**4.2 Stage 1—Exploring users' pains and needs**

*Concepts, Applications and Emerging Opportunities in Industrial Engineering*

immense accounts of information gathered.

**4.1 Stage 0—Constructing the organizational 'picture'**

ization; and a specific case study contextualization.

The proposed methodology was explored on an empirical case study of an airport's Baggage Handling System (BHS) of a European airport. This application followed the methodological procedure presented in **Figure 4** and used some of the tools and practices presented on **Table 1** to help on making sense out of the

This initial stage had the objective of giving an overview of the case study environment, and it was divided in two main topics: an overall industry contextual-

The first topic started with an overview of the air transport industry and its current trends, followed by a deep-down into the airport environment. Here, a brief overview over its control and supervision was made, and some generic characteristics were laid down concerning multi-airport systems, international differences, stakeholder's mapping and economic chain, and the main regulatory drivers. Afterwards, the focus was directed to the airport's Baggage Handling System (BHS), where a generic description of its objectives, work process and primary functions was made, along with the specific regulatory drivers and upcoming regulatory frames on BHS industry. The second topic intended to be company-specific within the case study boundaries. It started by giving a technical description of the equipment and controlling systems used for its operation. Then, the specific BHS ecosystem was portrayed, along with its main stakeholders and their influence/impact on BHS operations. Following, the BHS' operational performance and its main dimensions were addressed. Afterwards, a thorough description of the operational characteristics and the workflow along the various stakeholders involved on a baggage journey was made, depicting the intrinsic dependencies between the various BHS stakeholders. At last, an organizational overview was done showing the overall structure and the insite hierarchy of operation and management, as depicted on **Figure 5**. Subsequently, a final description of the BHS operational management is addressed, in order to understand how the daily processes and concerns were managed and resolved.

**4. Case study**

**104**

**Figure 5.**

*In-site organizational hierarchical structure of BHS management.*

After constructing the initial industry picture and understanding the surroundings where the BHS is included and its basic characteristics, the next step was to empathize with the various stakeholders which interact or impact the system's operations. A collection of tools was employed throughout this stage to harness on the gathered data and provide profound insights about the intuitive and relational aspects of human relations and people's reasoning when performing their jobs.

Approximately 7 hours of shadowing were undertaken to carefully observe BHS control room operators while operating the BHS. Here, the researcher had the opportunity of first-hand contact with the real-time management, which allowed him to pose some specific questions about the interactions between the CR operators and the information systems utilized for the BHS' operations management. Additionally, several semi-structured interviews with the BHS' main stakeholders were conducted to obtain a rich and broad view from the various perspectives that come into play when taking BHS operation performance into account. A total of 12 interviews were performed by the researcher, segregated according to the **Table 2**.

Each semi-structured interview was used to construct a mind map that allow for a summary of the gathered information and themes discussed, which helps to mentally process and reflect over them to generate deeper insights. Alongside these interviews, the author also performed a total of five customer narratives with different passengers, with the objective of understanding some common pains and needs that the final customer of the whole baggage journey could have. In order to get a broader picture and minimize biases throughout these collected stories, the author diversified the customers' profiles along the five customer narratives (e.g. old vs. new, frequent flyer vs. novice, family travel vs. business).

A total of seven personas were constructed in order to have a concise and coherent definition of the information gathered for further presentation and discussion with a focus group composed by experts. They comprised the main concerns and the global opinion between the interviewees belonging to each stakeholder's persona. Some stakeholder's categories were aggregated according to the relative importance to the BHS operational environment and management, resulting on the personas' set depicted on **Figure 6**.

#### **4.3 Stage 2—Defining and ideating on the problem-solution space**

This stage focuses on developing an iterative and collaborative environment for the discussion and refinement of the perceived problem space, as well as for the exploration of the solution space. This was an iterative process of exploration and discussion, in order to clarify and improve the problem definition and consequently, the solution ideation.


