1 Introduction

The COVID-19 pandemic prompted governments to impose movement restrictions, leading to an unparalleled surge in e-commerce usage, which subsequently resulted in the increase of last-mile deliveries and pickups. According to Villa and Monzón’s (2021) estimates, the number of parcels delivered in Madrid soared by 98% from Q2 2019 to Q2 2020. This growth has attracted the attention of scholars and retailers towards research in last-mile logistics. But even before the health crisis, academic interest in the sector was already on the rise, given the surge in omnichannel retailing, increasing urbanization, changing consumer behavior and growing concern for sustainability (Olsson et al. 2019).

The rise in delivery and pickup demands has led to an increased need for couriers and a significant increase in courier workloads. According to Wang et al. (2022), excessive workloads negatively affect couriers’ quality of life and contribute to a low courier retention rate, a situation that companies must address in today’s increasingly competitive labour market.

The main objective of this study is to investigate the correlation between the workload and the unfulfilled attempted services in last-mile and first-mile reverse logistics, i.e., the incompletion rate. Furthermore, this article aims to evaluate if there is a relationship between the distance from the warehouse to the service point and the outcome of the service. The service density at each area, that is, the frequency with which deliveries or pickups are performed at an area, is also posited as a potential determinant of the operational performance of couriers, represented by their completion rates, that is, the proportion of service orders allocated to them that they are able to fulfill. To the best of our knowledge, there is no previous research on the correlations of these variables and last-mile delivery or pickup effectiveness. Through a quantitative analysis of the data of a carrier operating in the capital city of Spain, Madrid, this work delves into the potential causes of one of the most common service disruptions in the B2C last-mile and first-mile reverse logistics: service incompletions due to the couriers’ performance. It should be noted that these disruptions are not necessarily due to the couriers’ underachieving during their daily routes: the carriers’ managers play an important role in the eventual effectiveness of their workers and, therefore, on the operational efficiency of the carrier.

The rest of the article is structured as follows: Sect. 2 provides a discussion of the academic works relevant to the study, focusing mainly on previous analyses of last-mile and first-mile reverse logistics operational efficiency and effectiveness, as well as on the diverse applications of Operations Research (OR) tools aimed at their improvement; in Sect. 3, the methodology followed in the article is presented; Sect. 4 summarizes the results and findings of the study; finally, the conclusion drawn from the analysis are presented, as well as the main limitations of the study and perspective for future research in the addressed topic.

2 Literature review

Due to the significant growth of e-commerce, optimizing B2C freight delivery and pickups has become increasingly necessary for logistics operators. In Spain, there has been a rise in e-commerce traffic and subsequent last-mile logistics volume. According to Deloitte’s report (2020), before the pandemic, approximately 1.5 million e-commerce parcels were delivered daily in Spain. This number has likely increased since then. Even if there are not any published figures that determine so, it would be logical to expect that the e-commerce increase has also brought with it a rise in the number of parcel returns and, thus, in the first-mile reverse logistics volume.

However, it seems that growing market opportunities and revenues have made efficiency an afterthought for many logistics operators, especially the smaller, emerging and inexperienced ones. Carriers should nevertheless strive to achieve the improvement goals of last-mile logistics: reducing costs, increasing sustainability and improving the level of service offered to customers (Ros-McDonnell et al. 2018).

Yet, the ramifications of the health crisis on the economy, geopolitical instability, soaring fuel costs and, more broadly, an environment of inflation in the global economy have curbed the aspirations of logistics operators around the world. In such an unstable environment such as nowadays’ last-mile logistics sector, companies should opt to prioritize their operational efficiency as the most stable path to sustained competitiveness in an overcrowded market.

But what is the definition of operational efficiency in last-mile logistics? Surprisingly, research in this area is rather scarce and recent. There are three primary sources of inefficiency in last-mile logistics, as identified by Cepolina and Farina (2015):

  • Several services are independently performed at the same service points in short timeframes. Due to the limited lead times provided by carriers (and expected by consumers) and the lack of transparency and integration between carriers’ and logistics providers’ data management systems, the consolidation of multiple parcels into a single shipment is rather uncommon, especially in B2C environments.

