PARCEL SORTING PRODUCTIVITY
Faculty of economics and business MSc. Technology Management August 2011 Wouter Voortman Student number: 1779907 Supervisor: Dr. G.C. Ruël 2nd supervisor: Drs. J. Drupsteen ABSTRACT
In the global 24-hour parcel delivery business 700 000 permanent post offices and sorting centers are utilized. At depots parcels are received from the network of a parcel delivery company. Subsequently, they go through the inbound sort, were they are sorted to the correct vans, which deliver them to the customer. The focus of this thesis is on increasing productivity at the inbound sort in order to achieve a high on-time performance for this critical part of the organization.
Demand per delivery round and the input mix were both found to be over dispersed, putting the layout under pressure. Furthermore, warehouse employees were found to be unmotivated due to job cuts, management and the reward system; solutions to these problems were proposed. The proposed solution to the layout problem was the introduction of a circular conveyor belt. By means of extensive simulations of the current and future situation, it was established that the circular conveyor belt would increase productivity with 41%.
Index
1. INTRODUCTION ... 1 1.1 Inbound logistics ... 2 1.2 Negative consequences ... 3 1.3 Research focus ... 4 1.4 Research question ... 51.5 Conveyor belt sorting process ... 7
2. METHODOLOGY... 8
2.1 “Hard” factors ... 8
2.2 “Soft” factors ... 11
2.3 Conceptual causal model ... 12
3. EMPIRICAL STUDY... 13
3.1 “Hard” factors ... 13
3.2 “Soft” factors ... 17
3.3 Empirical causal model ... 21
4. SOLUTION ... 22
4.1 “Hard” problem ... 22
4.2 “Soft” problem ... 22
5. SIMULATION ... 26
5.1 Simulation of the current linear layout (S1) ... 26
5.2 Model data ... 28
5.3 Model description ... 29
5.4 Simulation of the proposed circular layout (S2) ... 30
5.5 Results of the simulation ... 31
6. DISCUSSION... 33 6.1 “Hard” part... 33 6.2 “Soft” part ... 35 6.3 In general ... 36 6.4 Advice on implementation ... 38 REFERENCES ... 39 APPENDICES ... 42
1. Introduction
Worldwide more than 10 billion parcels are delivered every day, through a network of about 700.000 permanent post offices and sorting centers (Morlok et al, 2000). The major competitors in the business use a hub-and-spoke configuration to handle the flow of packages efficiently (McWilliams, 2008). The main merits of this methodology are that it requires a relative low amount of routes to connect all nodes, it is easy to connect a new node and it is intuitive to understand and the complex sorting operations are carried out at the (central) hub. The most typical feature of the parcel delivery business is its 24-hour parcel processing cycle and generally has three sorts (Vrgoc and Ceric, 1988). This process is illustrated in figure 1 and is described in the following paragraph. During the afternoon trucks and vans of a parcel delivery company collect parcels throughout the service district of their depot. At the depot, the first sort, or outbound sort, is performed according to whether a parcel needs to be delivered in the service district of the depot or outside. The second stream is transported to the hub. After receiving goods from all depots the central sort is performed at the hub during the night, which essentially sorts the parcels according to the service district where they have to be delivered. In the early morning, the inbound transport moves these parcels to the depots, overseeing the service district. Here the packages are sorted for the third time during the inbound sort, which involves sorting according to the various local delivery routes (Gue, 1999). The last step is the delivery of the parcels to the posted address in the morning. Note that, systems with a few hubs follow a similar procedure, only adding a few possibilities to the outbound sort.
Location Process name Responsibility Service district
Pick up Depot
Depot Outbound sort (1st) Depot Outbound transport Transport
Hub Central sort (2nd) Hub
Inbound transport Transport
Depot Inbound sort (3rd) Depot
Delivery Depot
Service district
Since the delivery (last) process and pick up (first) process share resources (drivers, trucks and vans), there is a dependency between these two steps; resulting in a circular dependency throughout the whole cycle. In order to maintain a 24-hour business cycle, each process has to be carried out before a deadline. A delay in one step of the supply chain will result in a smaller window to perform operations in the next step, in turn increasing risk of delays in that step. Therefore, it is “critical” that each of the operations is performed on time.
The first of all consecutive processes performed at the depots during each 24-hour cycle is the inbound sort. If this process is not performed on time, the following process (delivery) will have a shorter timeframe to perform their work in. This makes it harder for them to deliver on time. Since the negative consequences of missing the deadline by the inbound sort are incurred by the depot itself, this process has extra importance. Furthermore, the inbound sort has more sorting destinations than outbound sort (as there are more delivery rounds than hubs). This makes the inbound sort the most complex sorting process performed at depots. Due to its complexity and the other processes dependency on the performance of the inbound sort, it can be seen as the most important primary operation performed by the depot. Therefore, the subject of this thesis is on time delivery performance of the inbound sort.
This was investigated on the basis of an in-depth case study of a Scandinavian depot of a global transportation and distribution company. Management at this depot was willing to participate fully on the condition of anonymity and will henceforth be referred to as “the depot”.
1.1
Inbound logistics
The connection to the company-wide delivery network is made via two hubs: a “road hub” for non- priority parcels and an “air hub” for priority parcels. However, volume between the subject depot and some neighbouring depots is big enough to have separate connections, saving a detour to the hubs. Line haulers with parcels from these places arrive at the depot during the inbound sorting shift. They do not arrive at the same time because the distance and means of transportation (cargo plane / line hauler) vary among other things (distance, traffic, weather conditions etcetera).
The service district of the subject depot compromises only a small area of Scandinavia. The area is covered by two layers of routes and a third group of specials is handled by a business partner. The first layer are vans for smaller/lighter parcels and the second layer are trucks for heavy/pallet-size parcels and major customers. In total the depot employs about sixty routes for total coverage of their service district.
Upon arrival at the depot, line haulers are backed up against the depot and unloaded by five forklift drivers. Big parcels arrive in the form of pallets and are brought directly to the dock corresponding to their delivery truck. Smaller parcels arrive in cages (roughly: a wheeled square meter base, with two meter tall fences and a door) and are moved to the conveyor belt. Here the cages are opened and the parcels are put on a conveyor belt, along which the sort is performed by ten employees with the help of lifting-aids. On one side heavy and multiple piece consignments are loaded onto pallets, whilst on the other side other parcels are put into cages, each corresponding to a delivery round. When the sort is done the van-drivers collect their cages and the truck-drivers load their pallets. The just described inbound logistics process at the depot is illustrated in figure 2.
