Transforming Telco Network Management Decisions with Location Allocation

Managing a modern telecom network requires balancing cost, coverage, and operational efficiency. Every network node—a set of cell towers—represents demand that must be effectively served by strategically placed facilities.

In this tutorial, we’ll explore how network planners can determine the optimal locations for maintenance hubs or support facilities, ensuring that each area of the network is monitored and maintained efficiently through Location Allocation, a toolkit available in the Territory Planning Extension Package. With this approach, telecom operators can compare different strategies—from minimizing operational costs to maximizing coverage—making data-driven decisions that improve performance, reduce downtime, and enhance service quality across the network.

Before we dive into the use cases, let’s take a closer look at what Location Allocation is and how it works. If you’re already familiar with this type of optimization problem, you can skip this section and move straight to the practical example!

Understanding Location Allocation

At its core, Location Allocation is like solving a puzzle: you want to place facilities (such as warehouses, stores, service centers, or hospitals) in the best spots while making sure customers, delivery areas, or network points they serve (namely, demand points) are properly covered based on:

1. Objectives (what you’re trying to achieve):

   • Minimize total or maximum costs: This strategy focuses on reducing expenses. The algorithm chooses facility locations so that the total cost—or the highest individual cost—of serving all demand points is as low as possible. Costs could represent travel distance, delivery time, or operational expenses.

   • Maximize coverage: This strategy aims to serve as many demand points as possible. Facilities are placed to cover the largest amount of demand within a certain distance, ensuring that the majority of demand is efficiently served.

1. Constraints (rules you must follow):

   • Capacity limits: Each facility can only handle a certain amount of demand. For example, a warehouse can only store so many products, or a service center can only handle a certain number of clients.

   • Number of facilities: You might be limited by space or resources, so only a specific number of facilities can be opened.

   • Budget constraints: The total cost of opening facilities may be limited by a fixed budget. Say, even if opening more facilities could improve service coverage, the combined fixed opening costs cannot exceed the allocated budget.

   • Forbidden/required assignments: Some facilities may not be able to serve certain demand points due to restrictions like regulations, geography, or compatibility. For instance, certain locations might be off-limits for a facility, or some demand points might require special handling that only specific facilities can provide. Conversely, some facilities may be required to serve particular demand points, making those assignments mandatory.

It’s important to note that Location Allocation is a very broad topic used in many industries—from retail and logistics to healthcare and telecom networks. There are countless ways to define objectives, set rules, and model demand, depending on the problem at hand.

The Location Allocation component covers the most general and widely applicable use cases: placing facilities to either maximize coverage or minimize costs while ensuring demand points are properly served. These core strategies form the foundation of most Location Allocation problems and provide a solid starting point for understanding how this powerful tool can support smart, data-driven decision-making. If you are interested in custom-specific modelling, please reach out to the CARTO team to discuss tailored solutions for your unique business needs!

Location Allocation in action!

Now that we’ve covered the basics of Location Allocation, let’s bring the concept to life with a telecom network example from Connecticut. In this case, demand is modeled using H3 cells, where the density of cell towers within each hexagon reflects network needs. While the network already has existing facilities, we are exploring potential new sites, both represented as simulated management sites.

We will explore two optimization strategies used for different purposes:

1. Selecting facilities for emergency response, for which we aim to maximize network coverage so that whenever an emergency occurs (i.e. outages, equipment failures, or natural disaster impacts), the nearest facility can quickly respond and restore service. Our objective will be to open 8 facilities from a set of candidates that can act as Rapid Response Hubs. Access the template here!

2. Opening facilities for periodic maintenance, for which we aim to minimize total operational costs for ongoing inspections and servicing, respecting resource capacities, and ensuring that routine maintenance is delivered cost-effectively. Our goal will be to expand our existing facilities by adding one selected site per county in Connecticut to serve rising network demand. The template is available here!

Together, these approaches illustrate how Location Allocation can adapt to different business priorities, balancing efficiency with resilience depending on the needs.

Setting up your workflow

1. Sign in to CARTO at app.carto.com

2. Head to the Workflows tab and click on Create new workflow.

3. Choose the CARTO Data Warehouse connection or any connection to your Google BigQuery project.

4. Install the Territory Planning extension package if you do not have it already.

To perform location allocation using CARTO, we must prepare the necessary data using the available components to do so: Facilities Preparation, Demand Points Preparation, Cost Matrix Preparation, and Constraints Definition. In the next sections we will see how to use each of these.

Preparing facilities data

As a first step, we’ll prepare the data for our facilities. While not all of this information will be used in both use cases, having everything ready upfront simplifies the workflow and reduces computational effort.

