Ocado Group · London · 2022 — 2024
Network digital twin
Design and delivery of a decision product that let major retailers simulate and optimize their network before committing capital.
Context
Major retailers needed to reason about network design, cost-to-serve and inventory placement across highly complex, fast-changing operations.
Intervention
- Led design and delivery of the digital twin from concept to enterprise deployment.
- Connected engineering, data science, business and operational stakeholders around one fact base.
- Translated complex optimization capabilities into decisions operators could actually make.
Outcome
- Up to 40% lower cost-to-serve.
- Up to 30% lower working capital.
A digital twin of the distribution network
Retail networks fail slowly: capacity fills, bulk purchase quantities outgrow single sites, and third-party warehouses quietly absorb margin. None of that is visible on a spreadsheet. The digital twin made the whole flow — supplier to cross-dock to fulfilment centre to customer — simulatable, so network design, cost-to-serve and inventory placement could be optimised before capital was committed.
The multi-echelon distribution work is the clearest example: a decision that looks operational is in fact a P&L decision worth around half a point of EBITDA.
The questions the twin had to answer
Where should stock physically sit?
A large share of warehouse space is immobilised by slow-moving units bought in bulk — around 30% of stockholding rotates more than 30 days after inbound, while grid utilisation in some fulfilment centres reached 98%.
Direct, third-party, or in-house consolidation?
Three replenishment tactics coexist: supplier direct to a fulfilment centre, outsourced cross-docking through a 3PL, and in-house consolidation where one fulfilment centre supplies the others.
What does each option actually cost?
Cost-to-serve is modelled at unit level across inbound, outbound, third-party billing and transport, so a network decision can be argued in pence per unit rather than in opinions.
What the simulation showed
3PL billing is the expensive path
£0.11 per unit of third-party billing against under £0.07 for inbound and outbound handled in-house — roughly 40% cheaper, and up to 70% cheaper in the proof of concept run across three sites.
Consolidation is not optional
At 11% of throughput already consolidated the network still ran above 86% utilisation. Absorbing growth means more consolidation, so the only real question is which mode carries it.
The two modes are complements
If a bulk order can be consumed by one site in a week, ship direct. If the network can consume it but no single site can, consolidate in-house. If neither can, use the 3PL.
Simulation before capital
Scenarios at 11% and 20% consolidation, split across mixes of 3PL and in-house, expose an EBITDA swing worth roughly £10M a year — before a single pallet moves.
From model to technology pack
- Replaced manual, spreadsheet-based network planning with an optimiser-driven multi-echelon distribution feature in the supply-chain platform.
- Paired the software with a materials-handling change so storage totes move between sites without double-handling the stock.
- Framed the business case around three explicit assumptions, each tested empirically rather than asserted.
- Sequenced the roadmap to a production release, with the optimiser's in-situ performance held as the main delivery risk.
digital twin · optimization · enterprise product