"Last-mile delivery" is the leg of the journey from a local hub or store to the customer's door. It's the shortest distance and, weirdly, the most expensive. Industry analyses consistently put last-mile at 40–50% of total shipping cost. It's also where customer experience is made: a perfect warehouse and a flawless line-haul mean nothing if the driver shows up an hour late with a melted package.
This guide is a working playbook for delivery and courier operators on running last-mile well. It covers dispatch, routing, the driver experience, proof-of-delivery, customer communication, and the small set of metrics that actually predict whether your operation makes money.
Why last-mile is so expensive
Three structural reasons:
- Density falls off. A single truck moving 200 packages between two warehouses is incredibly efficient. The same 200 packages spread across 200 doorsteps means 200 stops, 200 chances for someone to not be home, and 200 chances for a wrong address.
- Failed deliveries are catastrophic. A single failed delivery can cost 2–3× a successful one — you've already paid for the trip out, you'll pay again to retry, and you'll likely deal with a refund or chargeback.
- Customer expectations are now Amazon-shaped. Free delivery, two-hour windows, real-time tracking. Whether or not you can deliver on those, customers compare you to that bar.
The four-stage last-mile workflow
1. Order capture & validation
The earliest place to save money is at the top of the funnel. Address validation, customer confirmation of the pickup window, and clarifying any access notes ("apartment 3B, ring the bell, leave with neighbor if not home") happen before a driver leaves. A 30-second SMS or WhatsApp confirmation can prevent a 45-minute failed trip.
- Geocode the address and confirm with the customer if confidence is low.
- Capture access notes as a structured field, not buried in chat.
- For high-value items, ask for ID requirements upfront.
2. Dispatch & routing
This is where software earns its keep. A good dispatcher asks five questions every time an order arrives:
- Who's the closest available driver to the pickup?
- Are they already on a route that this order fits into?
- Is the customer's window compatible with that driver's schedule?
- Does the order have constraints (cold-chain, fragile, ID required)?
- What's the priority — same-day, next-day, or scheduled?
For small fleets, manual dispatch with a good map view is usually fine. Once you cross 50–100 orders a day, manual dispatch becomes the bottleneck and you need auto-assignment with the option to override. Numus auto-dispatches based on driver location, capacity, and current route — with full manual override.
For multi-stop routes, true route optimization (not just sorting by distance) can save 15–25% on total drive time. The best routing accounts for traffic, time windows, and vehicle capacity.
3. Driver execution
The driver's mobile experience determines data quality for the whole operation. Drivers will only update statuses and capture proof-of-delivery if it takes 2 seconds, not 30. Things that work:
- Single-screen workflow. Map, current stop, big "Picked Up / Delivered" buttons. No menus.
- Offline-first. Driver phones drop network all the time. The app must queue actions and sync when signal returns.
- One-tap navigation. Hand off to Google Maps or Waze immediately — don't try to be a navigation app.
- OTP-verified handoff. A 4-digit code from the recipient is the simplest, cheapest proof-of-delivery you can get.
4. Customer communication
The single highest-leverage thing you can do is send a tracking link. A customer who can see the driver moving on a map sends 5× fewer "where is my order" messages and is dramatically more likely to be home when the driver arrives.
- Send the tracking link the moment the driver accepts the order.
- Auto-message at status changes: assigned, picked up, en route, delivered.
- Include the driver's first name and a photo (with consent) — humanize the experience.
- If the customer is on WhatsApp, send updates there. Don't make them download an app.
The five metrics that actually matter
Pick these. Track them weekly. Ignore the rest until these are healthy.
- On-time delivery rate (OTDR). Percentage of deliveries completed within the promised window. Target: 95%+.
- First-attempt success rate. Percentage of deliveries completed on the first try, with no retry. Target: 90%+.
- Cost per delivery. All-in cost (driver pay, fuel, software, support) divided by completed deliveries.
- Driver utilization. Active delivery time divided by total clocked-in time. Target: 65–80%. Below 65% means too many drivers; above 80% usually means burnout.
- Net Promoter Score (NPS) on the delivery experience. A one-question survey after every delivery: "How likely are you to recommend us?"
Common failure modes
The cheapest way to fix a problem is to recognize it. Most struggling last-mile operations are stuck on one of these:
- Address quality is bad and nobody is fixing it. Drivers are spending 10+ minutes per stop hunting for the building.
- Drivers are running solo. No team WhatsApp group, no exception handling. When something breaks, the driver shrugs and moves on.
- Cash reconciliation is happening on paper. Cash drifts daily. By month-end, you're missing thousands.
- The dashboard is a vanity metric show. Pretty charts, no operating decisions come out of them.
- Customer support is a personal phone number. Doesn't scale past 50 orders a day.
Where Numus fits
Numus is built around exactly the workflow above. Order intake from web, WhatsApp, and phone. Auto-dispatch with manual override. Driver app with offline support and OTP. Public tracking link for the customer. Cash reconciliation per shift. Operator dashboard with the five KPIs above front and center. See the full feature breakdown or check pricing.
If you do nothing else this quarter, send a tracking link the moment a driver accepts an order. It's the single highest-impact change you can make to last-mile delivery — and it's free.