Conveyor belt Line haulers
Pallets
Heavy & multiple pieces Cages
The inbound sort is operated by approximately fifteen employees on the busy days of the week (Wednesday to Friday); five forklift drivers and ten on the conveyor belt. The level of automation is low, with five forklifts, a conveyor belt and lifting aids. Therefore, the budget of inbound logistics consists for a significant part of labour cost. However, in 2009 the depot cut the employees’ hours in an effort to save costs; most went from full time jobs to half-time jobs. On top of that, because of regulation from the collective agreement of the transportation sector, the possibility of working extra in the outbound sort was removed. Due to the aftermath of the financial crisis of 2008 many employees had no other option than to accept their losses and were left with a feeling of resentment.
1.2
Negative consequences
Since all the line haulers are scheduled to arrive at least one hour before the hand over to the delivery department, the forklift drivers are ready before the this time with unloading these trucks. On the other hand, the conveyor belt sorting process is still busy at this point. Therefore the forklift drivers tend to help out their colleagues at the conveyor belt near the end of the shift. However, even with this help, the inbound sort has an on time performance of only 30%, according to company data. As a result, the delivery fleet cannot run their schedules properly, having to hire taxis and other substitutes to make up for the lost time. The manager of the pickup and delivery department (PUD) estimates that the overrun costs account for about six percent of their budget, which is much larger than that of inbound logistics.
Additionally, some trucks depart at fixed times; late goods will be left behind only to be delivered the next day. In these cases the company has to refund the shipment and incur a loss of reputation as a reliable distributor. Although this only happens with 0,4% of the total volume, the manager stresses that, it concerns the same trucks every day and as a result the same customers are affected. When a customer does not accept to wait another day, extra transportation has to be arranged to make the delivery anyway. These costs compromise a further one percent of the budget of PUD. The estimated seven percent of budget is relevant because margins in the transportation sector are relatively low (Crew, 2010) and the budget of inbound logistics is significantly smaller than that of PUD.
Moreover, the drivers start working already before the deadline, therefore an early finish will increase the ability of PUD to meet their targets. Subsequently, PUD has an overlap with outbound logistics as well; meaning that unless an overrun is solved by PUD the outbound sort is also affected by bad performance of inbound logistics. The following figure depicts the just described negative consequences of not performing the sort on-time.
Conveyor belt sort is not done on time
From this description it can be concluded that the current performance of the conveyor belt sorting process has serious negative consequences attached to it. De Leeuw would therefore define this as a “real” problem (2002); a situation that should be improved.
1.3
Research focus
From the description of inbound logistics at the depot it becomes apparent that the conveyor belt sorting process is the bottleneck, rather than the fork lift drivers. That is why improvement efforts will be focused on the conveyor belt sorting process of inbound logistics. Furthermore, most of the costs of inbound logistics are incurred via the wages of personnel and ten out of fifteen employees are occupied in the conveyor belt sorting process.
On time delivery is a measures whether work has been done before a deadline. The metric is often used in the parcel delivery business. An industry giant, UPS, saw it as paramount to customer satisfaction (McDaniel and Gates, 1998). According to Little’s law (1961) the required time for an amount of work depends on the throughput (Time = Work / Throughput). Throughput is defined as the average output of a production process. Since labour is the most important and costly resource in the conveyor belt sorting process, throughput was split up into labour productivity and labour capacity. Productivity is defined as the quantity of parcels sorted in a time period divided by the amount of labour hours required (Gaither and Frazier, 1999; Kals, Luttervelt and Moulijn, 1998). Labour capacity is defined as the amount of time periods of labour utilized for doing a sort. Combined, these concepts can be represented by a rectangular cuboid with the axes capacity, productivity and time; whereas the volumes denotes the amount of work (see figure 4).
Assuming that the amount of work stays the same, any future improvement in on time performance (whilst handling the same amount of work) can follow from any or a combination of the three following possibilities. The first option is to increase the time span in which work has to be done (move the deadline). The second option is increasing the capacity and the third option is improving the productivity. In order to focus research activities, only one of the vectors will be investigated in this thesis; the reasoning behind this is now elaborated on.
An improvement in the time span for inbound logistics, will, because of the 24-hour “business cycle” within the parcel industry, be at the cost of the time span of another part of the organization. Making the assumption that transportation times from the hubs to the depot cannot be shortened economically (the distance that has to be bridged stays the same). Then, a longer time for the inbound sort will result in a shorter time to perform the central sort at the hubs or a tighter window for the outbound sort at all the depots in the company network. This effectively moves the problem from a local level to a company-wide level. On the other hand, the time window can be improved by delivering later to the delivery department. This would give the delivery department less time to perform the distribution and as a result, customers will receive their parcels later, which is not acceptable for all customers.
The second option for future improvement of the on time performance of the inbound logistics, results from increasing capacity. In the case of the Scandinavian depot the sorting process is done manually, so increasing capacity will require the company to hire more staff. However, extra capacity acts, just as extra inventory, as a buffer and is “considered to be evil, because it covers up other kinds of wastes and encourages, or allows, wasteful practices” (Nicholas, 1998). With other words, the root problems are not addressed, but this “solution” only treats the symptoms.
The third option is to look for improvements in the area of productivity. When the capacity and amount of work stay the same, a higher productivity will result in the sort being done earlier. This way the problem is not mitigated to other parts of the organization and the solutions will treat the symptoms (De Leeuw, 2002). Therefore, the research focus of this thesis is on increasing labour productivity.
1.4
Research question
The subject of this thesis is the on-time delivery of the inbound logistics sorting process, taking place at depots in a hub-and-spoke configured parcel delivery network. In the previous sections reasons for focusing on the conveyor belt sorting process and labour productivity were stated. Combined this lead to the following research question:
- How can the labour productivity of the conveyor belt sorting process at depots be improved
in order to realise a process capable of delivering on time?
To this research question there are several boundaries. Most importantly, the desired increase in labour productivity should not be at significant costs to other performance metrics. There is for instance no point in a highly productive sort, which is not capable of sorting parcels to the right delivery round (quality of the sort). However, there might be a good reason to invest money for a certain solution. Therefore, any negative impact on other performance metrics following proposed solutions will have to be accounted for.
