; Coping vs_ith Demand Uncertainty at Sport Obermeyer
Learning Center
Plans & pricing Sign in
Sign Out
Your Federal Quarterly Tax Payments are due April 15th Get Help Now >>

Coping vs_ith Demand Uncertainty at Sport Obermeyer


parka, also known as the trench coat or a raincoat, outdoor sports enthusiasts is one of the essential equipment. Both urban and leisure family, or ordinary weekend picnic lovers, whatever you do in the long-distance hiking and mountain climbing, or professional adventure, climbing, or even seven or eight kilometers of mountain climbing, one for their "all-weather "parka are your must-have.

More Info
  • pg 1
									       Coping vs^ith Demand Uncertainty at Sport Obermeyer
        Longtime industry player Klaus Obermeyer charac-       us to generate multistage, risk-hased produetion
     terizes the skiwear market as extremely fickle. What      schedules.'
     possible use could formal statistical methods have in       To implement the approach described ahove, we bad
     such an unpredictable setting? You'd he surprised. The    to estimate the mean and tbe standard deviation. For
     trick lies in realizing that although demand for each     produets with extensive historical demand data, tbose
     product can be highly uncertain, the distrihution of      parameters ean be estimated using statistieal meth-
     demand follows a discernible pattern.                     ods. However, with only a judgmental forecast avail-
        At Sport Ohermeyer, we found that demand data          ahle, we had to devise a different approach. We started
     followed a normal distribution, whieh is defined hy its   by asking each member of Sport Ohermeyer's huying
     mean [average) and its standard deviation [a measure      committee to provide us with an individual forecast
     of the dispersion, or "width," of the distrihution and    for eaeh product.
     hence of the level of demand uncertainty).                  We treated tbe average of the committee members'
        The graph "Prohahle Sales of the Pandora Parka"        forecasts as tbe mean of the demand distrihution. We
     shows a forecast distrihution based on the demand         estimated the standard deviation for eaeh style to be
     predietions of the huying committee. The area under       twice the standard deviation of the buying commit-
     the eurve hetween two points is equal to or greater       tee's forecasts. We decided to scale by a factor of two
     than the probability of demand                                               beeause the average standard devia-
     falling between those points. (For                                           tion of actual forecasting errors in
     example, the shaded area repre-                                              preceding seasons was twice that of
     sents the probahility that demand                                            tbe buying committee's foreeasts.
     exceeds 1,28.S units.) If Sport Ober-                                           We believed that foreeasts would
     meyer were to have only one op-                                              tend to be more accurate for those
     portunity to produce Pandora par-                                            styles for which the buying com-
     kas, we would use this eurve in                                              mittee members bad similar fore-
     the following manner to find the                                             casts-that is, those whose fore-
     produetion quantity that maxi-                                               casts had a low standard deviation.
     mizes expected profitability by bal-                                         Tbis hypotbesis was eonfirmed
     aneing the risks of overproduction                                           with actual data from the 1992-
     and underproduction.                                                          1993 season. The close fit hetween
                                                                                  actual and predicted forecasting er-
        For the Pandora parka, Sport
                                                                                  rors gave us a solid basis for deter-
     Obermeyer earns $14.50 in mar-
                                                               mining wbicb products were safe to produce before ad-
     ginal profit for each unit sold and loses S.S.OO for
                                                               ditional sales data became available and which were
     eaeh unit produeed and not sold. The company should
                                                               not. Using this information along with detailed data
     keep producing parkas as long as it expects the gain
                                                               about minimum lot sizes and other production con-
     from eaeh parka to exeeed the loss. Expected profits
                                                               straints, we formulated an appropriate risk-based pro-
     are maximized hy producing up to the point where the
                                                               duction sequenee for Sport Obermeyer.
     expected marginal gain from producing a parka is
     roughly equal to the expected marginal loss from pro-       [ust as quick-response and just-in-time programs
     ducing that parka. For the Pandora, that occurs when      cannot realize tbeir full potential without correspond-
     tbe company produces 1,285 parkas, hecause the ex-        ing changes in planning systems, neitber should those
     peeted gain from producing the 1,285th parka is ap-       changes in analytical approach exist in isolation. Im-
     proximately equal to the expected loss from producing     provements in supply ehain speed and flexihility are
     that parka. That is, the probability of selling tbe       essential to achieving tbe full potential of an accurate
     I,285tb parka (25,7%) multiplied hy the profit if the     response program.^
     company sells that parka IS14.5O| is roughly equal to
     the probahility of not selling it (74.3%) multiplied hy   1, Fiir a descriptiiin of the mode!, see Marshall L, Fisher and Ananth
                                                               Raman, "Thu ValuL; of Quick Response for Supplying Fashion Prod-
     what tbe eompany loses ii it makes it and cannot sell     ucts; Analysis and AppliciStion," Departmt-nt of Operations and Infor-
     it ($5.00),                                               mation Management Working Paper No. 92-10-03, The Wharton
                                                               School, University of Pennsylvania, 1992,
       This analysis illustrates two critical components of    2. For a description of such .supply chain changes, see Janice H. Ham-
     an accurate response program: assessing a prohability     mond, "Quick Response in Retail/Manufacturing Channels," in
     distribution for demand, and estimating the eosts of      CJohahzaiion. Tecbnoiag.y, and Competition: The Fufnon of Comput-
                                                               ers and Telecomnninicntions in the 1990s, ed. Stephen P. Bradley, Icr-
     stoekouts and markdowns. We have embedded tbis            ry A, Hausman, and Richard L, Nolan (Boston: Harvard Business
     basic logic into a sopbisticated algorithm that allows    School Press, 199,1), p. 18S,

90                                                                                    HARVARD BUSINESS REVIEW            May-|une 1994

To top