TOOL KIT

Despite ever more demanding retailers and rampant product

proliferation, manufacturers have stayed with dangerously

indiscriminate production schedules and sourcing strategies.

A new approach leaves less money on the table.

II N V E N T O R Y

IN A W O R L D O F L E A N R E T A I L I N G

by Frederick H. Abernathy, JohnT Dunlop,

Janice H. Hammond, and David Weil

|ANUFACTURERS OF CONSUMER GOODS

are in the hot seat these days. 1 n the past,retailers would place large orders at thebeginning of each selling season, andfactories would simply produce to order.But the big chain stores are increasinglyadopting lean retailing practices, sothey're insisting that manufacturers fillorders to replenish retailers'stock on anongoing basis. Because factories usuallycan't produce goods fast enough to meetthese orders, manufacturers often holdlarge inventories for indefinite periods.

And the cost of holding these inven-tories is only growing. Consumers aredemanding greater variety in products,and their preferences are getting harderto predict. As products proliferate andbecome more susceptible to changingwhims,the risk grows that a given prod-uct line will have disappointing salesand have to be discounted. But if a man-ufacturer decides to go lean on inven-tories, it runs the risk of stockouts, lostsales, and endangered relationships withthe chains.

HARVARD BUSINESS REVIEW November-December 2000

TOOL KIT • Control Your Inventory in a World of Lean Retailing

It's a tough position, but a newapproach can help manufacturers pre-dict their inventory needs more accu-rately. Manufacturers tend to treat everystock-keeping unit within a product linethe same way-but in fact, these SKUsoften have very different levels of de-mand. By differentiating SKUs accord-ing to their actual demand patterns,you can reduce inventories on someSKUs and increase them on others-thereby improving your profitability forthe entire line.

Differentiating SKUs can also helpyou rethink your sourcing strategy.Instead of producing all the SKUs for aproduct line at a single location, eitheroffshore at low cost or close to marketat a higher cost, you can typically dobetter by going for a mixed allocation.That way, you can meet the demandsof retailers while controlling costs andinventory.

The Inventory DilemmaTo illustrate, let's consider the inven-tory problems of a hypothetical com-pany called jeansco. In the 1980s, thisblue-jeans manufacturer offered about1,000 different SKUs-a dozen styles ofjeans spread across a few dozen sizes,with total annual sales of 20 millionpairs. Each season,Jeansco built up itsinventories in preparation for big ship-ments to retailers. The inventories wereenormous just before the shipmentdate, butthe risk was small because allof those jeans matched actual ordersretailers had submitted several monthsbefore. Inventory, in fact, was just ameans of spreading out the demand sofactories could achieve a steady, effi-cient fiow of output. For jeansco, theonly cost of i nventory lay in the worki ngcapital tied up there and in the minorexpense ofthe warehouse. The retailers

bore the major cost of inventory-therisk that sales would prove disappoint-ing and the jeans would have to bemarked down below cost.

Then in the 1990s, partly to minimizethis risk, most of Jeansco's retailers beganto adopt lean retailing practices. Theyshifted most of their ongoing invento-ries - and risk - back to jeansco by keep-ing on-site inventories low and placingweekly replenishment orders. Since thelead time for manufacturing jeans wasseveral weeks, jeansco could no longer

annual sales have risen to 90 millionpairs, average sales per SKU have fallenfrom 20,000 units to just 3,000 - orapproximately 60 sales per SKU a week,much lower than the 1980 average of400. And that's just an average. PopularSKUs register hundreds or even thou-sands of sales per week, but less pop-ular, highly differentiated items maysell only ten across all retail stores. Thesmallerthe volume of sales for any indi-vidual SKU, the more those sales tendto vary each week because there is so

By fine-tuning inventories according toSKU-level demand, a manufacturer canincrease profits and reduce inventory risks.

make to order; it now had to predictthe weekly demand for jeans and setproduction schedules accordingly. Andeven if Jeansco got the average weeklydemand right, it also had to take intoaccount those weeks with unusuallylarge orders. To ensure that it could fillthose orders and keep its retail cus-tomers happy, Jeansco had to estimatethe weekly variability in demand andhold a safety stock of finished goods ininventory.