Mukherjee and Singh (1975) note that productivity is not completely controllable by internal factors, but is also affected by the government and the wider economy. However, it may be clear to the reader that this thesis will not dig deep into law or the wider economy.
this, the depot does not have any influence or complete overview into this centrally organized matter. Therefore, arrival and departure times of line haulers between the depots and the hubs will be considered as a boundary.
1.5
Conveyor belt sorting process
The inbound sort at the depot’s warehouse takes place around a linear conveyor belt. Both sides of the conveyor belt are flanked by a meter wide path, followed by a meter wide area designated for pallets and cages. This allows the employees to walk back and forward besides the conveyor belt to put parcels in the right pallet or cage. Each of the pallets and cages corresponds with a delivery round. The ten employees can be grouped into six workstations according to their activities; see figure 5. The following paragraphs will explain each of the workstations into more detail.
The forklift drivers move cages with conveyable parcels to the feeding workstation (1. Feed), at one end of the conveyor belt. Here two persons unload the parcels from the cages onto the conveyer belt. At the following workstation (2. Scanner) one person scans the barcode on every parcel, to notify its arrival at the hub. The scanner shows whether a parcel needs to be held for customs, in which case the person puts the parcels in a cage for customs. Packages that have a special service (early delivery and or safety related) are pushed from the conveyor belt onto a smaller non-moving conveyor, where personnel from other departments take over. After this initial sort, only parcels remain on the conveyor belt who are delivered by one of the sixty delivery rounds.
At the third workstation (3. Reader) a person decides whether parcels are to be transported by the company’s vans or trucks, depending on their weight and amount. To this extent the person needs to look at every parcels label. Heavy and multiple piece consignments are pushed to one side of the band and light parcels are kept on the other side of the band, parcels that are to be delivered by their partner are left in the middle of the conveyor belt. As a result the following people only need to look to the parcels on ‘their’ side of the conveyor belt.
At the fourth work station (4. Light parcels) two persons, helped by the ‘reader’ take light parcels from the band and put them into cages, each representing one of the 27 delivery rounds serviced by vans. To prevent excessive walking, each of the employees patrols a range. For instance the reader is responsible for the first 5 cages, the first employee of workstation four takes care of cage 6 to 18 and the second, cage 19 to 27. These ranges are not fixed and can change during a shift. The same approach also holds for the fifth work station (5. Heavy and multiple piece). They sort goods onto pallets, each representing one of 20 delivery rounds, serviced by trucks. On the sixth workstation (6. Partner) one person loads their goods onto a partner’s special cages and are delivered to the partner by truck.
1. Feed 2. Scanner
3. Reader 4. Light parcels
5. Heavy and multiple piece 6. Partner
Figure 5: Conveyor belt workstations
Cages
2. Methodology
In this chapter a literature search has been performed to find possible explanations of the occurring problem. Literature was combined with logic and facts to formulate questions and concepts, about how the problem is thought to come about. Next to this, methods for validating these theories are proposed. Finally a causal conceptual model summarizes these hypothesises.
Productivity is the ratio between output and input (Koontz and Weihrich, 2006) and is one of the main indicators of organizational success (Flynn, 2008). Prokopenko (1987) proposed to split up productivity into a hard and a soft part. The soft factors concern the company’s people, work methods and management style, whereas the hard factors compromise the product, plant and equipment (Prokopenko, 1987). This structure was adopted by this thesis, effectively dividing each chapter up into a hard and soft part.
2.1
“Hard” factors
One of the main hard factors determining productivity is the company’s product, because it dictates its production process (Prokopenko, 1987). It is easy to mistaken parcels as the company’s “product”. However, the industry provides a service; the timely delivery of parcels to a specified address. The difference being that the “raw materials” are not delivered by suppliers but by customers. This makes the situation more complex, as a company can exert more control over a supplier than a customer. The company is thus positioned in a “customer-company-customer” relation. In turn, customer demand is described in literature to be volatile (Harrison and van Hoek 2008; Sethi et al., 2005) and through this double relation the company is extra sensitive to customer demand.
Production systems on the other hand, are typically being optimized for one expected value of demand; a stable situation (Ravi and Rosenblatt, 2003; Gattorna, 1998). This contradiction becomes of extra interest when recognizing that each parcel’s routing through the sorting facility is based on its delivery address, which is determined by the customer. Thus, demand for each of the sixty routings (one per delivery round) is established by the customer. Moreover, only after the outbound sort are the amount of parcels and their specific delivery addresses known, resulting in a very short noticing time to possibly adapt the “production” process.
Not surprisingly, Walker and Vatter (1999) found that customer demand has an influence on productivity. Indeed changes in customer demand will require the company to respond (Koste and Malhotra, 1999; Tsourveloudis and Phillis, 1998). Therefore, customer demand is the first issue to be investigated. Q1: To what extent does demand for each delivery round fluctuate on a daily basis?
The combined demand of all delivery rounds, determines the overall input on a given day. This input is however, not necessarily put into the process in an orderly manner. This so called input mix, is ideal, when it is ordered in such a way that the workstations gain a constant balanced flow of work (Lu and Kyokai, 1989). However, due to the fact that the customer acts as a supplier to a parcel delivery company, it is likely that the parcels are not received in an orderly fashion. Moreover, the sheer number of customers, parcels and destinations, will make managing the input into an ideal input mix an impossible task. Therefore, it is likely that the workstations along the conveyor belt face a random input mix. According to Lu and Kyokai (1989) a random input mix is a mix that when put into a production system, will result in an unbalanced mix of work for workstations; figure 6 depicts both concepts.
One can imagine that the random input mix, as depicted above, will lead to more starvations and stoppages than the ideal input mix (Stoppages occurs when a workstation faces too much work). Lu and Kyokai (1989) conclude that the input mix has a significant influence on productivity, resulting in the second subject for investigation. When the ideal situation is seen as a constant distribution in the view of Cox (1966), then there are besides the true random situation (with a variance-to-mean ration of 1) also situations possible with input mix that are under- or over-dispersed. Q2: What type
of distribution does the input mix follow?
The input mix distribution can be determined when the order in which parcels are put in the process is known. For this, the data from the scanner could be employed. However, at the depot no knowledge was available to couple data retrieved from this scan to a parcel delivery round. Therefore, this method was not viable and an alternative had to be found. Since each parcel has a sticker attached to it with a code that holds information about the destination and routing, one can observe to which cage or pellet a parcel is destined. The more constant the frequency at which parcels flow towards a cage is, the more ideal the input mix is.
