Introduction

“Last mile” is the path of a parcel from the final warehouse/hub to the customer’s door (or to a pickup point). This is where logistics becomes expensive, nerve-wracking, and very human: traffic, parking, intercoms, elevators “under repair,” the customer “not home right now,” two addresses in one order, and the eternal classic — “can you come in 10 minutes?”. On long-haul transport you buy kilometers in bulk. On the last mile you buy minutes and attempts.

A typical mistake is trying to optimize it with the same methods as “big logistics”: push the rate down and hope it magically resolves itself. It won’t. The last mile is expensive by nature: few parcels across many stops, high unpredictability, lots of manual work, and huge downstream impact from small failures. Let’s break down what the cost consists of, which metrics to manage, which approaches actually work, and how to implement improvements without chaos.

Basic concepts and selection criteria

1) Cost per delivery (cost per delivery / cost per stop)
This is the total cost for one successfully completed delivery: courier time, fuel, depreciation, dispatching, support, repeat attempts. Measured as RUB (or your currency) per delivery. It matters because a cheap “price per trip” can turn into an expensive “price per successful delivery” if there are many failures.

2) First-attempt delivery rate
The percentage of orders handed over on the first visit. Measured as %. It matters because a second attempt is almost always a full repeat of stop costs (and sometimes a noticeable hit to the customer’s mood).

3) Drop density
How many deliveries occur per kilometer or per hour of route time. Measured as deliveries/km or deliveries/hour. It matters because the last mile gets cheaper when stops are closer together and there’s less “empty driving.”

4) Stop time
How many minutes one delivery takes: parking, getting to the entrance, the elevator, waiting for the customer, signature/code, photo proof, walking back to the vehicle. Measured in minutes. It matters because 3 extra minutes per stop across 60 stops = 3 hours into nowhere. Yes, math can be ruthless.

5) NDR / failed delivery rate
The share of deliveries that didn’t happen: customer unavailable, wrong address, refusal, reschedule, closed location. Measured as % and by reasons. It matters because NDR directly turns into repeat trips, storage, returns, and load on support.

6) Returns rate and processing speed
How many orders get returned and how fast the item becomes available for sale again. Measured as % and cycle time in days. It matters because returns are “reverse last mile,” usually even more expensive and complex.

Factor How it increases last-mile cost What to control
Low drop density More driving and time between deliveries Zoning, consolidation, selecting services by area
Delivery failures (NDR) Repeat attempts, storage, support Address validation, time windows, contact confirmation, notifications
Long stop time Fewer deliveries per shift Handover procedure, pickup points/lockers, route preparation
Complex geography (city center / private housing areas) Parking, entrances, traffic restrictions Delivery type by zones, schedules, micro-hubs
Returns Double logistics and handling Returns policy, pre-shipment QC, analyzing reasons

Approaches and solutions

Option 1

Focusing on door-to-door delivery as the main channel

When it fits: premium service, products where convenience matters (bulky items, expensive SKUs), strong competition on delivery experience, a customer base with predictable windows.

Pros: high service level; less friction for the customer; easier to manage the end-to-end impression.

Limitations: the most expensive model with low density and high NDR; scales poorly into “difficult” zones.

Risks: if confirmations, windows, and communications aren’t tuned, costs start growing not linearly, but emotionally.

Option 2

Mixed model: pickup points/lockers + door delivery for selected segments

When it fits: an online store with a broad assortment, many “small parcel” orders, regions/sleeping districts, the need to reduce last-mile costs without losing control.

Pros: higher handover density (many orders at one point); lower failure rate; easier planning.

Limitations: some customers don’t like self-pickup; you need clear communication and incentives to choose pickup points.

Risks: if pickup points are overloaded or inconvenient, you’ll get dissatisfaction and increased support load.

Selection criteria

Step-by-step implementation guide

Preparation

Execution

  1. Zoning segmentation: split the city/region into zones and define rules (where pickup points are default, where door delivery applies, where only by time windows). Checkpoint: rules are clear and applied consistently.
  2. Reducing NDR: implement contact confirmation, notifications, reschedule-before-dispatch, clear time windows. Checkpoint: failure reasons are recorded as codes, not as “customer didn’t answer.”
  3. Reducing stop time: simplify handover (code/QR), give couriers advance info on entrances/parking, standardize the procedure. Checkpoint: average stop time decreases without a rise in complaints.
  4. Routing and “waves”: build routes in batches, accounting for time windows, density, restrictions. Checkpoint: more deliveries per shift with the same resources.
  5. Pickup points/lockers as a lever: gradually increase the share of pickup points in zones with poor parking and high NDR. Checkpoint: conversion doesn’t drop and support load doesn’t spike.

Result evaluation

Cases / micro-examples

Scenario 1: baseline — a city with many deliveries in the center; high NDR due to parking and “couldn’t reach.” Actions — introduced mandatory windows, time confirmation before dispatch, increased pickup point share in the central zone, and kept door delivery for bulky items. Result — fewer repeat attempts, and couriers started completing more handovers per shift.

Scenario 2: baseline — residential districts with high density, but long stop time due to entrances and waiting. Actions — standardized handover via code, added structured notes (entrance/floor/intercom), implemented the “reschedule before dispatch” rule. Result — stop time decreased, and the first-attempt delivery rate grew. We’ve worked in this field for over 13 years, and sometimes “last-mile optimization” isn’t fancy software—it’s discipline in details everyone is too lazy to do. That’s the catch.

Common mistakes and how to avoid them

Mini-FAQ

1) Why is last mile more expensive than long-haul even though the distance is shorter?
Because you pay not for kilometers, but for time and uncertainty: many stops, small batches, complex handover conditions, and a high risk of repeat attempts.

2) Where to start optimizing if resources are limited?
Start with NDR and stop time. Two simple levers: improve address/contact quality and reduce waiting (windows, notifications, reschedule before dispatch). This often gives the fastest effect without “rebuilding the world.”

3) Are pickup points and lockers always cheaper?
Often cheaper in areas with poor parking and high NDR, but not universally. If your product requires door service or your segment expects it, the model should be mixed: pickup points as the main channel where it’s rational, and couriers where it’s justified.