Introduction

The importance of warehousing as a key activity for manufacturing companies is widely recognized by scholars and practitioners. Although in recent years a more pervasive automation under the label “Industry 4.0” is emerging, a large number of warehouses still strongly rely on human workers, such as forklift drivers, especially in picking activities. Indeed, the picking process is considered among the most laborious and expensive activities in warehouses, and estimates suggest that around 80% of the order picking is still carried out manually [1].

In such a context, where individuals are predominantly involved in material handling activities, individual and team dynamics are among the fundamental drivers of workplace motivation, wellness and system productivity [2]. As a result, there is a growing interest in taking into consideration human factors such as stress, health and collaboration dynamics in the logistic operations. Nevertheless, while such factors promise to achieve better business performance and high levels of safety and well-being of workers, there is still a lack of studies [3] that consider both human and working environment factors in such a context. Indeed, studying these variables quantitatively can lead to a clearer insight on which are the factors that affect productivity in such logistics operations.

The recent advent of wearable sensors, and other sensor-based measurement tools (e.g. sociometric badges and smartwatches), is offering researchers the opportunity of collecting and analyzing human factors through data-driven methodologies [4]. For instance, smartwatches are particularly suitable to measure acceleration and heart rate, useful as predictors for physical fatigue of workers [5], and they can efficiently collect a big amount of data in real time, avoiding the main bias linked to common approaches in behavioral studies and increasing the data richness, quality and reliability. On the other hand, more conventional instruments such as the thermo-hygrometer and lux-meter can be used to measure data about the quality of the environment (e.g. temperature, humidity and luminosity).

Background

Many researchers developed analytical models to improve picking activities and increase efficiency, suggesting different warehouse layouts, routes or archiving tasks, but only a minority of them focused their studies on the human factor, as highlighted in the study of Grosse et al. [3] and reported in Fig. 6.1.

Fig. 6.1
figure 1

Literature studies about human factors in warehousing (Source [3])

Full size image

This study suggests that future research should focus on the connection between the order picking system design and the aspects of human factors and it indicates opportunities of integrating aspects of human factors in management-oriented warehouse picking research. A more detailed study about human factors and ergonomics could be promising to achieve higher performance (time/productivity) and quality (reduced errors) and to improve safety and well-being of workers, as suggested by Neumann et al. [6].

Moreover, a logistics hub is a complex environment where external factors can affect performance, such as the carrier selection [7] or fragmentation and scarce collaboration among the extensive networks of suppliers and subcontractors [8]. Regarding human factors, there is a multitude of risks and dangers regarding the working environment that leads research conduct studies about environmental factors such as temperature, humidity and luminosity, as reported by Leather et al. [9].

Our goal is to embrace all these factors together and to collect quantitative data in order to explore possible connections among them.

Research Objective and Methodology

This research aims to empirically explore environmental and human factors that may affect productivity and more generally performance of operators in material handling activities. Toward this goal, an exploratory case study was conducted in a semi-automated logistics hub, located in Tuscany, Italy.

The research methodology includes the following steps:

  1. (1)

    Context analysis. A context analysis of the logistics hub has been done for setting the case up and to test preliminarily the innovative measurement tools in the specific application context (e.g. smartwatches, thermo-hygrometer and lux-meter with datalogger function).

  2. (2)

    Data collection. Data have been collected from employees through the smartwatches (e.g. body movement, acceleration, proximity, speaking, heart rate), while we used thermo-hygrometer and lux-meter for working environmental measures. Finally, productivity data are provided by the warehouse management systems (WMS).

  3. (3)

    Data pre-processing. Data from different sources have been properly pre-processed and homogenized to obtain comparable data useful for the analysis phase. Moreover, in order to avoid bias due to the physiological differences among workers (i.e. different heart rate baseline), smartwatch data have been normalized for each operator. Then, additional features were created through the combination of the measured variables, for example, the mood variables, based on Russell’s circumplex model [10].

  4. (4)

    Data analysis. Correlation and regression analyses have been conducted to propose an explanatory model that can provide guidelines for management with directions for possibly improving workers’ productivity, safety and well-being.

Case Study

The experiment has been carried out in a semi-automated logistics hub, located in Tuscany, Italy. The hub is owned by a 450+ million revenue company that produces and sells paper-tissue products all over Europe. The warehouse covers an area of 24,000 m2 with a height of 7 m and it is open 24 h a day from Monday to Saturday afternoon. Operators work in three shifts.

Operational activities carried out in the warehouse consist of the stocking of inbound pallets and the picking of ordered pallets that must be delivered. Each pallet is 80 cm × 120 cm and the weight is not considered critical due to the lightness of the paper. About 10 operators are employed for each shift, five operators for picking, four for stocking, one team leader, and people from administrative staff.

