Case study · Delivery
Better routes and ETAs customers can trust
Delivery businesses lose money on inefficient routes and customer trust on vague delivery windows. This is how we combined route optimisation with machine learning ETA prediction, so drivers follow better routes and customers get accurate arrival times.
Project overview
Routing and ETA prediction for last-mile delivery.
- Industry
- Delivery & logistics
- Client
- Confidential
- Solution
- Routing and ETAs
- Platform
- Optimiser + Azure ML
- Timeline
- 10–16 weeks
- Team
- 3–4 people
- Constraints
- Time windows, capacity, shifts
- Output
- Driver routes, customer ETAs
- Measurement
- Pilot depots vs control
Expected business outcomes
What the client set out to achieve
Measured in pilot depots against current planning.
-
5–15%
Fewer miles per delivery
Optimised routes respecting windows and capacity.
Target range compared with current planning
-
Narrower
Customer delivery windows
ML ETAs based on real drive and stop times.
Measured as ETA error
-
10–25%
Fewer failed deliveries targeted
Customers informed accurately and able to prepare.
Target range; depends on notification adoption
Business challenges
Where the client started
Routes are planned by experience; windows are wide.
Planning
Planners build routes by hand or with basic tools.
Windows
Customers get all-day or half-day windows.
Stops
Stop times assumed, not measured.
Failures
Customers out when the driver arrives.
Growth
More volume means more planners.
Proposed solution
What the build covers
Optimisation for planning, machine learning for timing.
-
Route optimiser
Daily routes that respect windows, vehicle capacity and driver shifts.
-
ETA models
Predicted drive and stop times from historical GPS and delivery data.
-
Driver app
Routes, navigation and proof of delivery.
-
Customer notifications
Accurate ETAs by SMS or email, updated during the day.
-
Planner tools
Planners adjust routes and see the impact.
-
Performance tracking
Miles, ETA accuracy and failed drops by depot.
Project requirements
What the client asked us for
The request was efficiency. Accurate ETAs deliver a second benefit: fewer failed drops because customers know when to expect the driver.
Planners keep control: the optimiser proposes, planners can adjust.
What the project had to deliver
- 01Optimised daily routes
- 02Accurate ETAs
- 03Customer notifications
- 04Planners in control
Key challenges & considerations
What made it harder, and what we did
-
Poor-quality address data
What we did
Addresses were cleaned and geocoded before routing relied on them.
-
Unpredictable stop times
What we did
Stop times were measured from real deliveries rather than assumed.
-
Driver familiarity with areas
What we did
Planners could adjust routes, and their changes were kept.
-
Integration with order systems
What we did
Routes and ETAs were exchanged with the order system through its API.
What we watched for
- Clean address data early
- Measure stop times, don't assume them
- Let planners adjust routes
- Compare with control depots
Development process
Pilot depots first
Results are compared with depots still planning the old way.
-
Data
GPS, orders and delivery history prepared.
-
Optimiser
Routing configured with real constraints.
-
ETA models
Models trained and validated on historical routes.
-
Pilot
Pilot depots use routes, app and notifications.
-
Rollout
Other depots follow with performance tracking.
Technology stack
Optimisation and ML together
Proven solvers and prediction models.
Optimisation
- Vehicle routing solver
- Constraint modelling
Machine learning
- ETA models
- Azure Machine Learning
Apps
- Driver app
- Planner console
Notifications
- SMS and email
- Tracking links
Quality assurance & deployment
Measured against a test set before launch
AI quality was measured, not judged from a demo. A test set of real examples with agreed answers decided whether the system was ready, and the same set was re-run after every change.
- A test set of real examples with agreed answers
- Accuracy thresholds agreed with the business before launch
- Safety checks: refusals, permissions and prompt-injection tests
- A person reviewing low-confidence output
- Cost and response time measured at real volumes
Deployment
- 01PilotPilot depots use routes, app and notifications.
- 02RolloutOther depots follow with performance tracking.
- 03Pilot, then staged rolloutA small group of real users first, with results reviewed weekly before widening access.
- 04Monitoring and warrantyAccuracy, drift, usage and cost tracked after launch, and six months of free bug fixing.
Final deliverables
Fewer miles, fewer missed drops
Drivers follow optimised routes, customers receive accurate arrival times and failed deliveries fall.
Planners handle more volume with better results.
- Optimised routes
- ML ETAs
- Customer notifications
- Depot performance tracking
Handed over to the client
- 01Production AI serviceDeployed in the client's own cloud tenant
- 02Source code and modelsIn a repository the client can access
- 03Test set and resultsThe evidence behind the launch decision
- 04MonitoringAccuracy, usage and cost
- 05Six-month warrantyFree bug fixing after delivery
Success metrics
How success is measured
| Measure | Outcome | How it is measured |
|---|---|---|
| Fewer miles per delivery | 5–15% | Target range compared with current planning |
| Customer delivery windows | Narrower | Measured as ETA error |
| Fewer failed deliveries targeted | 10–25% | Target range; depends on notification adoption |
Verified reviews
What clients say about our custom software work
Verified reviews from our clients on the same kind of work, published on Clutch.
-
Custom Software Development
Web-Based Coupon App Development
The work produced by the team was of exceptional quality.
Jon Thies
CTO, Model Rocket Feb 2022 – Mar 2023
-
Custom Software Development
Mobile & Web Platform Development & Design
Silver Scintilla was responsive, respectful, and enjoyable to work with, quickly addressing any questions or requests.
DeShawn Brown
CEO & Founder, Lithios Jun 2021 – Nov 2022
-
Custom Software Development
Web-Based HRM System Development
Silver Scintilla provided the best solution with scalable functionality, personalized features, and robust security.
Suheb Khan
Director, Collaborative Insight Technologies Jul 2020 – Jun 2022
Free discovery call
Have an idea? Let's turn it into AI-powered software.
Book a free discovery call with our experts. Share your idea and we will help you shape the scope, timeline and budget, under NDA.
- Free consultation
- NDA before we talk
- Transparent estimate