  • Lengthy waiting times are experienced by couriers at loading and unloading sites. In highly populated and congested urban areas, the transportation of goods or items can be significantly delayed due to the search for an appropriate loading/unloading site and the travel time to the service location. In these large cities, buildings with multiple floors are the standard, which results in greater distances between the street and the destination address.

  • Inefficient management of vehicle capacity. The challenge of vehicle load capacity has been one of the most extensively researched in last-mile logistics planning, as managers face this critical constraint when assigning services to couriers. Although many methods have been proposed to optimize routes taking into account the vehicle capacity constraint - in their exhaustive review of vehicle routing problems, Braekers et al. (2016) identify capacity constraints as the most frequently considered problem feature - the actual vehicle load seems to be far from its ceiling: the load rate of vans in last-mile logistics distribution was estimated to be 40% in Spain (Deloitte 2020) and 30% in Europe (Dablanc 2007).

In brief, the high degree of fragmentation of both the logistics operators and the demand for last-mile logistics, together with the complex urban environment, results in a very high number of trips with lightly loaded vehicles, and very inefficient loading and unloading times. However, the aforementioned factors overlook a pivotal aspect of the success of delivery and pickup services in last-mile logistics, especially in a B2C environment - the service completion rate.

Certainly, the inefficient use of fleet capacity, the lack of efficient consolidation of service orders and the complex delivery conditions have a significant impact on the efficiency of last-mile logistics. However, focusing solely on these factors would be to assume that all services are performed at the first attempt, without any incidents. At this point, the concept of effectiveness comes into play. From a courier point of view, the rate of completed services represents their effectiveness, understood as a measure of goal achievement, while efficiency relates to the optimization of the use of resources (Fugate et al. 2010; Seghezzi and Mangiaracina 2020). However, there are varying definitions and measures for logistic effectiveness and efficiency, which are partly contingent on the actor being characterised (Lagin et al. 2022).

In Spain, the rate of unsuccessful first delivery attempts is estimated to be between 10 and 15%, according to Deloitte’s report from 2020. A study conducted by Lorenzo-Espejo et al. (2023) determined that the percentage of failed service attempts (including both deliveries and pickups) incurred by couriers operating in 10 Spanish provinces throughout 2021 was, on average, between 8.1% and 20%. These results, while in line with those observed in Deloitte’s report, highlight a very significant variability between the different provinces observed. The couriers working in Madrid fall on the most efficient side of that spectrum, with an average of 9.1% of uncompleted services. Of course, as shown later in the results provided in this article, this figure encompasses the failed service attempts due to the multiple issues that can arise during a last-mile delivery or pickup: missing parcel information; wrong address of the service point; customer absences at the time of delivery or pickup; service rejection due to the noncompliance of the parcel or the preestablished service conditions; service modifications after arrangements made with the consumers; a lack of time in the couriers’ workdays; incomplete or fully missing freight; or even incidences simulated by the couriers in order to cover up their poor performance. These are some of the most common incidences found in last-mile logistics and, also, on the operational records analysed in this study.

As discussed later, the main goal of this work is to investigate the effect of three potential factors (courier workload, distance from warehouse to service point, and service density at the area of the service point) on the success rate of last-mile logistics services. This success rate, or rather, the incompletion rate, is drawn from the operational records as the number of service disruptions caused by a lack of time in the couriers’ workdays. To the best of our knowledge, no previous studies in the literature address this sort of incidences, other than the methodology for incomplete services due to a lack of time proposed by Pegado-Bardayo et al. (2023). In their work, the authors posit a hybrid machine learning clustering and regression methodology that determines the most-common routes for each courier in four Spanish provinces, clusters their assigned service points and predicts the number of services that are bound to be left unattempted in each daily route. This data-driven approach is then used to identify which services are, presumably, not going to be successfully performed each day. From such information, managers can either opt to remove said services from the courier’s allocation (thus saving the unfruitful handling costs or even distributing them more adequately amongst other employees) or, in a less drastic measure, to warn the end clients that their service is not likely to be performed.

