Classification of Bottlenecks

The review of literature with respect to definitions of “bottleneck” suggests that a common and unambiguous understanding has not evolved. Partly, this might

4Goldratt & Cox (2004) have chosen their wording very carefully. The reason why they chose to make the existence of bottlenecks a question of capacity lies in the fact that they distinguish bottlenecks from capacity constrained resources, the latter of which are delaying or disrupt-ing material flow not necessarily due to their internal capacity restrictions but to any other possible reason. This distinction is not made in this document. A more elaborate discussion of types of bottlenecks will follow in Section 2.3.3.

be explained by the differences between the various domains where bottlenecks can play an important role. Nevertheless, the lack of a unifying taxonomy poses a problem for the researcher who attempts to make system level comparisons or tries to adopt useful concepts from one system for application in another. Possi-bly, a classification of bottlenecks might contribute to a common taxonomy and remove ambiguity. Below the development of a classification of bottlenecks is attempted. Such a classification will facilitate bottleneck management as it con-tributes to a more precise understanding of the nature of bottlenecks emerging as well as of possible remedies.

A first rough distinction between different types of bottlenecks can be found in the tangibility of bottlenecks: Bottlenecks can be intangible or tangible.

Intangible bottlenecks are, for instance, processes that inhibit higher system throughput. Processes can impede the system from achieving higher through-put because they are badly designed or because the process is simply cumber-some and takes a lot of time (e.g., EPA approval for new drugs in the US).

Tangibleare those bottlenecks which impede higher system throughput due to physical limitations. Furthermore, system elements that are tangible bottlenecks can be either active or passive. Activetangible (physical) bottlenecks are those elements which can influence system throughput by their own actions and be-haviour (either deliberately or unintentionally). Examples are workers of a pro-duction station (on a high level of detail) or an entire factory (on a lower level of detail). Passivetangible (physical) bottlenecks, on the contrary, are not able to change system throughput by themselves since they do not exhibit will or power to do so. Examples are machines of a production station reaching their physical limitations and streets that slow down transportation due to high traffic density.

Another way to categorize bottlenecks is by origin: organizational bottle-necks, physical bottlebottle-necks, and operational bottlenecks.Organizational bottle-necks refer to situations where the root cause of constrained throughput can be found in processes, organizational directives, or established procedures. Exam-ples of organizational bottlenecks in a factory setting are maintenance processes that require significant downtime of a machine, large buffers, order release rules that increase WIP, and ineffective quality assurance processes leading to delays

or low yield. Put differently, organizational bottlenecks refer to the question

“How things are planned”. Physical bottlenecks refer to the physical capa-bilities or limitations of a system’s element. Physical bottlenecks refer to the physical capability of a resource (e.g., a machine or a worker) or limitations due to the general physical setup of production facilities. Put in different terms, physical bottlenecks refer to the question “What can theoretically be done”. So far, this categorization into organizational and physical bottlenecks at first sight appears equivalent to the previous categorization into intangible and tangible bottlenecks. The third category, operational bottlenecks, demonstrates the dif-ference, though.Operationalbottlenecks refer the actual handling of production assets and parts. Examples are careless use of tools and machines that reduces yield or requires frequent reworking, deviation from management directives, de-struction of finished products due to careless transportation and badly scheduled breaks (e.g., lunch breaks) at resources with high utilization. Put differently, op-erational bottlenecks refer to the question “How things are being done”. Hence, this classification addresses “how things are planned”, what physical limitations are met, and how the work is being conducted. However, it is important to note that the root cause of a bottleneck in the material flow system is not necessarily so obvious that one could easiliy identify the bottleneck as, say, organizational bottleneck. The bottlenecks would materialize in some element of the system that passes on or processes material. The root cause of the problem, as in the case of organizational bottlenecks, may still lie somewhere else in the firm or even in another organization. Liu (2011, p. 39) proposes another classifica-tion of bottleneck origins: bottleneck resources and logical bottlenecks. By bottleneck resources, he refers to organizational and physical resources such as processes and facilities; by logical bottlenecks, he refers to schedules and functions.⁵ It seems that these categories are well covered by the more sophisti-cated distinction between organizational, physical, and operational bottlenecks as proposed above.