**Table 2.**

*List of semi-structured interviews, grouped by project's stakeholder.*

*Concepts, Applications and Emerging Opportunities in Industrial Engineering*

#### **Figure 6.** *Total stakeholders personas constructed.*

The authors started by reviewing the current insights gathered during the exploration phase, to look deeper into any misconnection or discover profounder insights. Hence, it was reexamined the mind maps and personas created in the prior stage, in order to develop a motivational map. This was used to understand and connect the various stakeholders' perspectives around a central concern identified throughout different interviews and, deconstruct it to find opportunities and clear the problem definition.

After trying to find connections to continue developing the problem space understanding, there were already many pains and needs identified throughout the data collection processes. Therefore, the authors had to filter them and find the most critical ones, in order to have an efficient and effective dialog with the experts' group. With this purpose, every pain and need identified until this step were listed and a multi-criteria prioritization was made. As such, the authors created a set of four criteria consisting in:


From this exercise, five main problems were prioritized, which were then taken into discussion with a focus group of experts. This group gathered both managers and engineers with experience ranging from 5 to 20 years with this type of systems, and belonging to operational and back-office teams. The meeting started by presenting a summary of the overall organizational 'picture', followed by the presentation and discussion of the five main problems prioritized earlier. As a result, it was agreed on one most desirable opportunity to be pursued, which would be the one to be explored on the solution development stage:

**107**

*A Hybrid Human-Data Methodology for the Conception of Operational Performance…*

At this point, the pivotal opportunity had been identified and it was the one explored throughout this stage. This opportunity was related to an internal process of the BHS, thus having available automatically generated data to leverage on analytics tools, which was one of the essential research objectives. This stage had different progressive steps in order to delve deeper into this opportunity and approach different solution possibilities, according to their desirability and feasibility in terms of time and data resources available for analysis. Additionally, the BHS information systems' functioning and their integration had to be fully understood, to comprehend the type of data being captured, and how it could be combined and

The first step undertaken by the researchers was to further explore the identified opportunity, using the Motivational Map tool one more time. This enabled the deconstruction of the problem into some more specific issues and concerns, as well as the identification of solution paths to take into consideration. Additionally, it also allowed the narrowing down of the solution space into more concrete and defined opportunities to be further investigated. Hence, the initial opportunity was trans-

Consecutively, the authors had to build a mental model which allowed him to know the dataset he would need to have in order to tackle this opportunity and develop some analytics' prototypes. This was made with the use of a metaphor with a roundabout, through which the researcher could find an analogy between the real opportunity another common context to help him on reasoning and understanding

In order to gather the necessary dataset for further exploration and prototyping on the solution space, the authors had to firstly understand the underlying information systems and the type of data collected by each, and how these could be integrated into a complete coherent dataset for analysis. To help on this task, the author resorted to the Tableau Software™, an interactive visualization software for business intelligence, which allowed to visually display various combinations of variables to understand their relations and the real meaning of the virtual captions. The initial understanding of these IS, along with the mental exercise provided by the roundabout metaphor, allowed for the author to understand what types of data he would need and could have access to, in order to comprehend the sorter's conges-

• Data about the injection lines, i.e. the amount of baggage which enters the

• Data about the transfer chutes, i.e. the amount of baggage which leaves the

• Data about the sorter's congestion, i.e. combination of operational factors which indicate the existence of the incapacity of the sorter to deliver baggage

• Data about the assigned baggage destinations, i.e. the amount of baggage inside

The construction of the necessary dataset resulted on a combination of different

*DOI: http://dx.doi.org/10.5772/intechopen.93631*

leveraged with the power of analytics tools.

formed into a concrete solution to be explored:

tions. Thus, these could be summarized into:

sorter through each injection point;

the sorter which is destined for each transfer chutes;

sub-sets retrieved from the various IS, as illustrated on **Figure 7**.

sorter through each chute;

into their respective chute;

"Sorter X congestions' comprehension – causes and impacts"

the flow of baggage through a sorter and the possible causes for congestion.

