The causes and consequences of disrupted startups of new product and production processes are examined in relation to examples drawn from several, diverse industries. It is demonstrated that inappropriate management actions can often precipitate large deviations from expected patterns of productivity increases during startups, resulting in important short & long-run productivity losses. Based upon the discussion, several guidelines for effective startup management are suggested.
…Because of an intense concentration on mastering process variables and developing standard operating procedures, changes in product specifications and production factors may prove very disruptive during such startups.
An example of the potential consequences of product specification and mix changes is provided by the startup of a new temper mill in an integrated steel-manufacturing firm. Previous studies in this company had shown that startups typically conform to the startup model until the steady-state phase is reached.2 However, after a normal beginning, the temper-mill startup deviated from the usual pattern. As can be seen in Figure 1, it followed a log linear trend through the first 7 months of production, during which nearly 2 million “base boxes” of steel were produced (a base box is roughly 200 feet2). In the eighth month, the productivity index drops dramatically to some 60% of the previous level and then begins to increase slowly and irregularly over the next 20 months of manufacture (at which point this study ended).
This interruption of the startup was precipitated by a change in the mix of product being rolled on the mill. The first 7 months of production was limited to “hot rolled steel”. The 8h month marked the introduction of “cold rolled steel”, which required sufficiently different operating settings and procedures to confuse the entire process adaptation effort. The change resulted in a considerable disruption of the learning curve; following the interruption, production efficiency dropped on both product types and the overall productivity remained below previously attained levels for 8 months (and over 1 million base boxes of output).
Another example of the effects of changing product mix is provided by the labor-intensive production of wearing apparel. As in the steel industry, research has shown that startups of new “styles” or models of wearing apparel conform to the learning curve model under normal production procedures6 . In this case, however, normal procedures were not observed. 3 new styles of a basic type of apparel were introduced at different times on the same production line, whereas usually each processing line specialized in the production of only one style.
The consequences of varying the product mix in this way are illustrated in Figure 2, where the production history of the startup is summarized in relation to monthly measurements of the average unit cost index. The initial data points have been fitted to a least squares line for discussion purposes; the variations of the monthly data points about this line show the effects of the first and second style introductions. The first 3 months of the startup were devoted to the production of a single style, and the apparent trend of the 3 points was well below the regression line. When the second style was added in the 4th month, the data jumped above the line, remained there for 3 months, fell dramatically, and then showed a downward trend for the last 5 months of the regression line. The end of the line marked the introduction of the third garment style. The reaction was dramatic—costs drifted upward for 5 months until they ultimately doubled. The next 8 months of alternating production of all 3 styles was marked by a pronounced relearning phase that finally reduced costs to their previous level. The result: 15 months and 700,000 units of “high-cost” production following the introduction of the third style.
One may question the wisdom of changing product mix during this startup. Maintaining the policy of production specialization—even if it meant the creation of 3 smaller lines—could have been a more rational choice for the firm. Of course, the production managers had not expected such a violent reaction since the changes in style design and worker tasks seemed minor to them and each style was run in batches for several days, not intermingled chaotically.
Variance reduction/VoI …Variations in another factor of production—raw material—can also disrupt the learning process in some industries. In extreme cases raw material changes can perturb a startup so greatly that systematic analysis of the learning phenomenon becomes difficult. This appears to be true in food-processing industries; our examinations of food-processing equipment startups have indicated that the extreme variance in the raw produce attributable to seasonal, geographical, and other variables affect productivity to such a degree that one cannot define a true learning curve. We have also observed definite, though less extreme, effects of raw material variations on the startups of highly mechanized processes in other industries. The steel industry is a good example. In recognition of the problem some plants actually pre-select the material input to new processes during the startup period—essentially spoon feeding the baby in its infancy.
[Additional examples from bulb manufacturing, automated process, & steel omitted.]