A last-mile logistics service disruption that has garnered certain interest from the academic sphere is the absence of customers at their households at the time of delivery or pickup. In this regard, researchers have focused their efforts on the prediction of customer absences, or at least into accounting for the probability of customer absences in key aspects of last-mile logistics planning. The works by Van Duin et al. (2016), Pan et al. (2017), Florio et al. (2018), Praet and Martens (2020), Özarık et al. (2021), Kandula et al. (2021), and Seghezzi and Mangiaracina (2023) must be acknowledged regarding this scope of research.

Apart from the prediction of customer absences, there is a clear lack of research regarding the other multiple potential causes of service incompletions. In this regard, some recent studies have simply aimed at the prediction of the attainable courier workload, not necessarily devoting their efforts to analysing the determinants of said workloads, but by attempting to estimate service time. For example, Guo et al. (2023) develop a hybrid methodology that uses learning-based models to predict couriers’ routes service time and uses said estimates to allocate the different delivery areas to the couriers through a zone-partition algorithm. Ultimately, the goal is to allocate delivery zones so that the couriers’ routes are equitable, using the required workload (measured as the time needed to fulfill the services of an allocated delivery area) as the metric to assess fairness.

Along these lines, Liu et al. (2021) address the order assignment problem in food delivery services, again, combining prediction models to estimate the completion time of the assigned services with mathematical optimization and heuristic techniques.

Lastly, Song et al. (2019) address specifically last mile delivery service time understood as the time period from when the courier parks the vehicle to when it resumes its route, that is, the time required to walk to a household, perform the delivery, and walk back to the parking spot - it should be noted that this is a variable procedure and that several driving-walking tours strategies can be implemented. The authors refer to this process as the “last yard” of the supply chain. The last-yard service time can certainly present a high degree of uncertainty due to both expected (for example, having to walk a long distance because of parking spot scarcity) and unexpected factors (such as elevator breakdowns in tall buildings). Notably, the two experiments conducted by the authors using real-world data show that their machine learning based regression models for the prediction of last-yard service time, which utilize parcel-related and spatio-temporal information, present a significant degree of error: the mean absolute percentage error (MAPE) of their predictions ranges from 64% in the first case-study to 394% in the second experiment, which equates to a mean absolute error of over 10 min. Still, since an error reduction of around 14% is achieved by the K-nearest-neighbours model when compared to planners’ manual estimations, the proposal shows promise and could be enhanced by future works.

Additionally, certain researchers have underscored the need for improved efficiency in the last-mile logistics sector. Muñuzuri et al. (2012) emphasize the common perception of all stakeholders that corrective actions need to be taken to address the inefficiency of the urban delivery system in Spain. This is particularly critical in certain areas with a higher tourist appeal, dense commercial activity and narrow street grids. The authors state that in Spain, urban freight delivery faces challenges due to poor driving behavior and complex urban morphology, as well as insufficient enforcement, lack of cooperation and knowledge transfer between cities, and inconsistent and outdated regulations. This is particularly the case of the metropolitan area of Madrid, which expands to almost the totality of the Community of Madrid, with 6,751,251 inhabitants (Instituto Nacional de Estadística 2021), chosen as the basis for the case study of this research.

Finally, regarding the contribution of this article, it must be borne in mind that, while the literature review has focused on works related to last-mile logistics, the operational records analysed include pickups as well and, thus, this study also addresses first-mile reverse logistics, which refers to the collections made by the carriers at the customers’ households, usually as a result of an unsatisfactory e-commerce transaction. First-mile reverse logistics constitutes the initial stage of the reverse flows, arising from returns made by customers (Agnusdei et al. 2022), and represents a significant share of carriers’ business activity. Moreover, the first mile is hardly free from the problems of last-mile logistics: carriers providing first-mile services tend to suffer from high empty haulage rates and poor efficiency (Wang and Huang 2021).