**4.4 Stage 3—Solution exploration**

"Prevent the sorters' congestion"

*A Hybrid Human-Data Methodology for the Conception of Operational Performance… DOI: http://dx.doi.org/10.5772/intechopen.93631*

## **4.4 Stage 3—Solution exploration**

*Concepts, Applications and Emerging Opportunities in Industrial Engineering*

The authors started by reviewing the current insights gathered during the exploration phase, to look deeper into any misconnection or discover profounder insights. Hence, it was reexamined the mind maps and personas created in the prior stage, in order to develop a motivational map. This was used to understand and connect the various stakeholders' perspectives around a central concern identified throughout different interviews and, deconstruct it to find opportunities and clear

After trying to find connections to continue developing the problem space understanding, there were already many pains and needs identified throughout the data collection processes. Therefore, the authors had to filter them and find the most critical ones, in order to have an efficient and effective dialog with the experts' group. With this purpose, every pain and need identified until this step were listed and a multi-criteria prioritization was made. As such, the authors created a set of four criteria consisting in:

From this exercise, five main problems were prioritized, which were then taken into discussion with a focus group of experts. This group gathered both managers and engineers with experience ranging from 5 to 20 years with this type of systems, and belonging to operational and back-office teams. The meeting started by presenting a summary of the overall organizational 'picture', followed by the presentation and discussion of the five main problems prioritized earlier. As a result, it was agreed on one most desirable opportunity to be pursued, which would be the one to

**106**

the problem definition.

*Total stakeholders personas constructed.*

**Figure 6.**

• Relative reference frequency

• Relative occurrence frequency

be explored on the solution development stage: "Prevent the sorters' congestion"

• Impact on BHS operations

• Feasibility

At this point, the pivotal opportunity had been identified and it was the one explored throughout this stage. This opportunity was related to an internal process of the BHS, thus having available automatically generated data to leverage on analytics tools, which was one of the essential research objectives. This stage had different progressive steps in order to delve deeper into this opportunity and approach different solution possibilities, according to their desirability and feasibility in terms of time and data resources available for analysis. Additionally, the BHS information systems' functioning and their integration had to be fully understood, to comprehend the type of data being captured, and how it could be combined and leveraged with the power of analytics tools.

The first step undertaken by the researchers was to further explore the identified opportunity, using the Motivational Map tool one more time. This enabled the deconstruction of the problem into some more specific issues and concerns, as well as the identification of solution paths to take into consideration. Additionally, it also allowed the narrowing down of the solution space into more concrete and defined opportunities to be further investigated. Hence, the initial opportunity was transformed into a concrete solution to be explored:

"Sorter X congestions' comprehension – causes and impacts"

Consecutively, the authors had to build a mental model which allowed him to know the dataset he would need to have in order to tackle this opportunity and develop some analytics' prototypes. This was made with the use of a metaphor with a roundabout, through which the researcher could find an analogy between the real opportunity another common context to help him on reasoning and understanding the flow of baggage through a sorter and the possible causes for congestion.

In order to gather the necessary dataset for further exploration and prototyping on the solution space, the authors had to firstly understand the underlying information systems and the type of data collected by each, and how these could be integrated into a complete coherent dataset for analysis. To help on this task, the author resorted to the Tableau Software™, an interactive visualization software for business intelligence, which allowed to visually display various combinations of variables to understand their relations and the real meaning of the virtual captions. The initial understanding of these IS, along with the mental exercise provided by the roundabout metaphor, allowed for the author to understand what types of data he would need and could have access to, in order to comprehend the sorter's congestions. Thus, these could be summarized into:


The construction of the necessary dataset resulted on a combination of different sub-sets retrieved from the various IS, as illustrated on **Figure 7**.

#### *Concepts, Applications and Emerging Opportunities in Industrial Engineering*

**Figure 7.** *Process summary to get the necessary data from the different IS.*

From system A, it wasn't possible to get congestion variables since there wasn't a specific event or a coherent combination of current events being captured that could effectively indicate the congestion. Therefore, the author had to assume a different variable which could indicate congestion: the sorter's occupancy, i.e. the percentage of sorter trays being occupied relative to the total available trays. From system B, the objective was partially accomplished, since there was data characterizing the desired variables, but the time granularity with which some variables were captured wasn't enough for the detail level required. As such, the only variable possible to consider was the occupation variable. Finally, the author stepped into system C in order to collect every missing variable to complete the desired dataset. Here, the process of getting this data was more laborious and involved more complex queries in order to align the captured data for the proceeding data pre-processing. From this IS, the author could arrange the variables of the injectors, chutes and assigned destinations of the baggage going through the specified sorter.