For the management of bottlenecks, it is important to know whether effec-tive measures lie within or beyond one’s reach. Therefore, a distinction can

5Unfortunately, the examples given by Liu (2011) are rather vague.

be made between bottlenecks that lie inside and those that lie outside manage-ment’s reach. Borrowing from psychology, this parameter will be referred to asLocus of Control(LoC, Rotter 1966) . If a bottleneck has an external LoC, which may be the case if the bottleneck has its root cause outside the boundaries of the organization, then management might be able to influence throughput at the bottleneck through negotiation and “politics”, yet it is unable to directly apply technical or organizatiomal measures that elevate system throughput. In case of an internal LoC, the bottleneck lies within management’s reach and ef-fective measures can be directly applied. As an example, in 2010 and 2011 many car manufacturers suffered from a shortage of components from their suppliers that included microchips (cf. Beer 2011). The bottleneck were the production facilities of semiconductor producers. During the 2008/2009 crisis, semicon-ductor producers cut back production capacity which they were then unable to ramp up fast enough when some industries experienced resurgence in 2010.

Furthermore, the production capacity available was largely used to supply im-portant customers from industries such as consumer electronics: both margins and volumes there were considerably higher than in automotive. The reaction of automobile manufacturers was both individual and joint efforts in persuading semiconductor producers to dedicate more of their production capacity to chips that were needed for automobile industry supply. Because neither individual automobile OEMs nor the automotive sector altogether represent a major share of customers, they were unable to exert power over semiconductor producers.

Instead, high-ranked managers of semiconductor producers were invited for ne-gotiation and “event days” where automobile OEMs tried to influence capacity dedication of semiconductor producers in their favour.

Another fundamental difference is whether a bottleneck emerges “somewhere”

or isplannedby design. There are no material flow systems without bottleneck;

if there were, system throughput would be unlimited. Engineers planning a material flow system can make deliberate decisions as to what element of the system shall be the bottleneck. In a factory setting, for instance, firms may choose the station as a bottleneck that involves the most expensive equipment and machinery for reasons of depreciation. In a rather static environment where

little changes in the factory layout occur, firms may also choose the last process step as the bottleneck (Goldratt & Fox 1986). If a bottleneck is deliberately chosen, this will allow taking effective measures to protect throughput at the bottleneck. In contrast, if the bottleneck emerges in random, unforeseen places, then throughput may suffer since the bottleneck may not be identified as such (or only with some delay) and effective measures to protect throughput are thus difficult or impossible to apply. “Randomly” appearing (i.e., floating) bottle-necks are a typical phenomenon in multi-product manufacturing environments and are difficult to detect with certainty as the location changes with the product mix (Nakata et al. 1999, Hopp & Spearman 2008, p. 486). Firms may therefore choose to eliminate the possibility of bottlenecks in some stations where adding capacity is cheap, just to reduce the number of possible bottleneck locations if identification proves challenging. From the perspective of a powerful OEM, it might be beneficial for the OEM to design the supply network such that the OEM’s production facilities represent the bottleneck and thus enjoy high uti-lization whereas suppliers are required to provide some more excess capacity.

Quite apart from these considerations, a material flow system can be purposely designed such that it will not be able to meet market demand because the produc-tion output is intended to be scarce and valuable and not widely available. This often concerns limited editions of luxury products, but more mundane products can be subject to such considerations, too. Oil production, for example, had been curbed in the past to establish higher price levels and similar measures were discussed by OPEC member states after a sharp drop of oil prices in the second half of 2014 (Lawler et al. 2014). While decisions to lower production output were made deliberately despite higher production capacity available, the same principle can be applied to factory design.