Furthermore, the work presented in this article aims to attract attention to the prediction and control of courier workload and completion rate, that is, to the maximum feasible number of services that can be assigned to a courier. This article contributes to the scarce discussion on the topic by studying the correlation between certain operational variables and the incompletion rate, which can be seen as a first step to develop comprehensive models able to accurately predict adequate workloads. This essential indicator has been mostly unaddressed in the literature, and, in practice, it is a somewhat common strategy amongst carriers to overload their couriers’ route assignments. The belief is that, even if there is a high possibility of not fulfilling a significant proportion of the services, the chance of being able to do so compensates the excess costs of loaded but unfulfilled services. In truth, this practice masks an inefficient strategy with several detriments for carriers: additional loading/unloading and parcel handling costs; fuel costs; environmental impact and costs; customer dissatisfaction if they are warned of a pending service that is finally not performed; or staffing issues due to inaccurate workload estimation.

Therefore, accurately predicting the maximum achievable workload per courier should be a fundamental step in some the most common applications of OR to last-mile and first-mile reverse logistics: warehouse or distribution hub location; service-courier allocation; and routing. While academics have produced very proficient methodologies to optimize these tactical and operational decisions, it is our belief that more attention should be devoted to such a fundamental part of the data processing for said problems as is workload estimation. Therefore, we posit investigating the effects of operational variables on courier effectiveness given said workloads as the seminal step for this type of prediction.

3 Methodology

In this section, an account of the materials and methods employed in the study is provided. The dataset encompasses information concerning deliveries and collections made between January and September 2021 by a logistics operator based in Madrid. The carrier performs B2C deliveries and collections for multiple brands and e-commerce stores.

The data included in the analysis is derived from the operational records of the couriers performing the delivery and collection services. When couriers are allocated a delivery or collection order by their managers, they are tasked with registering the completion location and time of their assigned services. Additionally, if any service disruption takes place, they must record it, specifying its cause. This information is stored in the carrier’s database, and has to be checked later by the carriers’ agents to determine the validity of such notifications.

The available information encompasses the following variables for all the service attempts performed by the carrier in Madrid in the discussed period:

  • Courier ID,

  • Service type (delivery or collection),

  • GPS coordinates of the service location,

  • GPS coordinates of the warehouse,

  • Service completion status,

  • Disruption cause (only if attempt is failed).

These data are used to compute the variables relevant to the analysis, the couriers’ daily assigned workloads and incompletion rates; the service attempts assigned in a postal code, the distance from the warehouse to the centroid of the postal code and the incompletion rates at the postal codes. A more detailed description of the calculation of these variables is provided when the design of the analysis is discussed later in this section.

It is important to note that, while various disruptions may hinder completing a service, we solely consider those that couriers themselves have specified as “failed services” due to insufficient time as our target in this work. This has been done with the main goal of the study in mind: analysing the correlation of courier workload, distance from warehouse to service point, and service density at the area of the service point with the incompletion rate of services. These variables can be presumed to have certain influence on the number of services that courier cannot perform due to a lack of time in their workdays. Perhaps the remaining services are too far from the warehouse, to which the courier must return before a certain time of day. Also, an excessive workload might force couriers to leave some services unperformed. Finally, couriers might be more inexperienced in areas in which the number of parcels delivered or collected is lower. These are precisely the questions that should be addressed with this work, and which justify the use of the proposed incompletion rate.

In terms of the couriers’ routes, it is worth noting that they alternate between deliveries and pickups without a predetermined pattern. For this analysis, we treat both deliveries and collections equally as services, without distinguishing between them.

