Inventory Planning in Rocket Science Retailing

Offering the right product in theright place at the right time for theright price is retaiiing's formula forperfection.The ideal rennains elusive,but an elite rank of retailers is gettingcloser to it every day.There's muchto be learned from what they do.

Some retailers (we'll refer to retailers and e-tailershenceforth with the broader term) have dramaticallyimproved their performance in ordering, distribution,and merchandising. But those companiesare still a small, elite rank. The next step? An industrywidemove toward something we call rocket scienceretailing - the act of blending traditional forecastingsystemLS, which are largely based on theintuition of a handful of employees, with the prowessof information technology. Rocket science retailingfuses data and instinct with computer modelsand analysis to create a high-tech forecastingsystem supported hy a flexihle supply chain.

Inventory planning involves deciding when andhow much to order, or how much to produce, ofvarious raw materials, components, and finishedgoods. Inventory planning differs from forecastingbecause a planner might find it beneficial to stockmore or less than predicted demand. In planninginventory for a household, for example, you mightdecide to stock far more medicine than you anticipateneeding in case you become sick. Or you mightbuy certain items - batteries, for instance - manymonths' demand at a time while other items-breadand milk, for instance - might be ordered everyweek. Inventory planning at most retailers suffersfrom several shortcomings. One of tbe most glaringis that many retailers don't track stockouts andtbe resulting lost sales. Only 15 of the 32 companiesin our study said they track stoekouts, and 11 of the 13 used this information to estimate the resultinglost sales.

Lost sales are endemic among retailers, especiallyfor products with short life cycles. Tracking stockoutscould help retailers set optimal inventory levelsand could help them see the value in improvingsupply-chain responsiveness. So why aren't thesemetrics studied carefully? One reason is that it'shard to know how much of a product would havesold if supply had been plentiful. The figure canbe estimated using sophisticated statistical techniques,but retailers generally can't find such capabilitiesin commercial software, especially in thecase of short-life-cycle products.

There is a way over that hurdle. We developed amethod to estimate lost sales. Our procedure worksin two steps. First, it calculates the underlying demandrate for a product based on the sales patternsthat occurred when the product was in stock. Second,it combines the estimated demand rate withthe duration of the product stockout at a particularstore to derive the lost sales. To estimate demandrate and lost sales, the technique has to be modifiedfor factors such as the variation of demand on differentdays and at different times within a day. Inour experiments with real retail data, our techniqueestimated lost sales to within 2% at the store leveland with higher accuracy at the chain level or for acategory of products.

The benefits of tracking lost sales, and increasing inventory levels systematically to reduce thoselosses, can be substantial. One retailer found thatsales could be improved by roughly io% simply byincreasing inventory at the stores, suggesting thatlost sales-before the inventory boost-would haveaccounted for at least io% of sales. At Rome-basedjewelry manufacturer Bulgari, stockouts on a singleitem at one store had been high enough to reducethe store's revenue hy 3.5%. As a result, Bulgari isseeking ways to improve its planning processes.