As noted previously, analyses of organizational learning that use cumulative output as the measure of organizational experience have been found at times to contain large errors. The L-1011 TriStar produced by Lockheed throughout the 1970s and early 1980s is an example of a production program for which predictions based on the classic learning curve formulation were not realized.
As can be seen from Table 3.1, the number of L-1011 TriStars produced by Lockheed each year varied considerably. Applying the classic learning-curve formulation to the TriStar program led the firm to predict that unit costs would fall below price around the time Lockheed produced the fiftieth plane (Fandel, 1974). Although Reinhardt (1973) criticized Lockheed’s analysis of the L-1011 program for omitting the opportunity cost of nonrecurring expenses for develop- ing technology and facilities for the L-1011, he used the conventional cumulative output measure in his forecast of recurring production costs. Reinhardt arrived at a conclusion similar to Lockheed’s: production costs would be below price at about the fiftieth plane. This prediction was borne out: Lockheed reported in 1975 that unit costs were less than the price at which each plane was being sold (Sansweet, 1975).
Table 3.1 Lockheed’s production of the L-1011 TriStar
Cuts in production occurred in late 1975 (see Table 3.1). Costs rose to exceed price and appeared to remain above price for the rest of the production program (Harris Jr, 1982; “Lockheed loses hope,” 1980). In 1982, the L-1011 was sold for $50 to $60 million per plane. Converted to 1975 dollars, this corresponds to $29 to $35 million. By contrast, each plane sold for $20 million in 1975 before the cuts in production occurred. Thus, the unit cost of producing each plane was less than $20 million in real terms in 1975 but greater than $29 million in real terms in 1982! Clearly these data do not follow a learning curve because costs rose rather than fell with increases in cumu- lative output. In 1980, the Chairman of Lockheed, Ray Anderson, stated that the L-1011 would never break even if all costs incurred since 1968 were taken into account but that the company hoped to reach a point where each plane was selling for at least its current production costs (Hill, 1980). Apparently, this never happened. In December 1981, Lockheed announced that it was phasing out production of the L-1011 (Harris Jr, 1981b), making it the last commercial aircraft Lockheed produced.
In the classic learning-curve model, unit costs decrease as a function of increas- ing cumulative output. Thus, according to the classic model, unit costs should have continued to fall because cumulative output of the L-1011 continued to rise. Instead, unit costs rose rather than fell with increasing experience.
Many factors contributed to the increasing costs Lockheed experienced. The L-1011 program was plagued by shortages of personnel and parts (“TriStar’s trail,” 1980). For example, the L-1011 production line was shut down when the company that produced its engines, Rolls Royce, went into receivership (Harris Jr, 1981c). This shutdown delayed the delivery of the first L-1011 aircraft. During this delay in the introduction of the L-1011, McDonnell Douglas’ DC-10 made significant gains in the market (Harris Jr, 1981c).
Although personnel and parts shortages hurt every aircraft manufacturer, Lockheed appeared to suffer more than the others because of its attempt to increase production so dramatically in the late 1970s (“TriStar’s trail,” 1980). Lockheed went from employing 14,000 personnel to employing 25,000 in a 2-year period in California. In order to accomplish this dramatic increase in personnel, Lockheed hired many workers who had no previous experience in aircraft construction and who had not finished high school (“TriStar’s trail,” 1980). Business Week concluded that these inexperienced workers hurt Lockheed’s learning curve:
Inevitably, green workers have wreaked havoc with the TriStar’s “learning curve,” dramati- cally boosting the man-hours required to build each airplane (“TriStar’s Trail,” 1980, p. 90).
Indeed, Lockheed executives spent a significant amount of their time “fighting fires” on the production line (“TriStar’s trail,” 1980). There was a 6-week strike in 1977 and a threat of one in 1980 (“TriStar’s trail,” 1980). Further, Lockheed’s build- up in production came after Boeing had already purchased a storehouse of materials. Lockheed had to pay a premium for materials it needed (“TriStar’s trail,” 1980).
Another factor that may have favored a competitor of Lockheed’s, McDonnell Douglas, is its shared production line between the DC-10 (a commercial product) and the KC-10 military tanker (Harris Jr, 1981c). Thus, the DC-10 had a larger experience base from which to learn and improve than the L-1011. McDonnell Douglas could transfer knowledge from the KC-10 tanker to the DC-10 and thereby reduce the unit cost of production of the DC-10. Further, production of the DC-10 was not characterized by the “roller coaster” production that characterized produc- tion of the L-1011 (“TriStar’s trail,” 1980).
In addition to these factors that seemed to disadvantage Lockheed in particular, other conditions hurt all firms producing wide-body jets. Deregulation (Harris Jr, 1981c) and high fuel prices (Harris Jr, 1981b) favored smaller aircraft. The market for wide-body jets, such as the L-1011, was not as large as had been anticipated.
The classic model that assumes that knowledge is cumulative is too simplistic to capture the dynamics of organizational learning in firms such as Lockheed. The L-1011 is regarded as a technological—but not a financial—success (Harris Jr, 1981a). The L-1011 data are more consistent with a model that assumes knowledge depreciation than with one that assumes knowledge persistence.
Subsequent empirical work by Benkard (2000) confirmed that depreciation occurred in the L-1011 production program. Using newly available data he obtained from Lockheed, Benkard demonstrated very convincingly that organizational for- getting explained the upturn in labor requirements observed in the Lockheed data. Given the large losses Lockheed sustained on the L-1011 program, averaging more than $15 million per year over the decade ending in 1980 (“TriStar’s trail,” 1980), improved strategies for estimating learning rates are clearly of great importance.
The 3-year delay Boeing experienced in launching its 787 Dreamliner is reminis- cent of Lockheed’s problems with the L-1011 TriStar. Boeing began assembly of the 787 in 2007 and planned the first delivery for mid 2008. Many problems arose that caused the first delivery to be delayed over 3 years (Kesmodel & Michaels, 2011). Boeing outsourced many of the Dreamliner’s components. When assembly began at Boeing’s plant, mechanics discovered boxes from suppliers with thousands of parts that should have been installed by the suppliers. Machinists went on strike for 2 months in 2008. Boeing discovered a structural flaw in 2009 that delayed test flights. In 2010, Boeing identified problems with Rolls-Royce engines and experi- enced a fire on a test flight, which required an emergency landing. After spending billions correcting problems in its supply chain and compensating customers for delivery delays, Boeing finally delivered its first Dreamliner in September 2011.
Boeing has been able to produce about two jets a month. It aims to increase pro- duction to ten aircraft a month by 2013. Analysts predict that Boeing will reach its ten-a-month target a year late and that Boeing will lose money on the first 1,000 Dreamliners (Kesmodel & Michaels, 2011). Thus, similar to Lockheed’s L-1011, analysts argue that Boeings’ Dreamliner will not be profitable for many years.
Source: Argote Linda (2013), Organizational Learning: Creating, Retaining and Transferring Knowledge, Springer; 2nd ed. 2013 edition.