The study undertakes two main analyses. Firstly, we examine the correlation between the daily workload allocated and the incompletion rates of the couriers. Since a significant number of couriers were not yet available in the first three months, we have computed the average daily workload for each employee in the second quarter and the third quarter of the year, i.e. from April to June (Q2) and from July to September (Q3), in order to base our first analysis on a stable workforce. However, it should be noted that the first quarter is included in the rest of the analyses, as the focus is then set on the postal codes.

For the calculation of the daily assigned workload, we have computed the number of delivery and collection service attempts allocated to each courier each day. Of course, the same delivery or collection can be attempted in subsequent dates if a service disruption takes place. Thus, the distinction between service attempt and service order.

Moreover, we calculate the proportion of unsuccessful services caused by insufficient time (herein referred to as the incompletion rate or IR) for each courier by dividing the number of delivery and pickup efforts that failed due to time constraints by the total number of attempts made by each employee. Again, it is important to note that the IR should be based on failed attempts rather than failed services, since the vast majority of services are bound to be completed at some point.

Subsequently, the correlation between the IR and the distance is analysed. In order to study this relationship, we have clustered the deliveries and collections made between January and September in each postcode. This enables us to compute the distance between the centroid of each postal code and the corresponding logistics operator’s warehouse, and then to associate it with the couriers’ performance records for each day. The postal code incompletion rate is obtained as the proportion of service attempts located inside the postal code which are left unfulfilled due to insufficient time, as is the case of the courier incompletion rate.

The centroid is determined as a point within each postal code that can be obtained as the average of the coordinates of the points in the perimeter of the area. The Euclidean distance between the coordinates of these centroids and the corresponding warehouse is then calculated. However, this measure does not strictly represent the distance travelled by road between these points. To address this problem, we implemented a coefficient computed as the ratio between the actual road distance and the Euclidean distance in a sample of postal codes. Our calculations were conducted using SPSS Statistics version 26, and the distances were computed using the Transact-SQL geometry methods in Microsoft SQL Server.

Finally, the correlation between the postal code IR and the average number of daily assigned attempts at each postal code has also been computed. For the calculation of the daily assigned attempts at a postal code, we have computed the number of delivery and collection service attempts allocated to the courier inside each postal code in the analysed period, and obtained the average of all the days in the dataset.

4 Results and discussion

As previously stated, there are two main aspects to our analysis: firstly, we examine the correlation between the daily workload assigned to the couriers and their incompletion rates (Sect. 4.1); secondly, we analyse the correlation between the IR and two other characteristics of last-mile and first-mile reverse logistics: the distance between the service points and the warehouse, and the service density of an area (Sect. 4.2).

4.1 Correlation between workloads and Incompletion Rates

To investigate the correlation between workloads and courier IRs, we have analysed the Q2 and Q3 data for 2021. We have summarized the relevant data in Table 1.

Table 1 Summary of data regarding service attempts made in Q2 and Q3
Full size table

Table 1 shows that the service volume of the logistics operator increased by 60.2% from Q2 to Q3, while the IR remained almost constant. Given that the workforce increased by only 5 couriers, this growth in volume can be attributed to the fact that the average number of days worked by each courier rose significantly from Q2 to Q3.

Figure 1 displays a dispersion plot illustrating the average daily workload against the IR for each employee in Q2 and Q3. In addition, the trend lines for each series are depicted in their respective colours.