That's difficult enough, but productproliferation only made things worse. Inthe 1980s, most of Jeansco's i,ooo SKUsgarnered fairly high sales. Big volumetended to smooth out the inevitablepeaks and valleys of demand. Thatmeant the composite weekly demandwas fairly predictable and variabilitywasn't so great, so the safety stock heldin inventories was relatively small.

Today, Jeansco manufactures farmore styles and sizes than before- itnow offers 30,000 SKUs. And while total

Frederick H. Abernathy is a professor of engineering and John T Dunlop is a professor

of economics at Harvard University in Cambridge, Massachusetts. Janice H. Hammond

is a professor of technology and operations management at Harvard Business School

in Boston, and David Weil is a professor of economics at Boston University All are affil-

iated with the Harvard Center for Textile and Apparel Research in Boston. They are the

authors 0/A Stitch in Time: Lean Retailing and the Transformation of Manufactur-ing (Oxford University Press, 1999). For more information on the research supporting this

article, including the simulations, go to vi/ww.hctar.org.

much less demand to pool together. Thatmeans Jeansco has to stock a lot morethan ten pairs of those slow-selling jeansto meet sudden upsurges in demand-or risk angering important customerswith stockouts. For the same overalllevel of sales, the company now has tohold a much bigger overall inventory.

And what happens to Jeansco whencertain styles go out of fashion? Retail-ers stop placing replenishment ordersand all those multiple SKUs in inven-tory have to go to discounters, eatingaway most of jeansco's profit. To takean example from the real world, look atwhat happened recently when demandfor athletic shoes suddenly dropped asconsumers moved to brown shoes. Nikeand its competitors had to take a hugefinancial hitto dispose of their bulginginventories. In the absence of lean retail-ing and product proliferation, thosetosses would have been far smaller.

Product proliferation has transformedretail categories far beyond apparel,from office products to pasta. And thetrend isn't going away, despite the fer-vent wishes of many manufacturers,who complain about erratic orders fromretailers. We recently met with an exec-utive who told us business was good,except that low-volume items werecausing him fits because retailers keptasking for greatly varying quantities.

170 HARVARD BUSINESS REVIEW November-December 2OCX)

Control Your Inventory in a World of Lean Retai l ing • TOOL K I T

T H E I M P O R T A N C E O F S K U - L E V E L A N A L Y S I S

The inventory demand for SKUs within the sameproduct line can vary significantly. These chartsshow the weekly demand across more than a thou-sand retailoutletsfor one style of men's blazer intwo different sizes. The top chart shows the demandfor 46-regular, which is one ofthe most popularsizes. The bottom does the same for 43-regular, amuch slower-selling size. To highlight the variationin demand, we've expressed the sales numbers inunits of weekly demand. While 46-regular sells alot more than 43-reguiar, those sales are relatively

stable-peak sales are only about twice the weeklyaverage. Sales for 3-regular varya lotmore-upto four times the i\eekly average. As a result, themanufacturer has tu keep a much bigger inventoryof 43-regular, relative to average weekly sales, thanit does for 46 regular. If we were looking at43-long,the inventory would be bigger still. Although thistype of analysis can be done on a simple spread-sheet, manufacturers generally ignore this variationand assign the same inventory targets for all SKUsin the product line.

Blazer, Size 46-Regular

units of 4weekly

demand • average weekly demand

iiMllllillllliill.nweek 4 6 8 w 12 14 i6 i8 20 22 24

Blazer, Size 43-Regular

units of 4weekly

demand

week 4

I I average weekly demand

• • l l l - l i14

HARVARD BUSINESS REVIEW November-December 2OOO

• I l l II8 20 22 24

171

TOOL KIT • Control Your Inventory in a World of Lean Retailing

" I f I could just smooth out thatdemand," he said, "I'd be fine." Evennow, most manufacturers don't thinkthis is their problem to address. Becauseinventory costs are often hard to mea-sure, losses like Nike's are usually dis-missed as special, unpredictable charges.But the long-term negative effect onprofit can be substantial.

