Data Department helps distributors, wholesalers and delivery companies make better inventory, fleet and routing decisions with machine learning (ML), operations research and AI (artificial intelligence). Our work in this sector includes a spare parts forecasting system covering more than 100,000 parts and fleet optimisation work whose analysis showed that the same service level could be met with about 30% fewer vehicles. We integrate with your ERP (enterprise resource planning) and WMS (warehouse management system).
Where does AI pay off in distribution and logistics?
Distribution margins are thin and the decisions are many: what to hold, where to hold it, how many vehicles to run and which vehicle goes where. Small improvements, repeated daily, add up. Some of these decisions need prediction; others need optimisation. We use whichever fits.
| Decision | What ML, AI or optimisation does | Typical data |
|---|---|---|
| How much stock to hold, per item and location | Probabilistic demand forecasting that sets stock by acceptable risk | Order history, lead times, stock levels, substitution and compatibility data |
| How to allocate stock between warehouses and branches | Forecast-driven allocation and replenishment rules | Branch sales, transfer history, WMS stock positions |
| How many vehicles to run | Fleet sizing with MIP (mixed-integer programming) under service-level constraints | Delivery volumes, time windows, vehicle capacity, depot locations |
| Which vehicle serves which deliveries | Clustering plus route assignment optimisation | Addresses, delivery windows, priorities, capacities |
| Which customers or outlets need attention | Early prediction of underperformance or churn, so teams act sooner | Outlet sales, visit history, asset placement |
| Whether a large plan will hold | Simulation that estimates the probability of hitting milestones | Resources, schedules, dependencies |
| How to process shipping and supplier documents | Document AI that extracts fields from waybills, invoices and orders | Scanned or digital documents, ERP master data |
Relevant case studies
- Spare parts inventory forecasting: a large automotive distribution group with 100,000+ active parts across several brands and channels. Probabilistic forecasts and segmented models gave ~30% better forecast accuracy than the legacy method and ~10% lower inventory holding cost. For workshop planning and other after-sales work, see our automotive page.
- Fleet and route optimisation: a parcel delivery company. We combined custom clustering with mixed-integer programming, respecting delivery windows, priorities and capacity. The analysis showed that the same service level could be met with about 30% fewer vehicles.
- Outlet cooler performance: a beverage bottler. Our model flagged outlets whose coolers would underperform with 99% accuracy, ahead of the company's own review, with an estimated ~2.35% annual sales upside.
- Programme delivery simulation: for a global networking company, a Monte Carlo simulation of a large migration programme showed early that it would miss its targets with current resources. The same method applies to large logistics roll-outs.
What are typical first projects for a distributor?
- Inventory forecasting for slow and intermittent items. Spare parts and long-tail items are where point forecasts fail most and where probabilistic forecasting helps most.
- Fleet sizing for one region or depot. A contained optimisation study that shows how many vehicles the current service level really needs. See our optimisation and simulation service.
- Branch replenishment. Forecast-driven stock allocation between a central warehouse and branches.
- Document extraction. Automating data entry from waybills, invoices and purchase orders into ERP.
Each starts with a two-week Discovery: ranked use cases, a data check and a go/no-go decision. The free 30-minute AI opportunity review is a quick way to test fit.
Data and integration notes
- ERP holds orders, items, suppliers, lead times and costs. It is the starting point for most forecasting projects, and forecasts are written back into ERP or planning tools.
- WMS holds stock by location and movement history. It matters for allocation, replenishment and warehouse planning.
- TMS (transport management system) and telematics data supply routes, stops, times and vehicle capacity for fleet and route work. Where no TMS exists, delivery records and addresses are often enough to start.
- Master data such as part compatibility and substitution lists can be as valuable as sales history, especially for spare parts.
- Infrastructure: we deliver on AWS (Amazon Web Services), Google Cloud, Azure or on-premises. Our spare parts system was built on AWS SageMaker with daily updates.
Frequently asked questions
Is route optimisation an AI project?
It is usually an operations research project. Mixed-integer programming and related methods find the best plan under hard constraints such as delivery windows and capacity. We often combine it with machine learning, for example clustering or demand forecasts.
How do you forecast items that sell rarely?
For sporadic and intermittent items we use methods designed for sparse data, such as hierarchical Bayesian models that borrow information from related and substitute items. The output is a forecast range, not a single number.
Can you reduce fleet size without hurting service?
In our parcel delivery work, the service level was a constraint, and the analysis showed that the same service level could be met with about 30% fewer vehicles.
Do we need a TMS or WMS to start?
No. Many projects start from ERP exports and delivery records. A WMS or TMS helps, but it is not a prerequisite.