In order to visually observe what type of distribution the input mix follows, it is assumed that demand per delivery round is equal. Since there are sixty delivery rounds, the ideal input mix would send a parcel to a specific cage/pallet at a constant rate of one per sixty parcels. The chance that two or more consecutive parcels are bound for the same cage is than 1 / 60 = 1.6%. A regular occasion in a shift of a few thousand parcels. Therefore when two consecutive parcels are found both bound for the same delivery round the situation can be called “random”. The chance on a series with three
Work-station A
Work-station B
Ideal input mix
Random input mix
A B B A A A B B B A B A B A B A
equal numbers is 1 / 602 = 0.03%. This is a very seldom occasion and is therefore an indicator of an over dispersed situation. Under dispersion is a situation between the constant and random distribution. This is the case when no series are observed but when the amount of parcels between two parcels with the same destination varies (so are not zero or sixty). Therefore, a simple visual inspection will be used to determine what type of distribution the input mix follows. These results will then be verified by asking staff and management for their obeservations.
The possible fluctuating demand for delivery rounds (Q1) and possible random input mix (Q2) are
both basic input variables when designing a layout (Jaramillo and McKendall, 2009); affecting the choice of both the type and layout of the necessary manufacturing system (Slack et al., 2004). Optimising the layout as a means to improve productivity has received considerable attention in literature (Caron et al., 2000). Indeed, the layout has a direct impact on the operational performance (Benjaafar, 2002), especially on productivity (Raman et al., 2009; Kishel and Gunter Kishel, 2005). Therefore, the effectiveness of the current layout is evaluated. Q3: How effective is the current
layout in dealing with (external) variations (like the possible ‘random input mix’ (Q2) and possible
fluctuating demand for delivery rounds (Q1))?
The objective of a layout is, according to Kishel and Gunter Kishel (2005), to achieve a high utilization rate. This parameter can be measured with time and motion studies. However, the appearance of time clocks has in more than one instance, resulted in protests and personnel leaving work (Montgomery, 1980; Lewin, 1988). Next to the business and personal loss, this would also intervene with the stability of the situation, which makes research results unreliable. Therefore a less personal measure was used; conveyor belt stops. The conveyor belt can be stopped by employees when they face a problem or when they are overwhelmed by the amount of work. When this occurs, the other workstations starve and the amount of production will fall (Lu and Kyokai, 1989). In a perfectly effective layout there is no need to stop the conveyor belt, because the process is capable of handling all
(
external) variations (Ono, 1988). When the layout is not so effective, problems occur, that cause personnel to stop the conveyor belt. Therefore conveyor belt stoppages are assumed to be a good indicator for layout effectiveness.2.2
“Soft” factors
The soft factors refer to personnel, who are “the principle resource and the central factor in productivity improvement” (Prokopenko, 1987). In order to achieve high performance Benson (2008) noted that an employee needs to be driven (or motivated). “Motivation represents those psychological processes that cause the arousal, direction, and persistence of voluntary actions that are goal oriented” (Mitchell, 1982). Motivation in a business environment refers to the drive that individuals feel to perform their assigned tasks (Canzer, 2006). Edwards et al. (2000) found that mechanistically oriented job designs have a negative relationship with satisfaction-related outcomes. This implies that people with mechanistically oriented jobs are less motivated, which in turn implies that the employees will not deliver a high performance. Motivation, the pivotal factor in this line of reasoning was therefore investigated. Thus; Q4: How motivated are employees of
inbound logistics?
Measuring motivation can be done by means of a questionnaire (Hackman and Oldham, 1980). This kind of survey is regularly performed within the company by an independent bureau. This survey covers a wide range of subjects. The employees can answer questions as positive, negative or neutral. Most interesting are the negative answers, because they determine how much the current performance can be improved. Neutral answers are problematic since their true opinion regarding the subject cannot be known; they might agree with certain aspects and disagree with others. Therefore conclusions will be drawn on the percentage of negative answers out of all answers. In order to increase validity of the results, each of the research questions was investigated by selecting several questions from the survey. These questions were selected from the total survey on the basis of relevance with regards to the investigated subject.
The next step is to interpret the obtained data. To this extend normally a control group is employed (Wright and Marshden, 2010). The questionnaire was also held under the office staff of the depot. However, the job characteristics of office staff are significantly different from warehouse staff. Therefore, they cannot function as a real control group, but their answers can give an indication about ratings given by inbound logistics staff. Also, no fitting reference group with similar questions was found in literature. Therefore the ratings given by warehouse staff were interpreted on the basis of their value as compared to an expected value. Since the employees could give three possible answers, the expected value was one third; meaning that more negative answers will be perceived as a negative rating about a subject. The implications of this method will be further discussed in chapter 6. Data from this biannual company questionnaire was also employed to answer the remaining research questions (Q5, Q6 and Q7).
Recall that only one year ago the employees were cut from a full job position with the possibility to work overtime, to a part time job without that possibility. Many had no other option than to accept these changes. This historical event is expected to have affected morale in a negative way, which in turn is associated with lower productivity (Benson, 2008). The possible influence on motivation from the event can be observed by comparing the results from the 2008 survey, of before the events, to the 2010 survey. Giving answer to the following question Q5: How was the motivation of inbound
logistics staff influenced by these historical events?
extrinsic rewards are conceived to improve productivity. Indeed, Hassink and Koning (2009) found that reward systems can reduce absenteeism and Levi (2010) found that performance evaluation systems can provide incentives to work together. Therefore, performance evaluation and reward systems have significant influence on motivation. Thus Q6: How motivated are employees of
inbound logistics by their performance evaluation and reward system?
Management has an important role in a group structure (Shaw, 1981). Therefore Bass (1990) sees that a manager is the primary agent for determining atmosphere, goals, ideology and activities of the group. Benson (2008) confirms that it is part of the management task to motivate subordinates. Popper and Lipshitz (1998) go even further and say that it is the essence of leadership to motivate people to act by non-coercive means. Rungapadiachy (1999) defines leadership as influencing people towards the accomplishment of objectives. These objectives are related to organizational outputs, like productivity. Therefore Q7: How inspired are employees of inbound logistics by their
management?
2.3
Conceptual causal model
In this chapter hypotheses were developed on the basis of facts, logic and literature that could explain the current situation. When these suspected problems are combined the following causal conceptual model can be drawn up; showing how the situation is thought to come about.
