Last-mile routing is the logistics problem of sequencing many individual delivery stops into feasible vehicle journeys that satisfy capacity, timing, access and receiver constraints.
In one line: the last mile is hard because the route stops being one long movement and becomes many small promises that must all fit inside the same day.
This is Article 49 in eduKateSG’s 100-article logistics authority build. The canonical parent remains How Logistics Works. Batch 12 built custody and observability. Batch 13 turns to the receiver-facing edge of the network, where one efficient trunk flow fragments into many separate destinations.
Reader Status and Scope
- Reader job: understand why last-mile routing is a constrained sequencing problem rather than a shortest-path problem.
- Mechanism owner: stop sequence, vehicle capacity, time windows, service time, access constraints, traffic, route balance and dynamic re-planning.
- Boundary: this article does not own parcel-locker economics or failed-delivery analysis. Those follow separately in Articles 50–52.
- Evidence anchor: recent last-mile research continues to formulate delivery as vehicle-routing and location-routing problems with time windows, parcel lockers and other receiver constraints because the useful route is the one that satisfies the whole service system, not merely minimum kilometres.
Why the Final Ten Kilometres Can Be Harder Than the First Thousand
A container vessel can move thousands of containers between two major ports on one long trunk route. A delivery van in a city may have to satisfy one hundred individual receiving promises before the driver’s shift ends.
The distance is shorter. The decision surface is larger.
- Different addresses.
- Different time windows.
- Different parcel sizes.
- Different building access rules.
- Different parking conditions.
- Different service times.
- Different probabilities that the receiver is available.
This is why last-mile logistics often feels disproportionately expensive and fragile.
Trunk logistics moves volume. Last-mile logistics resolves individuality.
The Basic Routing Chain
Orders → delivery points → vehicle / courier capacity → service constraints → route sequence → departure → live execution → exceptions → proof of delivery → route completion.
The route is therefore an operating plan, not merely a line on a map.
The Shortest Path Between Two Stops Is Not the Same as the Best Route Across Many Stops
Navigation software can find a short road path from A to B.
Last-mile routing asks a different question: in what sequence should Stops A, B, C, D and fifty others be visited so the complete set of delivery commitments remains feasible?
The best next stop locally can produce a poor overall route globally.
Vehicle Capacity Couples the Stops Together
A van has finite weight and cube.
A route containing many bulky but light parcels may run out of space before weight capacity. A route containing dense goods may hit payload limits first.
That means route assignment depends on Weight vs Cube as well as geography.
The mathematically shortest route can be physically impossible if the wrong stops are assigned to the same vehicle.
Time Windows Turn Routing Into Scheduling
Suppose Customer A can receive only between 10:00 and 11:00 while Customer B requires delivery after noon.
The route cannot simply visit whichever address is geographically closer.
Time windows create precedence and waiting. A vehicle can arrive too late and fail the service—or arrive too early and wait unproductively.
The final-mile route is therefore a combined routing-and-scheduling problem.
Service Time Matters at Every Stop
Driving is only part of the route clock.
- Find parking.
- Walk to entrance.
- Check security.
- Use loading bay.
- Wait for lift.
- Call receiver.
- Unload.
- Scan parcel.
- Capture signature or photo.
- Return to vehicle.
A five-minute difference in average stop service time multiplied across eighty stops can change whether the route finishes within the shift.
Parking Is Part of Routing Even Though It Is Not a Delivery Address
The route may say “deliver to Building X”. The operational vehicle still needs somewhere legal and practical to stop.
A dense street with poor kerb access can create more route delay than a slightly longer suburban road with easy parking.
This is why route distance and route friction are different variables.
Building Geometry Creates a Hidden Final Metre
High-rise cities compress addresses geographically but can expand vertical travel.
A courier may reach several customers in one building with almost no vehicle distance yet spend substantial time moving through lifts, security, corridors and loading-bay rules.
The street map sees one point. The courier experiences a small internal network.
Route Density Changes the Economics
If ten customers live in one small district, the vehicle can spend more of its shift serving and less of its shift travelling between demand points.
If those ten customers are spread across a large area, distance and dead travel increase.
This is important enough to deserve its own reader job. Article 51, Delivery Density, will explain why close customers change cost per stop even when every delivery remains individually simple.
Customer Availability Changes Route Risk
Attended deliveries need a human or authorised receiving point to be available.
A route can be geometrically excellent and economically poor if several stops fail because nobody is available.
This creates a probabilistic routing problem: the planner may need to consider not only distance and time but the chance of a successful handoff.
Failed Stops Damage the Rest of the Route
A failed delivery is not always “zero time”.
- The driver still travelled there.
- Parking was still needed.
- The address was still approached.
- Contact may have been attempted.
- An exception record must be created.
- The parcel remains on the vehicle or enters a recovery process.
The remaining route now carries both its original work and a failed promise.
Article 50 takes this multiplication effect directly.
Stop Sequence Changes Parcel Accessibility Inside the Vehicle
A route can be optimised digitally and still perform badly if the vehicle is loaded in a way that makes the next parcels difficult to reach.