The layout of the warehouse is shown in Fig. 6.2. It is divided into six areas. Each of the zones numbered from 1 to 5 has one pallet shuttle, while zone 6 is managed using stacks, without any automation. The use of pallet shuttles provides more density of shelves in the warehouse, as well as a greater safety for the operators. We note that although the pallet shuttle partially supports warehouse automation, the presence of workers is still very necessary.

Fig. 6.2
figure 2

Logistics hub layout

Full size image

Data Collection

Human and environmental data have been collected from the warehouse 8 h a day, 5 days a week, 8 weeks from March to May 2019, for a total of 320 h monitored for each operator. Figure 6.3 provides an overview of the hardware and software used for data collection.

Fig. 6.3
figure 3

Hardware and software for data collection

Full size image

As for human factors, a group of five picking workers, including the team leader, has been equipped with smartwatches and monitored during working activities for the whole duration of the experiment. The tool allows researchers to conduct more in-depth quantitative analysis of interactions inside organizations leading to high-level descriptions of human behavior in terms of (i) physical activity/human movement, (ii) speech features (rather than raw audio), (iii) indoor localization and (iv) proximity to other individuals.

As concerning environmental variables, a thermo-hygrometer and a lux-meter, both with data logger function, have been positioned near to the outbound area of the warehouse for collecting environmental data.

Finally, productivity is evaluated through the hourly number of processed orders (data were extracted by the WMS).

Figure 6.3 summarizes our data collection instrumentation.

Results

The most significant correlations and p-values are presented in Table 6.1. Results show that there is a negative correlation with “worked_hours”, that is, the number of hours worked by an operator, while the correlation with “team leader connection”, that is, the measurement of how much an operator gets in touch with the team leader, is positive. As an addition, as concerning the features based on Russell’s circumplex model [10], the variable “nervous” seems to have a negative correlation with productivity. The last variable “tired” also shows a significant positive correlation with productivity.

Table 6.1 Correlations with productivity
Full size table

The regression model, that best explains productivity, is shown in Table 6.2.

Table 6.2 Regression model with productivity as dependent variables
Full size table

The regression model has an R-squared of 0.079 and an adjusted R-squared of 0.067 (p-value is 2.772 E–08). Considering that orders assigned to the employees are not always homogeneous, for example, they may differ regarding the difficulty degree, the personal factors influencing individual productivity and other possible confounding effects, a result of less than 10% is acceptable.

Findings show that the only positive coefficient is “team leader connection”, while the other coefficients negatively affect productivity. Performance seems to improve with the increase of connections between the operators and the team leader and if people work in an environment that allows them to be less nervous.

We controlled for the “week day” because the day of the week can characterize the type and complexity of work and intensity of operations; “team leader presence” because it helps to verify that the variable “team leader connection” is not connected to the physical availability (presence or absence) of the team leader in the warehouse; “luminosity” because usually lighting can contribute to improving workplace condition and overall operation performance; “hour” to double check the “worked_hours” variable. We note that “worked_hours” has a greater coefficient than “hour”; therefore it can be assumed that it is intended as a measure of the tiredness of an operator.

Discussion and Conclusions

This work explores the human and environmental factors affecting the productivity of employees in a semi-automated warehouse by using wearable sensors. As an addition, it methodologically contributes to empirically testing the innovative hardware and software for data collection in this specific setting.

Preliminary results show that, as expected, individual productivity depends on the worked hours, and hence is affected by the fatigue of operators, but it also seems to be influenced by coordination mechanisms and by individual variables such as nervousness and stress of employees. Specifically, an increase in the connection of operators with the team leader, which seems to act as a facilitator of operations, increases the productivity. Also, a high level of stress and nervousness leads to a decrease in individual productivity. Finally, no clear evidence about the role of environmental factors emerges in the experiment.

Some managerial implications can be derived, such as the adoption of part-time workers to reduce the amount of worked hours and avoid high fatigue of employees, as well as incentivizing effective collaboration and interactions of operators with the team leader. While it is difficult to confirm a direct causal relationship between human and environmental factors and operators’ productivity, we emphasize the opportunity for managers to care for operators’ needs and working environmental conditions in order to improve performance.

Further works should improve this study by extending the time window for data collection in order to cover more variability in the handling activities, by increasing the number of observed workers and by extending the investigation to different warehouses with different type of goods, automation level, and so on. Also, a greater number of environmental sensors could be adopted to cover the areas of a warehouse to provide more detailed insight on the role of environmental factors.