Note that, productivity is only one of three factors that could be investigated for achieving the objective of this thesis; improving on time delivery of the inbound sorting process. The choice to focus solely on productivity was motivated in the paragraph “research focus”.
Delivering late Productivity Employee motivation Layout effectiveness Historical events Input mix randomness Demand fluctuations Inspiring reward system Inspiring management
3. Empirical study
In the previous chapter research questions were formulated and methods were proposed to find the answers. This chapter elaborates on the obtained results from the proposed studies, which will be used to formulate answers to the questions. The chapter is concluded with an empirical model, which shows what problems were found to cause a low productivity and ultimately a low on time performance.
3.1
“Hard” factors
The first question was: To what extent does demand for each delivery round fluctuate on a daily basis? This was investigated by means of data mining the computer system of the company for input of the parcel scanners. The index of dispersion (D) was calculated from the available data, by dividing the variance of the population (σ2) by the mean (µ) (Cox, 1966). When this ratio is
equal to zero the variable is considered constant. If this ratio is equal to one the variable is considered random and can be modelled by the Poisson distribution; this is logical, as the variance is equal to one in the Poisson distribution (Montgomery
and Runger, 2006). When the ratio is between zero and one the variable is under-dispersed and follows a binominal distribution (Cox, 1966). Over-dispersion occurs when this ratio is greater than one; in this case the data could be clumped, and thus modelled with a negative binominal distribution.
Drawing on several data sets, the demand, numbers of parcels, per day per delivery round were obtained. The collected data spans a four-week period, unfortunately 10% of the data was missing. Delivery rounds with missing data for more than two days out of the four-week period were deemed inappropriate. This meant that from the 59 delivery rounds, 45 data series could be used as they had zero, one or two data points missing. Combined, the useful data series had 1080 data points of which 15 were missing. The missing data points were filled with the average of their delivery rounds in order to calculate the index of dispersion over the four-week time span.
The graph above shows that the data points have a higher variance than average, because all data points are below the line indicating a fully random distribution (with D=1). Indeed there are no delivery rounds with constant demand (no variance) or under-dispersed demand (variance smaller than average). A few data points are very close to following a random distribution (including the data point that is seemingly on the line). However, all delivery points have a variance greater, than average and are thus over-dispersed. The average index of dispersion data is 2.95, significantly higher than a random situation with D=1. The distribution followed by demand per delivery round is thus a negative binominal distribution and data might be clumped (Cox, 1966). This is not an unexpected outcome, as there probably is a weekly recurring pattern; different numbers of staff are scheduled for each day of the week. Therefore, also the index of dispersion per day of the week was calculated from the same data. This resulted in the following data:
Indeed, the total index of dispersion has come down from 2.95 to 1,81, which means that part of the over dispersion can be explained by the weekly pattern. Also, the adjusted index of dispersion per workstation was calculated; here it was found that demand for heavy and multiple piece consignments was much more susceptible to over dispersion than the other two workstations. A possible explanation for the over dispersion may stem from the fact that many clients sent or receive more than one parcel at a time. Workstation five is getting most of these larger series and is therefore more likely to be influenced by this. The bottom line is however, that no matter how the data is viewed, it shows that demand is over dispersed and thus has a greater presence of variability (statistical dispersion) than would be expected in a Poisson distribution.
The second question was: What type of distribution does the input mix follow? The method of establishing this was by visual inspection of the parcel routing codes. The employees are standing in between the belt and a row of cages, each corresponding with a delivery round. They check the parcel routing code and put parcels in the cage (or pallet) corresponding with this code. Following the fact that there are almost sixty delivery rounds, assuming equal demand, the ideal situation would be if one parcel would be sent to a cage every sixty parcels. A less desirable input mix is when two consecutive parcels are bound for the same cage; then the distribution would be deemed random. In between these two situations is under dispersion; no series will be found but the amount of parcels between parcels bound for the same cage would vary. The situation can be called over dispersed (possibly clumped) when series of more than two consecutive parcels bound for the same cage are observed.
During the five week manual measurement of conveyor belt stops (Q3) also the visual inspection of
the input mix took place. Series of more than ten consecutive parcels, bound for the same cage were observed during each of the asset shifts. This whilst the chance for such an occurrence is only ( 1 / 60 ) ^ 10 when assuming equal demand per delivery round. When assuming that these parcels were bound for the cage with highest average demand, responsible for approximately 10% of total demand, this observation equalled a chance of ( 1 / 10 ) ^ 10. Clearly the chance that these series
with a chance of one on ten billion happens every shift is not likely. Therefore the situation is far beyond a random situation and follows a highly over dispersed pattern. Employees and the manager also confirmed these observations. Therefore, both demand per delivery round and the input mix have been found to be over dispersed.
The third question was: How effective is the current layout in dealing with (external) variations (like the possible ‘random input mix’ (Q2) and possible fluctuating demand for delivery rounds (Q1))?
The selected indicator for layout effectiveness was the total minutes that the conveyor belt was stopped as a percentage of the shift time. Because such data was not available on the computer system of the depot, the data was gathered by manually monitoring the inbound logistics sorting process during a five week period. A positive side effect of doing this intensive research was that the workings of the inbound logistics sorting process became better understood. The results that were obtained during the observations of the conveyor belt process are displayed below in figure 11. Also the figure concerning the conveyor belt workstations was displayed for convenience.
It shows that during a 240 minute shift the conveyor belt is stalled for about 50 minutes. Therefore, the current layout is not effectively dealing with (external) variations for 21% of the time. When the conveyor belt is stopped, five workstations are starving, as only the workstation causing the stoppage can continue working. As a result of this, the average workstation wastes 42 minutes or 17% of their time by waiting for stoppages. Note that just as an overload of work, workstations can
1. Feed 2. Scanner
3. Reader 4. Light parcels
5. Heavy and multiple piece 6. Partner
Figure 10: Conveyor belt workstations
Cages
Pallets
also face times with a lack of work. In such cases the conveyor belt is not stopped but it will add to the total unproductive time of that workstation.
During the manual measurement another interesting observation was made: The main reason for a workstation to stop the conveyor belt is due to an overload in work. In fact, besides a very few motivational instances, no other reason for stoppages was observed during the weeks of observation. Typically, a workstation would come to face a series of parcels bound for that workstation and have to stop the conveyor belt in order to prevent the parcels from passing the workstation. This implies that the layout is especially bad in dealing with variations resulting from over dispersion. The employees and the supervisor confirmed this observation.