Good last-mile execution links route sequence to loading sequence so the first deliveries are not buried behind freight intended for the end of the day.
This is a small but important bridge between packing, staging and route execution.
One-Way Roads and Turn Restrictions Break Euclidean Intuition
Two addresses can be physically adjacent and operationally far apart because the road network forces a long detour.
Bridges, tunnels, gated estates, restricted loading zones and vehicle-height limits can all change effective distance.
The route must follow the usable network rather than the geometric map.
Traffic Makes Travel Time Time-Dependent
A road segment may take ten minutes at 10:00 and thirty minutes at 17:30.
This means the cost of visiting a stop depends partly on when the route reaches that area.
Dynamic routing can use live or predicted traffic, but it should not constantly rearrange the route in ways that destroy driver clarity or customer promises.
Route Re-Optimisation Needs Stability
As new information arrives, the mathematically optimal future sequence can change every few minutes.
But repeatedly changing the route has its own cost:
- Driver confusion.
- Customer ETA volatility.
- Loading-order mismatch.
- Broken planned breaks.
- More system interaction.
Dynamic optimisation should therefore include a stability cost, not assume changing the route is free.
Priority Deliveries Create Asymmetric Consequence
An urgent medical shipment, expensive service part or timed business delivery can deserve priority over an ordinary parcel even if that increases route distance.
The route should therefore minimise consequence-weighted failure, not blindly minimise kilometres.
Route Balance Matters Across Drivers
If one driver receives forty easy stops and another receives twenty difficult high-rise stops, stop count alone says the first route is heavier.
Actual workload can be the opposite.
Route balancing should consider service time, building type, parcel weight, walking distance and difficulty, not merely number of drops.
Driver Knowledge Is Real Data
Experienced drivers know that one building’s loading dock closes at lunch, one security desk takes ten minutes, one street has no legal stopping space and one receiver is consistently available only after 15:00.
Routing systems improve when that repeated local knowledge becomes structured operational data rather than disappearing when a driver changes route.
Locker and Pickup-Point Stops Change the Route Topology
If ten household parcels can be deposited into one shared locker bank, ten individual home stops become one delivery node plus ten customer collection journeys.
That can dramatically change vehicle routing, but it moves some travel and inconvenience from the carrier to the receiver.
Article 52 will test that trade-off directly.
Route Quality Needs More Than Distance
- Kilometres travelled.
- Vehicle hours.
- Stops completed.
- Service time per stop.
- On-time stop rate.
- First-attempt delivery success.
- Parking / access delay.
- Overtime.
- Vehicle utilisation.
- Customer promise compliance.
One route can be shorter and still be worse if it produces more failed or late deliveries.
Last-Mile Routing at Three Zoom Levels
One stop
What travel, parking, access, service and receiver constraints make this stop expensive or risky?
One route
Does the sequence fit vehicle capacity, time windows and driver hours while preserving enough slack for variation?
One city network
How should depots, delivery regions, lockers, access rules and demand density be designed so routes remain productive without externalising unacceptable congestion or receiver travel?
A Singapore Lens
Singapore compresses physical distance but intensifies access complexity. High-rise buildings, security, loading bays, parking restrictions, lifts and dense demand mean the final-mile clock often depends more on service and access time than on driving distance.
This is why a geographically tiny route can still be operationally long.
Hostile Test: “The New Route Is 12% Shorter”
Did route time fall?
Did late stops increase? Did drivers spend longer finding parking? Did failed attempts rise? Did overtime move to a smaller number of difficult routes? Did customer ETAs become unstable?
Shortest distance is one optimisation objective. The real route must return to completed service.
Last-Mile Routing Audit
- What stop set must be served?
- What weight and cube constraints apply?
- Which stops have hard or soft time windows?
- What service time is expected by stop type?
- Which access and parking constraints matter?
- How does traffic change travel time by hour?
- Which receivers require attended delivery?
- What is first-attempt success by stop type?
- Does vehicle loading match stop sequence?
- How much route slack absorbs variation?
- Which route changes are allowed during execution?
- Are routes balanced by real workload rather than stop count?
- Does the route improve completed on-time deliveries rather than kilometres alone?
Evidence and Further Reading
Recent last-mile operations research continues to model urban delivery as a constrained routing problem rather than pure shortest-path travel. See Zang, Jiang and Meng, Urban last-mile delivery services with time windows and parcel lockers (2025), and Yağmur, Joint vehicle and courier routing problem in last mile delivery with parcel lockers (2025).
Return to the Logistics Hub
Last-mile routing begins Batch 13 by turning the final delivery problem into a constrained stop sequence. Return to How Logistics Works for the complete mechanism. Continue next to Failed Delivery Attempts | How One Missed Handoff Multiplies Cost.
Final compression: the final mile is difficult because every stop is a small constrained handoff. Routing succeeds when the complete sequence fits vehicle, time, access and receiver reality—not when the map merely draws the shortest line.