Case Study #3: Order Management Optimization for Supply Chain
Assigning orders to plants and routing them across transportation lanes decides both delivery time and fulfilment cost. This manufacturing client relied on a linear optimisation model that was not scientifically suited to the task and lacked the features it needed. We designed a transportation lane cost prediction model and a new stochastic optimiser, informed by historical data and predictive maintenance. Prediction error fell from 30% to 9%, and the client saved millions in transportation and operations cost.
At a glance
Client: manufacturer shipping from multiple plants across transportation lanes · Industry: Supply chain, manufacturing · Scope: order-to-plant assignment and lane cost optimisation
9%
Prediction error
down from 30% with the new cost model
Millions
Saved in USD
on transportation and operations cost
2
New models
lane cost prediction and stochastic optimisation
Business challenges
- Fulfilment cost was high because orders were not assigned to the best plant and lane.
- The existing linear optimisation model lacked the features needed and was not scientifically suitable for the task.
- Transportation cost per lane was hard to predict, so plans were built on unreliable inputs.
Project objectives
- Predict transportation lane costs accurately.
- Assign orders to plants and lanes at the lowest total cost while keeping deliveries on time.
- Replace the linear model with an optimisation approach that handles uncertainty.
Our approach
1. Model assessment
We identified the missing features in the client's current model and concluded that the linear optimisation model in use was not suitable for the order management task.
2. Lane cost prediction model
A new transportation lane cost prediction model was designed from historical data and predictive maintenance signals.
3. Stochastic optimiser
We introduced a stochastic optimisation approach and built a new optimiser that outperformed Gurobi and significantly improved task handling within the client's application computing system.
Results
- Prediction error reduced from 30% to 9%.
- Multi-million dollar savings on transportation and operations cost.
- Faster, lower-cost deliveries from data-driven plant and lane assignment.
Business impact
Strategic order assignment and lane optimisation, backed by an accurate cost model and a stochastic optimiser, cut prediction error to 9% and delivered multi-million dollar savings while keeping deliveries on time.
Conclusion
Replacing an unsuitable model with the right one turned order management from a cost centre into a lever for profitability.
Burning cash on fulfilment?
We optimise order assignment and transportation with predictive models. See our Supply Chain Optimization solution, or talk to us about your order data.
Tools used