Another distinction can be made in terms of necessity– between avoidable and unavoidable bottlenecks. Bottlenecks appear to be unavoidable if they result from demand (that could not be reasonably anticipated) exceeding supply capac-ity, so that physical limitations become binding. If, however, the emergence of a bottleneck is due to sloppy preparation or other operational problems, or pos-sibly even due to not well thought-through organizational policies, it seems to

have been largely avoidable. Companies could make such a distinction in order to identify levers for improvement.

Bottlenecks differ in the duration of their existence. The duration of their ex-istence can be determined by changes in product mix, by demand fluctuation, by changes in production capacity, by competitors entering or leaving the market, and by a variety of other incidents that affect the firm’s ability to meet demand.

Lawrence & Buss (1995) assert that long-term bottlenecks cannot exist: “either work will increase without bound or there will be sufficient loss of business to reduce demand rate below capacity” (p. 342).⁶ Accordingly, they suggest bottlenecks can only exist in the short-run and these can be managed through the use of appropriate shop-floor control techniques. The authors ignore two important aspects, though. First, everymaterial flow system is constrained by a bottleneck as otherwise throughput would be infinite. The bottleneck can be internal or external to the system, but it does exist and does limit the system’s throughput. If we accept this premise, then the second point is that bottlenecks can exist by designand hence independently of short-term demand or supply fluctuations. Markets can experience long-term upswings and a factory system that is planned based upon (old) market might be unable to meet full demand.

This would be an example of an unintended(or unplanned) bottleneck by de-sign. Löffler et al. (2002)⁷, for instance, discuss “static bottlenecks” that persist over a full time period as demand continues to exceed supply capability. As explained earlier, however, bottlenecks can also be purposely built-in by design.

Therefore, bottlenecks can be distinguished betweenshort-term,medium-term, andlong-term bottlenecksprovided these categories are useful in the context of the problem and they are assigned more specific time horizons.

In a similar fashion, Nakata et al. (1999) argue in terms of lead time until ap-pearance(“appearance cycles”) of bottlenecks and propose different manage-ment approaches, particularly in terms of time horizon, for long-cycled,

mid-6Probably, Lawrence & Buss (1995) refer to self-regulating market mechanisms that will, in the long-run, match demand and supply. In addition to the arguments that follow it shall be noted that markets often are imperfect and not regulate themselves but also depend on external control and steering mechanisms to maintain their functionality. Therefore, the authors are disagreed with unless further clarification is provided as to the exact meaning of “long-term”.

7As cited in Schultheiss & Kreutzfeldt 2009

cycled,and short-cycled bottlenecks. Long-cycled bottlenecksaccording to the authors are those which emerge due to, for instance, planned changes in prod-uct mix or changes in process design, i.e., changes that are known long⁸before the consequences (i.e., the emergence of the bottleneck) occur.Mid-cycled bot-tlenecks emerge due to unplanned changes at short notice, such as changes in customer orders that arrive after the production process has already been started which leads to delays for the new order. Short-cycled bottlenecksare the third case the authors describe. This case refers to machine breakdowns and similar occurrences that will delay production.

Wanderingorshiftingbottlenecks are a problem widely discussed in the pro-duction planning and control literature (e.g., Roser et al. 2002, 2003, Hopp &

Spearman 2008). Therefore, bottlenecks can be devided by their steadiness.

In dynamic production environments with changing product mix and varying production processes, possibly not one particular station limits throughput but a variety of different stations depending on the specific situation. Such dy-namically shifting bottlenecks pose problems for bottleneck detection meth-ods (Roser et al. 2002). Static bottlenecks, on the other hand, are more likely to emerge in low-dynamic environments. Where production schedules don’t change because of stable demand patterns, bottlenecks are less likely to shift in normal operation mode. Having said that, other stations can emerge as tempo-rary bottlenecks due to process variability such as rework, low yield, or machine breakdowns. Static and shifting bottlenecks pose different challenges for bottle-neck management. Generally, shifting bottlebottle-necks are more difficult to identify, to resolve, and to protect than static bottlenecks unless the root cause (mostly variability) is resolved or mitigated.

