What Is a Network Analysis? Routing and Service Areas Explained
A road is not just a line on a map. It has direction, speed limits, turn restrictions, and connections to other roads. When you model roads — or pipelines, rivers, utility lines, or any connected system — as a network, you can start asking questions that geometry alone cannot answer.
Which route gets a delivery truck from warehouse to customer in the least time? Which hospital can an ambulance reach from any given street corner in under eight minutes? How many households fall within a 15-minute drive of a proposed clinic?
Network analysis answers these questions. It treats infrastructure as a graph of nodes and edges, applies cost functions to movement through that graph, and finds optimal paths or reachable zones. This article explains the core concepts: how networks are modeled, how routing works, and what service areas tell you.
The Network Model
A network is a graph. It has two components:
Nodes (also called vertices or junctions) represent points where edges meet. In a road network, nodes are intersections. In a utility network, they are valves, pumps, or substations.
Edges (also called links or segments) represent connections between nodes. In a road network, edges are road segments between intersections. Each edge has attributes that define the cost of traversing it.
The cost attribute is what makes a network model useful. Distance is the simplest cost: longer edges are more expensive to traverse. But travel time is usually more meaningful. A 10-kilometer highway at 100 km/h costs less time than a 5-kilometer urban street at 20 km/h. Speed limits, congestion factors, road class, and surface type all feed into a time-cost model.
Other cost types exist depending on the domain. Elevation gain for cycling or hiking networks. Pipeline diameter and pressure for utility networks. Traffic signal delays for pedestrian networks.
Directed vs Undirected Edges
Some edges are bidirectional: you can traverse them in either direction. Others are one-way. A road network is partially directed. Most roads allow travel in both directions, but one-way streets, highway on-ramps, and turn restrictions create directed constraints.
Turn restrictions add another layer. Even if two road segments connect at a node, a left turn from one to the other may be prohibited. Network models that ignore turn restrictions produce routes that are geometrically valid but operationally impossible.
Network Topology
Topology is the set of connectivity rules in the network. Two edges must share a node to be considered connected. Two roads that cross on a map but do not share a node in the data are not connected in the network — they pass over or under each other, or the data has an error.
Clean topology is a prerequisite for correct network analysis. A missing node at an intersection, a road segment that does not snap to its neighbor, or a duplicate node where there should be one are all topology errors that produce incorrect routing results. Before running any network analysis, auditing and cleaning the network topology is essential.
Routing: Finding the Optimal Path
Routing (also called shortest path analysis) finds the lowest-cost path between two or more points in a network.
How the Algorithm Works
The most widely used algorithm for single-source shortest paths is Dijkstra’s algorithm. It works by expanding outward from the origin node, assigning cumulative costs to each node it reaches, and always expanding the lowest-cost unvisited node next.
At each step:
- Start at the origin with cost zero.
- Examine all edges connected to the current node.
- Calculate the cumulative cost to reach each neighbor.
- If the new cost is lower than the previously known cost to that neighbor, update it.
- Mark the current node as visited. Move to the unvisited node with the lowest cumulative cost.
- Repeat until the destination is reached.
The result is the lowest-cost path from origin to destination. Backtracking through predecessor nodes gives you the actual route.
For large networks, Dijkstra alone is too slow. Practical implementations use optimizations: bidirectional search (expanding from both origin and destination simultaneously), A* (using a heuristic to guide expansion toward the destination), and Contraction Hierarchies (preprocessing the network to enable millisecond-level queries on continental road networks). Modern navigation apps use Contraction Hierarchies or similar hierarchical approaches to return routes in real time on networks with tens of millions of edges.
Multi-Stop Routing
Finding the shortest route visiting multiple stops in a fixed order is straightforward: solve a shortest path for each consecutive pair of stops and concatenate the results.
Finding the optimal order to visit multiple stops — the Travelling Salesman Problem — is computationally hard. For small numbers of stops (under 15 or so), exact solutions are feasible. For larger stop counts, heuristics like nearest neighbor, or metaheuristics like simulated annealing or genetic algorithms, find good-enough solutions in practical time.
Most GIS tools expose this as a Vehicle Routing Problem (VRP) solver, which adds real-world constraints: vehicle capacity, time windows at each stop, driver shift limits, multiple vehicles.
Impedance and Barriers
Impedance is the cost function applied to edges. You can apply multiple impedance types to the same network and switch between them depending on the analysis. Minimize distance for a cycling route. Minimize time for an emergency vehicle. Minimize fuel consumption for a logistics fleet.
Barriers block or penalize traversal of specific edges or nodes. A road closure, a flood, a locked gate. Point barriers block a specific node. Line barriers block all edges they cross. Polygon barriers apply an added cost or restriction to all edges within the polygon.
Service Areas: Mapping Reachable Space
A service area is the zone reachable from a facility within a defined cost threshold. If the cost is time, a 10-minute service area shows every location reachable within 10 minutes of travel from the facility.
Service areas answer a different question than routing. Routing asks: what is the path between these two points? Service areas ask: what territory does this facility cover?
How Service Areas Are Computed
Service area computation starts like Dijkstra’s algorithm. Expand outward from the facility through the network, accumulating cost. Stop when the cumulative cost exceeds the threshold.
The result is a set of network edges and partial edges that fall within the cost threshold. These edges are then buffered and dissolved to produce a polygon representing the service area.