Rethinking Your ProductCategoriesManufacturers generally classify prod-ucts in terms of broad product lines,developing a single marketing strategyand production plan for each I i ne. Thatmakes sense for marketing, but it's amistake for production. Different SKUswithin a product line can have very dif-ferent inventory needs.

Take, for example, a large Americanmanufacturer of men's blazers. As partof our research into lean retailing, wetracked the demand for different sizesof a blue blazer Far from a trendy fash-

ion item, the blue blazer is a staple ofthe wardrobes of millions of men. Butfrom the perspective of actual consumerbuying patterns, a blazer in an atypicalsize actually has more in common with

even though in absolute terms it willhold much moreof46-regular. But mostmanufacturers, including this one, tendto assign the same inventory policy forall products in a product line.

While the offshore option remains the mostdesirable for the lowest-variation SKUs, productproliferation raises the value ofthe optionto produce closer to the market.

a fashion-driven product than with thesame style jacket in a popular size. Forexample, sales for 46-regular, one ofthemost popular sizes, vary only by twicethe average weekly demand, while salesfor43-regularvaryas much asfourtimesthe average demand. A rare size, such as43-long, would vary even more. {See theexhibit "The Importance of 5KU-LevelAnalysis.") To satisfy retail customers,the manufacturer must hold a propor-tionately iarger inventory of 43-regular,

Byfine-tuning inventories accordingto SKU-level demand, a manufacturercan increase profits and reduce inven-tory risks. To demonstrate that improve-ment, we ran a computer simulationthat tests various inventory policies forthree groups of SKUs in the same prod-uct l ine-one group with low variancein demand, another with medium vari-ance, and the third with high variance.(See the exhibit "A Better Way to Man-age Inventory.")

A BETTER WAY TO MA N A G E INVENTORY

This table shows the effects of dif-ferent inventory policies on a set ofthree SKUs within a product line.The first case focuses on achievingvery high order fulfillment for allproducts to satisfy demanding

retailers, but atthe cost of highinventories. The second focuses onmeeting demand for high-volumeproducts, and the third seeks tomaximize profits by balancing thecosts of stockouts and inventory.

In each ofthese cases,a singleinventory policy is set for all threeSKU groups. The fourth case setsinventory policies appropriate foreach SKU, maximizing profits whilereducing inventory risks.

1. Minimize stockouts(Single inventory policy)

Sales

ÍI.761

Productioncosts

$1.198

Inventorycosts

$79

Average order-fulfillment ratio

97%

Totalinventory Profitwcski of demand

18 $485

2. Minimize inventory costs(Single inventory policy)

3. Balance stockout andinventory costs(Single inventory policy)

4. Maximize profits andreduce inventory risk(SKU-level Inventory policy)

$1,612

$1,739

$1,728

$1,062

$1,158

$1,148

$55

$70

89% $494

95% 16 $512

$66 $515 H

dollar amounts are weekly, m thousands

172 HARVARD BUSINESS REVIEW November-December 2000

TOOL KIT • Control Your Inventory in a World of Lean Retailing

The first test shows a scenario inwhich a manufacturer is most concernedabout keeping its big retail customershappy by maintaining very high orderfulfillment rates. The manufacturer setsa single inventory pol icy to ensure thatits highest variance SKUs have plentyof finished goods on hand-say ninetimes the expected weekly demandfor those SKUs. Following that inven-tory policy, the other two groups ofSKUs in that product line also carryinventory of nine times the expectedweekly demand even though their vari-ation is never more than four times theaverage.

The second test reflects a manufac-turer whose concern is maintaining in-ventories at a level appropriate for itshigh-volume, low-variability SKUs-saythree weeks of demand. That meansmuch lower inventories in general anda savings in working capital and risk. Butthe trade-off is that the manufacturerfrequently runs short on its medium-and especially its high-variability items.That means lost sales and maybe a can-celed contract with a prized customer.

In the third test, the manufacturerfocuses on balancing the costs of stock-outs and inventory by setting a singleinventory policy for all SKUs at sevenweeks. I n the case of blazers, the i nven-toryofthe43-regular is just about right,but there are too many 46-regulars andstockoutsof43'longs.

