Campbell Soup took over refilling its retailers' warehouses and cut stock by two-thirds

In the early-1990s grocery trade, promotions and batch ordering made the orders reaching Campbell Soup surge for reasons unrelated to soup eating: at four retailers, Tuesday alone carried 22% of weekly demand, versus 15.5% on an ordinary day. Retailers answered with buffer stock and still suffered stockouts. Campbell's Continuous Replenishment attacked the distortion itself: retailers paid a constant EDLP price, kept their own promotions, and sent daily stock and withdrawal data over EDI; Campbell shipped full truckloads to hold each DC above a minimum — the supplier now ran the resupply.

In spring 1993 Campbell sponsored a Wharton project by Gérard Cachon and Marshall Fisher to test whether the rules could be improved. On a year of data covering several hundred SKUs at four retailers, they found Campbell's early tests had already cut DC inventories 50% while lifting fill rates from 98.7% to 99.5%; their rebuilt decision rules cut inventories another 33% — 66% below retailers' original levels — while maintaining service, worth about 1.2% of a retailer's cost of goods. Inventory needs rose sharply if service targets pushed past 99.5%.

The practice stopped being Campbell's alone. By 1997 MIT Sloan Review listed Campbell Soup with M&M/Mars, Nestlé, Quaker Oats, Nabisco, P&G and Scott Paper as suppliers running continuous replenishment with some or most of their customers, with inventory reductions up to 25% common in such alliances. A practice that inverted a century of buying tradition — the seller decides when the buyer restocks — had become the grocery industry's default answer to the bullwhip.

Let the supplier refill from the retailer's own data

Reading demand from withdrawals and stock removed the distortions planners were compensating for: order batching (the Tuesday spike) and forward-buying on promotions.

Incentives made surrender painless: EDLP kept the retailer's purchase economics unchanged, promotions stayed the retailer's weapon, and Campbell absorbed the planning work.

Daily re-computed decision rules beat intermittent human buying — they never panic before a promotion and never hoard; the 66% came from mechanical consistency.

EDI made the data flow cheap and routine, so replenishment became a computation rather than a negotiation, and fill-rate gains built trust.

A partner's orders are decisions, not demand. Whoever sees real consumption and controls refill can dampen the chain's swings — if incentives are aligned and service visibly improves.

Vendor-managed replenishment spread through consumer packaged goods and became the template for Efficient Consumer Response and later CPFR, while the data pipe thickened from EDI files to point-of-sale feeds and APIs. Cachon and Fisher's evaluation became a standard reference for what sharing demand and inventory data is worth, and its caution — that service beyond 99.5% makes inventory requirements explode — still marks the idea's practical boundary. The deeper legacy is conceptual: planners learned to treat customer orders as noise around demand.