The shape of a service area reflects the network structure. On a grid street network, the service area is roughly diamond-shaped (following the Manhattan distance pattern). On a radial road network like many Indian cities, the service area extends along arterials and shrinks between them. In areas with rivers, highways, or other barriers, the service area shows the constraint clearly.
Multiple Threshold Rings
Service areas are often computed for multiple thresholds simultaneously: 5, 10, and 15 minutes from a hospital, for example. The result is a set of concentric rings, each representing the incremental zone reached between thresholds.
These rings are useful for understanding tiered coverage. The innermost ring might receive ambulance response in under 5 minutes (high coverage). The outermost ring receives response in 10 to 15 minutes (lower coverage). Areas outside all rings are uncovered.
Travel Direction
Service areas can be computed in the outbound direction (where can I go from here?) or the inbound direction (what can reach this facility?). For most facilities, the distinction matters.
An ambulance depot computing an outbound service area shows where the ambulance can go. A hospital computing an inbound service area shows where patients can come from within a given drive time. These are the same zone only on an undirected network. On a real road network with one-way streets and turn restrictions, inbound and outbound service areas differ.
For emergency services and public facility planning, inbound service areas are usually more meaningful.
Applications in Practice
Emergency Service Planning
Where should a new fire station be located to maximize coverage of uncovered areas? Generate service areas for all existing stations, identify the largest gaps, then evaluate candidate sites by the additional area their service area would cover. This is a coverage optimization problem, and service area analysis is the spatial foundation.
Logistics and Delivery
Which orders can a vehicle deliver in a single shift given its starting location, stop sequence, and time constraints? Network routing with time windows and vehicle capacity constraints answers this. Territory design for sales regions and delivery zones uses service area analysis to define balanced, contiguous zones.
Retail Site Selection
Where is the nearest competitor within a 10-minute drive of each proposed store location? Routing analysis. What share of the target population falls within a 15-minute drive of each candidate site? Service area analysis combined with population data.
Urban Infrastructure Planning
Where are the gaps in transit coverage? Compute walking-distance service areas from all transit stops and identify populated areas outside the covered zone. Where should cycling infrastructure be prioritized? Identify origin-destination pairs with high demand but long network routing paths due to missing connections.
Utilities and Telecommunications
Trace downstream from a pipe break to identify all affected customers. Find the shortest path for a new cable route that avoids environmentally sensitive areas. Both are network analysis problems, applied to utility and telecom networks rather than road networks.
Tools for Network Analysis
ArcGIS Network Analyst
The most feature-complete commercial option. Supports routing, closest facility, service area, origin-destination cost matrix, vehicle routing problem, and location-allocation. Requires a licensed ArcGIS Pro installation and the Network Analyst extension. Network datasets are built from polyline feature classes with defined connectivity and turn restrictions.
QGIS with GRASS or pgRouting
QGIS exposes network analysis through the GRASS GIS integration (v.net tools) and through the QNEAT3 plugin, which wraps pgRouting functions. For service areas, the ISO-Area tools in QNEAT3 produce isochrone polygons directly.
pgRouting
A PostgreSQL extension that adds routing functions to PostGIS. Operates directly on spatial tables. Supports Dijkstra, A*, bidirectional routing, and driving distance (service area equivalent).
-- Find shortest path between two nodes using Dijkstra
SELECT seq, node, edge, cost
FROM pgr_dijkstra(
'SELECT id, source, target, cost FROM roads',
source_node_id,
target_node_id,
directed := true
);
-- Find all nodes reachable within cost threshold (service area basis)
SELECT node, agg_cost
FROM pgr_drivingDistance(
'SELECT id, source, target, cost FROM roads',
start_node_id,
max_cost,
directed := true
);
NetworkX (Python)
For custom analysis and research, NetworkX is a Python library for graph construction and algorithm implementation. It does not handle spatial data natively but integrates with GeoPandas. Useful for non-standard network problems where off-the-shelf tools do not fit.
OSRM and Valhalla
Open-source routing engines designed for high-performance, production-scale routing on OSM data. OSRM uses Contraction Hierarchies. Valhalla uses a tile-based graph and supports multimodal routing. Both can be self-hosted and queried via HTTP API. Useful when you need programmatic routing at scale without commercial API costs.
Common Mistakes
Ignoring turn restrictions. A network without turn restrictions produces routes that cut across prohibited turns, particularly in urban areas. Model turn restrictions explicitly or accept that routes may be operationally invalid.
Using Euclidean distance as a proxy. Straight-line distance is not network distance. A facility that appears close on a map may be far by road due to river crossings, highway access constraints, or sparse road networks. Always compute service areas on the actual network.
Treating the network as undirected. One-way streets and directional constraints matter, especially in dense urban areas and for inbound vs outbound service area calculations.
Skipping topology validation. Disconnected edges, incorrect snap tolerances, and missing nodes produce silent errors in routing results. Validate topology before analysis, not after you have acted on incorrect outputs.
Confusing cost types. Distance-based and time-based service areas look different and answer different questions. Be explicit about which cost you are using and whether it matches the real-world question you are trying to answer.
Summary
Network analysis models infrastructure as a graph of nodes and edges with cost attributes. Routing finds the lowest-cost path between points using algorithms like Dijkstra, A*, or Contraction Hierarchies. Service areas map the zone reachable from a facility within a cost threshold, showing coverage in terms of real network travel rather than straight-line distance.
Together, routing and service area analysis underpin logistics optimization, emergency planning, infrastructure siting, and urban design. The spatial questions they answer — how to get there, and what can be reached from here — are among the most practically consequential in GIS.