Conclusion
It was found that both the input mix and demand per delivery round were over dispersed. Combined they will create a stream of parcels with highly unequal demand per workstation, which was observed to cause stoppages. As a result of stoppages, the average each workstation suffered 42 minutes or 17% down-time during each shift. Meaning that if the underlying problems were taken away and the arrival of line haulers allowed it, the sort could be ready 42 minutes earlier. This would be a significant contribution to the on time delivery capability of inbound logistics.
3.2
“Soft” factors
Biannually an independent bureau holds a survey under the companies’ personnel, with the focus on the soft part of the organisation. Each question is answered by filling in; positive, negative or neutral. In the methodology it was argued that the amount of negative answers are interesting, as they show the improvement potential. Therefore conclusions will be drawn on the percentage of negative answers out of all answers. Without a proper control group to compare the ratings of inbound staff with, a mathematical standard was chosen. To all questions staff can give three answers, therefore mathematically one would expect one third of negative answers(for more information see chapter 6. discussion). This means that more negative answers will be perceived as a negative rating about a subject and less as a positive rating. Data from this biannual company questionnaire will also be employed to answer the remaining research questions (Q4, Q5, Q6 and Q7). To this extend several
question regarding each subject were select from the questionnaire on the basis of relevance. Since the questionnaire is also held under the office staff of the depot, the results for this group were included. They cannot function as a real control group, because their job characteristics are significantly different from those of inbound logistics. However, they can still give an indication about the ratings given by inbound logistics staff. To ensure objectivity, the questionnaire was filled in anonymous by the employees and handled by an independent bureau. A total of 23 employees of inbound logistics filled in the questionnaire and 66 depot staff.
The fourth question was: How motivated are employees of inbound logistics? This was measured by selecting six questions with relevance to motivation out of the 2010 questionnaire.
Figure 12 shows that staff of inbound logistics answer to the questions regarding motivation with an average of 37% negative ratings. This is more than the 33% that would be expected and therefore it can be concluded that the staff is unmotivated. When taking a closer look at the results it can be
observed that employees of inbound logistics are especially negative about their overall satisfaction and the company’s ability to energize them to “go the extra mile”, with 48 and 57%.
When comparing the answers given by employees of inbound logistics with depot staff, it has to be concluded that they answer more negative to every question, but the sixth. This poses a remarkable contradiction to the overall conclusion: inbound employees are less motivated but are less willing to leave the company than depot staff. A possible explanation can be found in the turn of events. Recall that in 2009, staff of inbound logistics were faced with both a significant cut in their working hours and were denied the possibility to work overtime. With such conditions, as well as the through physical working conditions one would expect, that employees who could get a full time job elsewhere would have chosen to leave the company. The reality is however, that most employees could not find another job and had to accept the changes. Therefore, a hypothetical explanation for why the personnel of inbound logistics are not considering leaving the company is that they know their chances at the job market are low.
The fifth question was: How was the motivation of inbound logistics staff influenced by these historical events? The previous paragraph showed that employees of inbound logistics were less motivated. This paragraph will explore whether this can be explained by the cut in working hours of inbound logistics personnel, which was also discussed in the previous paragraph. The impact of the historical events was measured by comparing the results from the 2008 survey, of before the events, to the 2010 survey.
Figure 13 shows that in 2008 the average negative answers was 26% compared to 37% in 2010. This is a 41% increase in negative answers; a significant drop in motivation (paired t-test: p < .05). Since no other noticeable event has occurred during the time span, it can be concluded that the cut in
answered less negative. To put this in perspective: even though personnel of inbound logistics are less motivated and less proud to be part of the company, they answer that they are more willing to work beyond what is required. This is counter intuitive and might be the result from a feeling of resentment over the past events.
The sixth question was: How motivated are employees of inbound logistics by their performance evaluation and reward system? The answers to six questions relevant questions are displayed in the following figure.
Figure 14 shows that inbound logistics staff answers 39% negative on these questions regarding the performance evaluation and reward system. The only question far below the expected 33% line is number 11 with only 14%. This implies that staff thinks they are held accountable for delivering what is expected, even though they do not feel motivated by the other aspects of the performance evaluation and reward system. Furthermore, compared to the depot, inbound logistics staff give a higher percentage of negative ratings to all questions.
The seventh question was: How inspired are employees of inbound logistics by their management? Questions regarding direct management were gathered from the 2010 questionnaire. Figure 15 shows the answers given to these questions by 23 employees of inbound logistics and 66 depot staff.
The average percentage of negative answers regarding management is 43%. Comparably, the depot staff only gives 28%. Therefore it can be concluded that direct management of inbound logistics does not succeed as well in inspiring their staff, as direct management of the depot does. Furthermore, the same pattern can be observed for all questions. Inbound logistics staff are especially negative (57%) about their communication with management (13).
Conclusion
With the results from the biannual questionnaire the soft part of inbound logistics was investigated. Compared with the expected 33% of negative answers, the graphs show that on average inbound employees give a 37% negative rating to questions regarding motivation. Two years ago this percentage was only 26, showing a significant decrease in employees’ motivation as a result of the historical events. Especially negative were the questions regarding overall satisfaction and whether the company can energize the employee to “go the extra mile”; with 48% and 57% respectively. The performance and reward system was also rated negative, with an average of 39% negative answers. The employees were especially negative on getting recognition when doing a good job and seeing a link between reward and performance (50% and 57%). However, they thought they were held accountable for delivering what is expected (14%). This implies that only negative performance is being observed by management.
The last subject investigated was how inspiring direct management was. Here, the average percentage of negative answers was even 43%. All questions were met with a higher percentage of negative answers than the expected 33%. Most negative answers were given to a question about
In general it can be concluded that motivation was found to be low and all the, in the methodology proposed causes of this low motivation have been confirmed.
3.3
Empirical causal model
Finally, the results obtained in this chapter were used to form a causal model, which shows the empirically proven situation. Note that figure 16 compared to the casual conceptual model, in the methodology, is exactly the same; as each proposed factor’s existence has been proven. See also chapter 6 for a discussion of the results.
Delivering late Productivity Employee motivation Layout effectiveness Historical events Input mix randomness Demand fluctuations Inspiring reward system Inspiring management
4. Solution
The previous chapter showed that a lower than desired productivity at the Scandinavian depot was a result of several causes, both on the “hard” and “soft” soft side of inbound logistics. In this chapter solutions are proposed to deal with these causes, in order to improve productivity and ultimately on time performance.