From prior data collection, which was gathered from different IS, the author had different data types (i.e. some were counts, others were events) which could not be integrated into a whole coherent dataset to be analyzed. Furthermore, this data came with inherent redundancy, lack of normalization, and even outliers which had to be processed. Hence, different specific steps had to be taken to cleanse and integrate these loose data sub-sets into a whole dataset. Different software tools were used for the various tasks implied on this pre-processing step:


In the end, a total of 66 variables were taken into the next stage, comprising the following variable sets:

**109**

*A Hybrid Human-Data Methodology for the Conception of Operational Performance…*

• Baggage assigned destinations counts, with a total of 24 destination variables;

This stage was intended to undertake an initial analysis of the relationships within the constructed dataset, followed by the exploration of deeper analytic techniques. This procedure is necessary to build some concrete prototypes which will be tested and evaluated, thus becoming the basis for an iterative co-evolution

This stage unfolded in three main steps. On the first step, some descriptive statistics and the bi-variate correlations matrix were performed to better understand the variables within the dataset. This was performed using the Spearman correlation coefficient since all the variables' distributions were skewed. This is explained by the large number of zeros when there is no baggage passing through the collecting points, and the existence of a tail towards the positive integer numbers, given the relatively few but dangerous levels of high-traffic. The matrix showed some strong relations (i.e. coefficient bigger than 0.6) between some explanatory variables and the target, which is a good indicator of the possibility of building a

On the second step, the author tested some linear regressions to create an explanatory model of the sorter's functioning. From the initial dataset, encompassing every-minute counts from a one-and-a-half-month period along the 66 variables, four subsequent datasets were derived by sampling different time periods available on the initial dataset. This was aimed at mitigating the risk of overfitting, which can occur when a regression model rightfully explains the specific dataset, but it poorly performs when given a slightly different dataset. The resultant models created for each of the four datasets and their discussion gave a good overall confidence over an initial explanatory model for the sorter's functioning. Thus, considering the regression model created for the whole dataset, since it has the biggest and most complete time period, it is represented

min 1,191 0,88 ( ) 159158 0,749

∗ + ∗ + ∗ +

*Destination Chute Destination Chute Destination Chute Destination MES*

Basically, this expression provides a succinct and easy way of controlling the sorter's occupancy level and consequently, its congestions. It means that the manager can concentrate his efforts on controlling these five variables (i.e. the number baggage that is inside the sorter with destination to the respective eight transfer chutes) and, consequently, he will be able to assure the sorter does not surpass critical levels which can result on congestions. By keeping a focus on these counts and by guaranteeing the respective chutes aren't jammed, the manager can reduce the sorter's congestion probability and, consequently, the number of delayed baggage.

∗

*Occupancy level Destination Chute*

= +∗ +

 151152 0,899 153160 1,008 172 1, 235 125

(1)

• Injectors counts, with a total of 14 injector variables;

• Sorter's occupancy level, with one 1 variable;

• Transfer chutes counts, with a total of 27 chutes variables;

*DOI: http://dx.doi.org/10.5772/intechopen.93631*

**4.5 Stage 4—Prototyping and testing**

of the final solution.

regression model.

on Eq. 1.