Another rather practical distinction can be made based upon options that ex-ist to resolve the bottleneck. Regarding the conception of bottlenecks as a mis-match between demand and supply, it seems straightforward that a bottleneck can be resolved by an increase in capacity. Goldratt & Cox (2004) suggest that an increase in capacity often is neither necessary nor sufficient since the root

8“Long” is a rather subjective term and its meaning depends on context and perception. In this context, “long” means about several days or a week in advance. The context of Nakata et al.

(1999) is semiconductor production.

Parameters Characteristics

Location Internal External

Origin Organizational Physical Operational

Locus of Control Internal External

Intention Planned Unplanned

Necessity Avoidable Unavoidable

Duration Short-term Medium-term Long-term

Appearance Short-cycled Mid-cycled Long-cycled

Steadiness Static Dynamic („wandering“)

Exploitation Options Only through capacity increase Through various options

Financial Implications High Medium Low

Figure 2.3.1.– Morphological Classification of Bottlenecks

cause of the problem tends not to be a lack of capacity of the bottleneck sta-tion but rather how the capacity available is utilized. Nevertheless, it can be that an increase in capacity of the respective station is the only effective mea-sure to increase throughput, especially when other options have already been explored. Therefore, it is suggest to classify bottlenecks further according to the options available for effective throughput improvement into bottlenecks that can be relaxed (i.e., throughput can be increased) by a combination of measures and bottlenecks that can be relaxed only by adding additional capacity.

Bottlenecks can lead to severe financial consequences. On the other hand, they could also be harmless. If, for instance, there is some slack between arrival time of supply and begin of production, or if inventory can buffer delayed arrival of supply, the bottleneck might not even invoke action on part of the focal firm.

If the bottleneck starves the focal firm’s production, however, leading to delayed order fulfillment for customers, and upon arrival of supply the bottleneck shifts to the focal firm which then has to catch up with production schedules, works overtime, stretches maintenance cycles (possibly leading to compromised qual-ity), and frustrated customers turn to competitors, then the financial implications for the focal firm are intense. Many examples are conceivable where the exis-tence of a supply bottleneck falls into either category.

Bottleneck Idle Bottleneck Working

Bottleneck down Bottleneck pausing Bottleneck starving Bottleneck blocked Bottleneck busy (Utiliz. > 80%) Bottleneck normal (Utiliz. < 80%)

Generally, the following states of a bottleneck are conceivable. Generally, a bottleneck can either be working or not working (i.e., idle). If the bottleneck is working, a meaningful distinction can be made between “normal” operations with “normal” utilization (say, less than 80%) and the bottleneck being fully engaged with extremely high utilization. If the bottleneck is idle, it could idle due to a

breakdown, it could be pausing (i.e., it has been consciously stopped by someone), it could be starving, and it could be blocked.

With exception of “normal” operations, each of these states provides a starting point for a more detailed investigation as to why the state is so. Moreover, it can provide the fundament for a typology of reasons for the emergence of bottlenecks. It should be kept in mind that we are already dealing with a subset of bottlenecks, namely unplanned bottlenecks (as opposed to planned bottlenecks; cf. section “Classification of Bottlenecks”). That is, we are dealing with bottlenecks that emerge unintendedly and have not been planned into the system with full consciousness (and possibly good reasons).

(Hier kann ich Beispielhaft die weiteren Möglichkeiten einer Klassifizierung erläutern, wie in der Mindmap dargestellt).

The states of bottleneck as explained above represent theoretical cases. Furthermore, these cases are typical for material flow systems in factories. In the following, it will be examined which of these cases is of practical relevance in real supply networks – and to what extent.

(…)

(Kann ich das eigentlich als Petrinetz darstellen? Bottleneck-Darstellung als Petrinetz sinnvoll?) (Kann ich dies hier auch weiterverwenden für ein Paper „Classification of Bottlenecks“? Generell:

Mail schicken an den Editor vom Journal, das Jan Frick vorgeschlagen hat, und nach Potenzial zur Veröffentlichung fragen. Jan Frick fragen, ob so eine Nachfrage Sinn macht).

Figure 2.3.2.– Conceivable States of a Bottleneck