4.1
“Hard” problem
The inbound logistics was found to face both an over dispersed demand and input mix (see empirical study for more details). It appeared that the current layout of the sorting process could not handle this variation, as the conveyor belt stood still 21% of the time. However, the root causes (over dispersed demand and input mix) cannot be taken away as they stem from customers demand, which is almost impossible to change drastically (Ravi and Rosenblatt, 2003). Therefore, the non-constant demand and random input mix have to be seen as a “fact of life”. Any possible solution thus has to deal with this inherited variety.
During the data gathering phase it was observed that the conveyor belt stopped, not necessarily because there were more parcels on the belt, but because one workstation would face a lot of parcels. The result was that the busy workstation would stop the conveyor belt, starving the other workstations. Currently, the conveyor belt stop is used to prevent parcels bound for a workstation to be carried past that workstation. If a parcel passes the workstation where it should have been sorted, the employee needs to walk the entire extra distance it travelled, collect it and walk the distance back again to put it the right cage. However, a stoppage affects all employees and therefore personnel tends to focus their activities on preventing stoppages, by prioritizing parcels furthest down the belt. This whilst staff should focus on sorting efficiently; by prioritizing parcels close by their own position, which minimizes walking time. Therefore, stoppages as a result of parcels moving past their workstation should not occur. A solution to this problem can be found by letting the past parcel return automatically. This proposal will look like a luggage conveyor on an airport (see figure 23). Parcels are thrown on to the circle shaped conveyor and circulate until someone takes off the load. This removes the need for stopping the conveyor belt entirely since the missed parcel will come back at a later moment (perhaps when the workstation is not so busy). As a result, employees can focus on productivity, instead of preventing conveyor belt stoppages. This proposed solution will be put to the test in an extensive simulation, in section 4.3.
4.2
“Soft” problem
In the previous chapter it was proven that staff of inbound logistics has a low motivation. Causes for low motivation were found in management, the performance evaluation and reward system and due to historical events. The implications of each of these causes will be discussed further and possible solutions will be formulated.
Figure 17: Simplified overview of the shifts
Even with an extra buffer hour (to get the sort done if the deadline is not met), there is not enough work for the employees to give them a full time job at the inbound sort.
The same phenomenon also occurs at the outbound sort, meaning that employees in both shifts have part time jobs. However, people are not allowed to work both shifts as the regulation from the collective agreement states that employees in the transport sector need to rest at least 12 hours before the person is allowed to work again (even when working a few hours). In fact, employees that worked “overtime” before actually worked the outbound sort, which is now forbidden.
A possible solution to this problem for both shifts can be found, since the van drivers work a full time shift in between the working hours of the sorting shift. Splitting up the van driver’s shift will accommodate two full time jobs for warehouse staff, which would
work as following. In the early morning a staff member performs the inbound shift, after which he or she takes the wheel of a delivery van to work another four hours. At lunch this staff member is ready and can go home, having worked a full eight hour job. A second employee will take over the van to do the afternoon pick up and will come back to the depot and work the outbound sorting shift. The result being that both employees have a full time job and should therefore become more motivated. Next to this, both employees will get more job variety, generating even more motivation (Morgeson, 2006). And on top of this, employees will have contact with the customer, which is also found to be a motivator (Ryan and Deci, 2001; Wrzesniewski, Dutton and
Debebe, 2003). And last but not least, the employees working the inbound sort get more knowledge about the delivery rounds, possibly improving the quality of his sorting capabilities. If this solution were to be implemented for the Scandinavian depot, approximately fifteen of the forty van drivers would lose their jobs. Since this is not an cheap (emotionally and economically) thing to do in a Scandinavian country, this will be discussed further in chapter 6.
Another discrepancy found to affect the motivation is the performance evaluation and reward system. A good system will let the employees know what to expect and will make them feel that their efforts will pay off (Grant, 1989). However, from the analysis result as presented in figure 14, it appears that a major part of the employees do not see a link between performance and reward. Therefore these employees do not feel their efforts pay off.
Motivation theory (Ryan and Deci, 2001; Wrzesniewski, Dutton and Debebe, 2003) recognizes that work is not performed in isolation. Rather, work is done within organizations and for other people and motivation comes from feeling useful, social interaction and status. Since there is no social interaction with the customer, recognition mainly has to come from the management team. However, Figure 14 shows that 50% of the employees answers negative to the question whether good work is recognized. Contrary, only 14% answers negative to the question whether people are held accountable for delivering what is expected. This lead to the conclusion that there is a discrepancy between (management) attention given towards negative performance as opposed to positive performance.
The main solution to this is that management will balance their attention better between positive and negative performance. Assuming that the intentions of management are good and realizing that balanced performance recognition serves both the employees and the company, there should be no problem in making them want to change their habits. Awareness has to be raised under management about the importance of their role for the employee’s motivation and the discrepancy in attention given to positive and negative performance. The person ultimately responsible for this is the operation manager, who can raise awareness, simply by calling a meeting and discussing the matter. A more substantial solution is to make performance recognition of employees a subject to discuss between the managers every month. Besides management wanting to balance their performance recognition, also trust in management and their communication skills play a role, which will be discussed in the next paragraph.
Conclusion
5. Simulation
In order to prove the usefulness of the “hard” layout solution (the circular conveyor belt), it was simulated and benchmarked against a simulation of the current layout. First a simulation for the current situation was designed, which formed the basis of the model incorporating the proposed solution. The following method was derived from Robinson’s (2004) book, with added justifications per item.
5.1
Simulation of the current linear layout (S1)
Item Justification
Objective: Productivity comparison between
linear and circular layout.
Time-scale: Shift (4 hours) Problems occur over time due to unequal workloads.
Time-unit: Seconds (4h *60m *60s =14400s) To have enough detail to see events unfolding,
because parcels are in motion for only a few minutes.
Input: Parcel destination In the real process they also arrive in an unordered
manner and are the cause for unequal workloads.
Output 1: Stoppages caused by each workstation Allowing workload comparison with the real
workstations to show validity of the model.
Output 2: Parcels sorted during the shift Allowing productivity comparison with results from
the circular layout simulation (S2).
Output 3: Idle time per workstation As an indicator for improvement potential.
Model scope
Component Include? Justification
Parcels Include
Conveyor belt Include
Stoppages Include For greater accuracy and will be required to derive
output 1.