The better approach,of course, isforthe manufacturer to assign an individ-ual inventory policy for each SKU.Thefourth test optimizes the profit of eachSKU according to the estimated costs ofstocking out versus holding inventory.Inventories for some SKUs go up, whileothers go down, but overall inventoriesfall-And net profits rise.

We know of no manufacturers thathave fully implemented what we pro-pose. Yet lean retailers like Home Depotand Wal-Mart already incorporate someSKU-level analysis in their own inven-tory decision making.Calculating SKU-level variation can bedoneonasimplespreadsheet, so moving toward thistype of inventory policy should be quitefeasible.

S O U R C I N G A T T H E S K U L E V E L

Impact of short Cycle Manufacturing on Profits and Inventory

Profits Total inventory

dollars $550 -per weekin thousands

450

350 ProfitTotal inventory

„ 6

_ 2

weeks offinal demand

0 so 100%

Percent of production in short-cycle facility

This graph shows the effects of sourcingdecisions on profitability and inventoryrisk. It simulates the scenario of a manu-facturer with two factories: an offshoreplant with an n-week lead time and ahigher-cost, short-cycle plant with a two-week lead time. The graph shows howprofits (the solid line) and inventory (thedotted line) vary as the manufacturerdraws on more production capacityfrom the short-cycle plant. At oneextreme, all production is made at theoverseas factory, thereby minimizingproduction costs. At the other extreme,

the short<ycle plant handles every-thing,thereby minimizing inventorycosts. The intermediate values repre-sent a mixof the two facilities, wheremost of the tow-variance SKUs aremade offshore and the high-varianceSKUs are made at the short-cycleplant. A mixed strategy actually hashigher profitability than the ioo% off-shore option as well as substantiallylower inventory costs. The greaterthe valuation of the inventory risk, thecloser the manufacturer will movetoward the 100% short<ycle option.

Rethinking Your SourcingStrategy

SKU-level analysis has big implicationsfor sourcing as well. For a long time,manufacturers focused on direct costswhen they located factories. As a result,many shifted production to developingcountries, where labor costs are low.Lately, partly in reaction to the pres-sures of lean retailing, they've learnedthe importance of delivering certaingoods quickly to the marketplace, sothey've moved some production closerto home. In the 1990s, for example, the

American apparel manufacturers shifteda full third of the industry's sourcingfrom Asia to Mexico and the Caribbean.And finally, U.S. manufacturers haveexperimented with flexible productionlines within a factory that allow for fastchangeover to make hot-selling lines.But all of these sourcing strategies stilltend to treat all SKUs within a productline the same. A better approach wouldbe to move low-volume, high-varianceSKUs close to markets, while produc-ing most high-volume, low-variancegoods offshore where it is most costeffective.

174 HARVARD BUSINESS REVIEW November-December 2000

TOOL KIT • Control Your Inventory in a World of Lean Retailing

To set an optimal sourcing pol icy fora product line,the first step once againis to determine each SKU's variability.Next, arrange the 5KUs into groupswith similar variations in weekly de-mand. Each group will have separateinventory policies, and the allocationamong different plants will depend oncapacity, capabilities, production costs,and lead times for each plant, as wellas profit margins.

To simulate this decision, we tookthe same portfolio of three groups ofSKUs as before. The manufacturer hastwo sourcing options. The offshore facil-ity has low costs but, at n weeks, a longlead time for production. The domestic"short-cycle"factory takes only two weeksto bring products to market, but its directmanufacturingcosts are 20% higher. Theresults for this case appear in the exhibit"Sourcing at the SKU Level,"which showshow assigning different percentagesof total production to the two sourcesaffects profitability and inventory posi-tion ofthe manufacturer.

At one extreme, the manufacturerdecides to minimize direct productioncosts, so it assigns all production off-shore. At the other extreme, it uses onlythe short-cycle line in orderto minimizelead time and inventories. The interme-diate cases represent a mix ofthe twofacilities, where most ofthe high vari-ance SKUs are made at the short-cycleplant, while most of the low-varianceSKUs go offshore.