*A Hybrid Human-Data Methodology for the Conception of Operational Performance… DOI: http://dx.doi.org/10.5772/intechopen.93631*


#### **4.5 Stage 4—Prototyping and testing**

*Concepts, Applications and Emerging Opportunities in Industrial Engineering*

*Process summary to get the necessary data from the different IS.*

From system A, it wasn't possible to get congestion variables since there wasn't a specific event or a coherent combination of current events being captured that could effectively indicate the congestion. Therefore, the author had to assume a different variable which could indicate congestion: the sorter's occupancy, i.e. the percentage of sorter trays being occupied relative to the total available trays. From system B, the objective was partially accomplished, since there was data characterizing the desired variables, but the time granularity with which some variables were captured wasn't enough for the detail level required. As such, the only variable possible to consider was the occupation variable. Finally, the author stepped into system C in order to collect every missing variable to complete the desired dataset. Here, the process of getting this data was more laborious and involved more complex queries in order to align the captured data for the proceeding data pre-processing. From this IS, the author could arrange the variables of the injectors, chutes and assigned destinations of the baggage going through the

From prior data collection, which was gathered from different IS, the author had different data types (i.e. some were counts, others were events) which could not be integrated into a whole coherent dataset to be analyzed. Furthermore, this data came with inherent redundancy, lack of normalization, and even outliers which had to be processed. Hence, different specific steps had to be taken to cleanse and integrate these loose data sub-sets into a whole dataset. Different software tools

were used for the various tasks implied on this pre-processing step:

the various subsets and checking for redundancies;

• SQL Developer™ to pre-filter and query the various databases;

• Spyder™ interactive development environment (IDE) for Python programming language to manipulate and transform data and its structure;

• Microsoft Excel™ to easily visualize and validate data and correct punctual

• Tableau Prep™ to provide a visual and user-friendly manipulation for joining

In the end, a total of 66 variables were taken into the next stage, comprising the

**108**

specified sorter.

**Figure 7.**

errors;

following variable sets:

This stage was intended to undertake an initial analysis of the relationships within the constructed dataset, followed by the exploration of deeper analytic techniques. This procedure is necessary to build some concrete prototypes which will be tested and evaluated, thus becoming the basis for an iterative co-evolution of the final solution.

This stage unfolded in three main steps. On the first step, some descriptive statistics and the bi-variate correlations matrix were performed to better understand the variables within the dataset. This was performed using the Spearman correlation coefficient since all the variables' distributions were skewed. This is explained by the large number of zeros when there is no baggage passing through the collecting points, and the existence of a tail towards the positive integer numbers, given the relatively few but dangerous levels of high-traffic. The matrix showed some strong relations (i.e. coefficient bigger than 0.6) between some explanatory variables and the target, which is a good indicator of the possibility of building a regression model.

On the second step, the author tested some linear regressions to create an explanatory model of the sorter's functioning. From the initial dataset, encompassing every-minute counts from a one-and-a-half-month period along the 66 variables, four subsequent datasets were derived by sampling different time periods available on the initial dataset. This was aimed at mitigating the risk of overfitting, which can occur when a regression model rightfully explains the specific dataset, but it poorly performs when given a slightly different dataset. The resultant models created for each of the four datasets and their discussion gave a good overall confidence over an initial explanatory model for the sorter's functioning. Thus, considering the regression model created for the whole dataset, since it has the biggest and most complete time period, it is represented on Eq. 1.

$$\begin{array}{c} \text{Occupancy } level \text{(min)} = 1, 191 + 0, 88\* \\ \text{ } & \text{Destination } Chute151152 + 0, 899 \\ & \text{ } & \text{Destination } Chute153160 + 1, 008 \\ & \text{ } & \text{Destination } Chute172 + 1, 235 \\ & \text{ } & \text{Destination } ME125 \end{array} \tag{1}$$

Basically, this expression provides a succinct and easy way of controlling the sorter's occupancy level and consequently, its congestions. It means that the manager can concentrate his efforts on controlling these five variables (i.e. the number baggage that is inside the sorter with destination to the respective eight transfer chutes) and, consequently, he will be able to assure the sorter does not surpass critical levels which can result on congestions. By keeping a focus on these counts and by guaranteeing the respective chutes aren't jammed, the manager can reduce the sorter's congestion probability and, consequently, the number of delayed baggage.

On the third and final step, this initial prototype was presented and discussed on a group meeting, which served as a catapult to explore the possibilities of other types of analytic models for further development, as well as to diagnose some general considerations about the overall analytic journey in order to enable future easiness and agility when trying new and different analytic ventures.
