CONSUMER GOODS & LOGISTICS OPTIMISATIONUtilising machine learning to reduce cost to serve in FMCG distribution
Using our ML-logistics platform; Grigora, we set out to identify an optimal network configuration and routing design for a UK-based consumer goods distribution business
How we realised £850,000 annualised savings from one Distribution Centre
A well-established food distribution business operating over 10 national DC’s in the UK asked us to pilot Grigora; our multi-scenario based logistics solution with one of their northern Distribution hubs.
Optimisation had been considered from varying persepctives, including distance and total cost.
Considering for some flexibility on fleet / driver utilisation, it was agreed to test a more complex verison of Grigora, which considered for a direct reduction in cost to serve, without compromising fulfilment capability.
Our client delivered to a challenging customer base, across a large territorial network, including reputable food service outlets and a highly competitive independent retail segment.
Deliveries could not be compromised…
Re-engineering logistics from a broader, more sophisticated optimisation perspective
The Client supplied nearly 400 regional customer outlets, most of which were receiving deliveries between 4 – 6 days per week.
Customer orders were usually processed and delivered the next day. It was important to be highly responsive but at what cost?
Transport planning had largely been undertaken on a manual basis, supported by head office and routing had been pre-determined through a previous exercise on network design.
There was some variation in the capacity individual vehicles could carry, which added compexity around right vehicle allocation to deliveries.
Key accounts were quite strict on delivery windows, some for the day shift and some for the night shift.
Driver shift length also had to be constrained to a target 7.5 – 8 hour shift period.
Using Grigora, Trusted Data, selected a solution methodology that would seek out total optimisation of routing, asset selection and asset utilisation.
The optimisation task was to generate the most profitable distribution strategy for 384 regular customer sites.
This needed to consider use of a permanent team of drivers and the relief team, varying vehicle load capacity, customer delivery window constraints, optimal routing by vehicle and a targeted arrival time back to the depot in <8hours.
Utilising a plethora of algorithms, to converge for the optimal outcome, the Client was able to road-test the solution for a 3 month period, prior to a successful review with HQ and group-wide implementation.
Annualised Savings on the test Pilot
average mileage reduction per delivery round
% of fleet rendered redundant
Transport planning time reduction