An end-to-end ERP that cut food waste by 30%

A multi-site grocery retail chain running fresh and ambient ranges across a distribution network, with per-store variation in assortment and footfall.

Grocery margin is thin enough that operational sloppiness shows up directly in the P&L, and this chain was leaking in three places simultaneously.

Staff rotas were built in spreadsheets that had never seen a sales forecast, producing overstaffed quiet days and understaffed peaks. Supply ran on standing orders adjusted by instinct, so deliveries arrived late or arrived wrong. And fresh stock was written off in volume because re-ordering reflected last month's assumptions rather than this week's actual sell-through.

Each site could see its own numbers, more or less. Nobody could see the network — which meant stock sat in one store while another ran out of the same item.

The three problems shared a root: no single demand signal. Discovery treated them as one system rather than three projects, which is why the result is an ERP rather than a re-ordering tool bolted onto a rota tool.

We built the data layer first, unifying sales, stock and supplier data across the network into one schema, then put the predictive modules on top of it.

• Predictive re-ordering — order proposals generated from real sell-through, seasonality and known local events, with a human approving rather than calculating.
• Demand-led shift planning — rotas built against the same forecast that drives purchasing, with labour cost per shift visible while planning rather than after payroll.
• Network-wide live inventory, making inter-store transfers possible early enough to prevent both a stockout and a write-off.
• Supplier performance tracking — lead-time reliability measured automatically and fed back into ordering, so chronic under-deliverers stop being planned around optimistically.