboard for easier visualization and analysis of the data.

Step 1: Pre-processing addresses using LinkSight

The initial step was to format the addresses to optimize the geocoding results. For this, we used LinkSight 2.0, a technology tool developed by Thinking Machines for matching addresses to their correct administrative boundaries. This is heavily inspired by LinkSight 1.0, a web app we created to clean up messy location datasets quickly and easily.

Step 2: Spatial Analysis using BigQuery

After geocoding, we used location-based rules and filters to ensure that the returned coordinates of the subscriber’s address are in the same area as the cabinets. The team was able to scale out the process to ~400,000 addresses by using BigQuery spatial analytic function.

Step 3: Develop an Interactive Geospatial Dashboard to allow Globe analysts to easily analyze and extract insights from the data.

The final step was to create an interactive Geospatial Dashboard which displays the processed location data. The dashboard was built with intuitive interactions which allow Globe’s analysts to search, filter, and zoom in to areas of interest. On top of these features, the dashboard allows the Globe team to update the location and locatability score of each record against the actual results of their on-ground validation.


The following tools were leveraged to maximize the provided information and provide high confidence coordinates from addresses.

A 6x increase in customer locatability and network matching has accelerated network migration efforts

From 11% matching to now 70% based on on-ground validation. This is a 6x increase from the initial number to speed up customer migration.

Through a machine-learning-driven approach, algorithmic data cleanup, and geocoding, more customers are now matched with available network facilities to serve these customers.

The significant improvement of Globe’s customer data will help in accelerating their migration efforts and decommissioning legacy technology faster; preventing churn by rebuilding customer relationships.


When are you most in danger on the road?

How likely is it that if you get in an accident on your daily commute, it will result in someone getting hurt or killed?

Thinking Machines is now a partner of CARTO

“CARTO and Thinking Machines both believe in the power of geospatial insights for better decisions in a complex world," says Stephanie Sy, our CEO. "We’re very excited to bring better spatial data analytics tools to everyone.”

How We Simplified Time-Tracking with Slack, BigQuery, and a Power BI dashboard

Also known as: time-tracking for lazy people who hate forms.