Our simulation reveals that the mixedstrategyyieldsthe highest profits whilestill reducing exposure to total inven-tory risk.The simulation looks explicitlyat inventory levels as well as profits,which brings to the forefront both theconsiderable risk of inventory obsoles-cence as well as the return on differentsourcing strategies.The higher the val-uation of inventory risk, the more desir-able the short-cycle option becomes. Thesimulation results show that inventoryexposure decreases dramatically as themanufacturer draws more on the short-cycle option. And note that as the num-ber of SKUs increases, so does the de-mand variability for the manufacturer.While the offshore option remains the

P R O D U C T PROLIFERATIONIN THE BOOK INDUSTRY

7HE BOOK INDUSTRY MAY SEEM A WORLD APART frOm Otber

consumer goods. But it illustrates what happens when radical productproliferation-more than i.2 million individual titles In print-combines withextremely high fixed costs for each batch of production. The weekly demandfor an average book, if it could be charted, would vary far beyond anythingdiscussed here. As a result, most manufacturers continue to produce books inbig batches, based largely on advance retail orders. But book retailers havebegun to adopt many features of lean retailing and also some aspects of SKU-level inventory policies. Manufacturers are likely to follow suit.

To minimize their own inventories, most bookstores offer three kinds ofavailabilitytoconsumers. Ahot newbook, such asa novel by a major writer,is likely to be stacked high on display tables. But a similar novel by a first-timeauthor may not be. The store will hold only one or two copies of that book.If it runs out ofthe book, the store can get a copy from the distributor fairlyquickly. And what about the previously best-selling novel that everyone's nowforgotten? The book is still in print, but the bookstore won't bother keepingany copies on hand. If a customer wants a copy, the store requests onedirectly from the distributor or publisher, who delivers the book in a fewweeks. You can see this type of SKU-level differentiation explicitly atAmazon.com. Books are listed as usually shipping within 24 hours, intwo to three days, or in one to two weeks.

As the lean retailing practices of bookstores intensify, distributors andpublishers are likely to come under pressure to hold bigger inventories andimprove their response to orders. What can they do about it? They canrethink their sourcing strategies. We can see the future in booktech.com,an upstart company that uses digital printing technology. Already popularfor customized textbooks and course packets, this printing process is theepitome of flexible manufacturing: it can provide booksellers with rapidreplenishment of small orders, but eliminates the need to carry inventories.As production efficiencies for this process continue to improve-and as bookpublishers work on reducing their own costly inventories-we expect moreand nfiore books to be printed this way. Books from major writers will con-tinue to be printed in large batches, but slower-selling books will be printedon demand. As in other industries,this differentiation strategy will allowpublishers to meet the needs of retailers while managing exposure toinventory risk.

most desirable for the lowest-variationSKUs, product pioliferation raises thevalue ofthe option to produce closer tothe market.

In this light, manufacturers would dowell to look at their product lines as port-folios of distinct goods. In satisfying thedemand of retailers for differentiatedproducts, manufacturers must evaluatethe risk that comes with producing thed ifferent items in thei r offeri ngs- By con-

ducting SKU-level analysis, companiescan understand the true risks and returnsassociated with each item, and managethem accordingly. Ü

Reprint R00601To otdet reprints, see the last page of this issue.

Join HBR's authors and readers in theHBR Forum atwww.hbr.org/forumtodiscuss this article.

176 HARVARD BUSINESS REVIEW November-December 2000

Harvard Business Review Notice of Use Restrictions, May 2009

Harvard Business Review and Harvard Business Publishing Newsletter content on EBSCOhost is licensed for

the private individual use of authorized EBSCOhost users. It is not intended for use as assigned course material

in academic institutions nor as corporate learning or training materials in businesses. Academic licensees may

not use this content in electronic reserves, electronic course packs, persistent linking from syllabi or by any

other means of incorporating the content into course resources. Business licensees may not host this content on

learning management systems or use persistent linking or other means to incorporate the content into learning

management systems. Harvard Business Publishing will be pleased to grant permission to make this content

available through such means. For rates and permission, contact permissions@harvardbusiness.org.