Workstations Include
Employees Include Each workstation consists of one or more employees,
modelling employees will therefore provide for greater accuracy and once one model is established can be easily copied to serve on every workstation.
Detail Include? Justification
Parcels Destination Size Weight Hazardous materials Priority parcels Include Exclude Exclude Exclude Exclude
Input (randomly derived from historical distribution) Influence not well understood
Influence not well understood
Small % of volume, only applicable to first two workstations and influence not well understood. Small % of volume, only applicable to first two workstations and influence not well understood.
Conveyor belt Length
Motion
Include Include
Required to determine parcels position
employee
Workstations 1. Feed
2. Scanner
3. Reader 4. Light parcels 5. Heavy and multiple 6. Partner Include Exclude Include Include Include Include
Required to put parcels in the process
Is responsible for a tiny portion of stoppages as a result of procedures surrounding special packages. Otherwise enough capacity.
Required to derive output 1. Required to derive output 1. Required to derive output 1. Required to derive output 1.
Employees Individual behaviour
Range Walking speed Pick up time Delivery time Exclude Include Include Include Include
Influence not well understood
Required to determine output 1 and 3. Required to move parcels to their destination. Required to differentiate between workstations Required to differentiate between workstations
Assumptions1
- Equipment failures occur rarely and so do not need to be modelled. - The conveyor belt can hold 180 parcels.
- The rate at which parcels can be put on the conveyor belt is equal during the whole shift. - Estimations on walking speed, pick up and put down time are accurate.
- Employees behavior is geared towards minimizing the number of stoppages
Simplifications2
- At the first workstation (feed) there are always enough parcels available to put onto the conveyor belt, therefore variability from both line hauler arrival times and the unloading processes are not modelled.
- The reading process, involving the third workstation’s sort to either side of the band and workstations 4-6 reading the parcels again to gain knowledge of its destination, was simplified by a mark up factor for the third workstation3.
- Each employee from the sorting workstations (3-6) covers a fixed range, instead of having more fluid borders.
- Since the productivity of the shift is the objective of this simulation (and not the on-time performance), the process stops immediately when the 14400 seconds are over.
1
Assumptions are ways of incorporating uncertainties and beliefs about the real world into the simulation (Robinson, 2004).
2
Simplifications are ways of reducing the complexity of the simulation (Robinson, 2004). 3
5.2
Model data
In order to easily determine a parcels position on the virtual conveyor belt, length was depicted in seconds. In this way the position of a parcel, relative to the beginning of the conveyor belt, is the amount of seconds it was on the conveyor belt minus the stopped minutes. It was measured that it took a parcel, without stoppages, 180 seconds to get to the other end of the conveyor belt. In the same fashion the position of each cage or pallet was determined. Next, the percentage of total demand per cage was calculated from a data set of four weeks. The following table shows this information, as grouped by the virtual employee responsible for servicing the cage. Note, that the pallets and the three virtual employees at workstation five are situated on the other side of the conveyor belt.
Next to the information per cage also the parameters for each virtual employee had to be obtained. These parameters were estimated with the help of the inbound logistics supervisor. Perhaps the most important parameter was the rate at which the first workstation puts new parcels on the conveyor belt. With 4000 parcels during each 14400 second shift, the average feeding rate would be one parcel per 3,6 seconds. However, parcels are not allowed to be stacked and therefore an allowance has to be made for when the conveyor belt is stopped. Therefore, the rate was established at three seconds.
The range of each virtual employee, taking parcels from the conveyor belt, was defined by the first and the last cage serviced. The starting position of an employee would naturally be at the beginning of this range, whereas the conveyor belt would stop when a parcel moved past the last cage. The average time it took an employee to pick up a parcel from the band, as well as putting it in a cage, was established at one second for light parcels and three for heavy parcels. The speed at which employees moved along the conveyor belt was estimated at twice the conveyor belt speed for employees handling heavy parcels, compared to three for light parcels. Two mark up factors were also incorporated: for the omission of the reading process a big mark up of five seconds was added to virtual employee 31 and for an extra scanning duty a two second mark up factor was added to
5.3
Model description
Without any simulation software at hand, the model of the conveyor belt sorting process was created in an extensive excel file. This paragraph will explain the logic behind the model.
The core of the simulation was a virtual conveyor belt, which was represented by 180 columns. Rows represented the time in seconds, during the entire shift (14 400 seconds). When the conveyor belt was turned on, parcels would move a distance of one column, in the space of one second (one row). It took 180 seconds for a parcel to travel the entire distance of the real conveyor belt, thus explaining why this number of columns was chosen to represent the virtual belt. Each of the 180 spaces on the conveyor belt could either be empty, represented by a zero, or be filled with a “parcel”; symbolized by a unique number (the first parcel was a one, the next two and so on). In this way each parcels location could be pin pointed at any time (row). Next to this unique number other information about the parcel was generated; a destination and a routing. Via the unique number it was therefore possible to trace back this information with lookup functions.
At the beginning of the virtual conveyor belt two employees load the parcels. The speed at which they load the parcels on the conveyor belt is one parcel every three seconds, embodied by a unique number. To this “parcel” a random destination was assigned, based on a distribution that was as shown in figure 18. On the basis of this destination routing was looked up; which virtual employee had to pick up the parcel and deliver it to its destination cage/ pallet. When the band was operational, parcels would be created at one end, and the virtual conveyor belt would “move” the parcels one column every second (row) towards the other end.
The next step was to create the workstations that performed the actual sorting. When designing these workstations it was recognized that the workstations consist of either one or a few employees and that these employees all performed the same sorting process. Therefore, this process was modelled once and copied. The differences between workstations was accommodated for by building in the parameters from figure 19.
Each of the virtual employees was set to deliver parcels within a certain range, for instance: between distance 100 and 132. The function of the “employees” was to determine what parcels needed to be “delivered” within their range, pick up such a parcel from the conveyor belt and move to the destination “cage/pallet”, where it subsequently was deposited. Since the virtual employees could only handle one package at a time, there was a risk that parcels slipped past them, just as in the real situation. When such a parcel was not picked up from the conveyor belt by the end of the “delivery range” (position 132 in the example); the conveyor belt would stop automatically (as in the case at inbound logistics). Due to this stoppage, parcels would stop moving columns and the mechanism responsible for loading the conveyor belt would pause. Naturally time would move on (rows) and although the conveyor belt was stopped the virtual employees would continue their work (just as in
