EN
Back to the archive

The encyclopedia · Product & Design · Operational decision · 2011–2013

Kroger stopped over-ordering and under-ordering pharmacy stock with a simulation

Kroger's pharmacies produce on simulation-and-optimization demand forecasts, saving the grocer over $170M a year.

Kroger

the move

A pharmacy's inventory problem is a mismatch between uncertain demand and a finite shelf of drugs that also expire. Overshoot and money is sunk into stock that goes unused; undershoot and the customer walks to a competitor.

Kroger built, with Wright State's Xinhui Zhang, a set of intuitive demand models and fast calculation routines to project pharmacy demand more accurately and set stock accordingly. A simulation-and-optimization engine tests ordering rules against the range of possible demand.

By treating demand as a distribution rather than a single guess, the model lets Kroger hold less stock while cutting the out-of-stock events that lose customers.

why it works

  • It modelled the uncertainty directly, so it could balance overstock and stock-out rather than guess.
  • Quick computation made the models usable day-to-day, not a periodic review.
  • It reduced inventory cost and out-of-stocks at the same time, instead of trading one for the other.
  • The work showed a university partner could build a system a big retailer runs every day.
the payoffStock to a probabilistic forecast, not a point estimateneat

what transfers

When demand is uncertain, don't forecast a single number and stock to it — simulate the distribution and optimize the order policy against the risk of both overstock and stock-out.

what came after

Kroger was already saving more than $170 million a year when the project was named a finalist for the 2013 INFORMS Franz Edelman Award. The Interfaces paper also reported an $80 million a year rise in revenue, more than $120 million cut from inventory, and a $10 million a year reduction in labor cost.

references

spotted an error? The archive wants to know.

same kind of clever