As mentioned, we already have some operating (required) facilities that we may want to keep, but we also have a set of candidate facilities that we aim to open to optimize a specific objective. First, use the Get Table by Name component to load the following sources:

• Candidate facilities: cartobq.docs.connecticut_candidate_facilities

• Required facilities: cartobq.docs.connecticut_required_facilities

Then, load the Facilities Preparation component and connect the input sources. Now, we need to indicate what information we have available. For our two use cases, we may need to consider some of the following, so we will select them all:

• Use required facilities: this will ensure that our facilities data contains both candidate and required facilities.

• Use facility groups: this will ensure that our facilities data has a group ID assigned to them. In this case, we will consider counties.

• Use facility maximum capacities: this will ensure that our facilities data contains the maximum demand each facility can serve.

• Use facility cost of opening: this will ensure that our facilities data contains the fixed costs of opening a facility, which will also influence the final decision.

Preparing demand points data

To simplify the workflow, we have already aggregated the cell tower data into H3 cells by counting the number of 3G, 4G and 5G antennas within each cell, from OpenCellId (available in the “demo data” cell_towers_worldwide dataset of your CARTO Data Warehouse connection). So, you can directly use the Get Table by Name component to load the following source: cartobq.docs.connecticut_demand_points.

Then, use the H3 Center component to get reference coordinates for each demand point in our network (each H3 cell containing telco antennas). Drag and drop the Demand Points Preparation component and select the Use demand option. This will ensure that our demand points data contains the number of antennas that need to be served per region.

Last, use the Join component to recover the county information of each demand point, which will be useful for the second use case.

Computing costs

As a next step, we need to compute the costs of serving demand. For now, we’ll base this on the distance between each facility and demand point. To compute the pair-wise distances in kilometers, we will use the Cross Join and Distance (single table) components as a proxy. As an alternative, you could use the Create Routing Matrix component for more specific cost computations. Then, connect the Cost Matrix Preparation component.

Use case 1: Selecting facilities for emergency response

With this, we are ready to run our first Location Allocation analysis. Simply load the Location Allocation component and specify the following settings:

• Use ‘Maximize coverage’ as the optimization strategy with a coverage radius of 30 km. This is our estimated distance that can be covered in a timely manner from the Rapid Response Hubs in case of emergency.

• Consider that we can only allocate a budget of $20M to open at most 8 new facilities.

• Use demand to prioritize covering H3 cells with largest amounts of cell towers.

The selected facilities, seen in the map below, are optimally distributed to serve nearly all demand points, with each facility covering its surrounding area. The visualization highlights how this approach ensures broad geographic reach and balanced service across the state, minimizing uncovered demand. In fact, with the chosen set of facilities, we are able to cover more than 99% of the demand efficiently, ensuring readiness in the event of an emergency.

Use case 2: Opening facilities for periodic maintenance

In this second use case, facilities must meet demand within capacity limits while serving only demand points in the same county, reflecting regulatory, logistical, or policy requirements for localized service. For example, when certain permits or licenses might restrict a facility’s operations to its own county, or there may be local policies favoring in-county service.

Define additional constraints

To add such constraints, we need to use the Constraints Definition component. We will select the facility-demand point pairs that belong to different counties by typing the following in the Where component:

• group_id != county_name_joined_joined

Next, connect the Constraints Definition component through the `Forbidden` relationships input channel and select Consider forbidden facility-demand point pairs.

Lastly, connect it also to the Location Allocation component and specify the following requirements:

• Use ‘Minimize total cost’ as the optimization strategy, including required facilities. This ensures that we reallocate assignments in current facilities while opening new facilities that help maintain all cell towers to cover increasing demand.

• Include the costs associated with opening candidate facilities to ensure minimum total spending.

• Limit the number of facilities to open per country (group) to 4. Since 3 facilities are already operating in each county, at most 1 additional facility can be opened.

• Include capacity constraints to ensure that each open facility meets demand without exceeding its maximum capacity limits.

• Select the ‘Use required/forbidden assignments‘ option to consider the compatibility constraints previously defined.

In this case, facilities are placed closer to dense demand clusters (shown in pink in the map below), reducing overall travel distances between demand points and their assigned facilities. Required facilities remain active, while additional candidate sites are selected to balance the workload without exceeding capacity limits. Actually, 90% of the capacity is being utilized.

Together, these two applications of Location Allocation—minimizing costs for periodic maintenance and maximizing coverage for emergency response—help network operators balance efficiency with resilience, keeping their infrastructure reliable under both normal and unexpected conditions.