Fig. 1
figure 1

Courier incompletion rate vs. average daily courier workload in Q2 and Q3

Full size image

Figure 1 demonstrates that there is indeed a direct correlation between the daily workload of each courier and their performance in terms of IRs. To support this claim, the Pearson’s correlation coefficient was calculated between the average daily workload and the IRs for Q2 and Q3 and for the two quarters combined. When combining the two quarters, the incompletion rate is calculated as the number of incomplete services performed in the two quarters over the number of days worked in the whole six months. Likewise, the same period is used to obtain the average daily workload refers to the same period. Table 2 shows the results of said analysis, along with the P-values calculated through a two-tailed significance test.:

Table 2 Pearson’s correlation coefficients between couriers’ average daily workloads and incompletion rates in Q2, Q3 and overall
Full size table

Table 2 shows a significant (95% confidence level) but moderate direct correlation between average daily workload and IR for Q2, Q3 and the combined data for both periods. These results indicate that a higher IR can be expected when couriers are assigned a heavier workload. In any case, the need to differentiate between correlation and causality should be stressed. The results show a significant correlation between both variables, but this does not imply that high workload is the determining cause of the incompletion rate.

4.2 Correlation between distances, daily attempts and incompletion rates

As mentioned in Sect. 2, the second analysis presented in this article focuses on the correlation between distance and IR. Furthermore, an additional analysis is carried out to determine whether the frequency with which postcodes are visited correlates with the IR experimented in them.

Figure 2 illustrates these trends. In said figure, each dot represents a postal code with at least 1,000 deliveries or collections. In total, there are 121 postal codes that reach such threshold. The shades of grey represent the incompletion rate for all the services corresponding to each postal code in the analysed timespan, which has been divided into four categories for representation purposes: postal codes with an incompletion rate between 0% and 5% (in the lightest shade); between 5% and 10%; between 10% and 15%; and between 15% and 20% (in the darkest shade). Furthermore, the horizontal axis represents the distance from the warehouse to each of the postal codes. Finally, the vertical axis shows the average number of daily attempts made at each postal code.

Fig. 2
figure 2

Incompletion rate per postal code. Average number of daily attempts at the postal code vs. Distance from warehouse to postal code

Full size image

Figure 2 illustrates how the majority of postal codes with less than 5% IR are located within 20 km of their respective warehouses. In contrast, almost all postcodes with an IR above 15% are situated more than 20 km from their respective warehouses. Nevertheless, it may be difficult to derive any strong trends from these results, as the postal codes with IRs between 5 and 15% are apparently evenly distributed.

In order to explore this possible correlation in more detail, Pearson’s correlation coefficients (together with the P-values indicating the significance level) have been calculated for the IR with the average number of daily attempts on the postal code and with the distances between warehouses and postal codes. The results of said tests are summarized in Table 3:

Table 3 Pearson’s correlation coefficients between distance and IR, and average daily attempts and IR.
Full size table

As one might expect based on Fig. 2, the correlation between IR and distance is weak and only significant at a 90% confidence level. On the other hand, the results show a slightly stronger correlation between the IR and the number of daily attempts, which is likewise significant at a higher confidence level (95%). Based on these results, it can be concluded that the further away a postal code is from the warehouse, and the more services a postal code receives per day, the more likely it is that the services at such postal code will be left unfulfilled due to lack of time. The cause of these correlations can be hard to pinpoint. Obviously, serving farther regions implies planning longer routes. In these longer routes, couriers are bound to encounter more obstacles, mainly, a higher exposure to traffic congestion, accidents and breakdowns. Still, one would expect the managers of the carriers to take these potential issues into account when planning routes including service points far from the warehouse.

Regarding the negative correlation between the daily attempts at a postal code and the completion rate of the services demanded from it, again, no definitive cause can be identified. On the one hand, couriers should have more experience in delivering and collecting parcels in areas they often visit (that is, assuming they are mainly assigned similar routes every workday). On the other hand, a higher number of parcels to be delivered in the same area can lead to tougher routing decisions, particularly when the couriers are allowed to decide and follow their own routes. These manual “routing” decisions are mostly based on experience, but frequently fail to grasp the multiple factors involved in such a complex decision-making task: traffic and weather conditions, delivery and collection time windows, order priority, etc. This issue can be solved through the use of vehicle routing tools based on accurate and even real-time data and optimization techniques. However, these routing recommendations are severely shortsighted if they fail to acknowledge, precisely, the number of parcels that couriers are not able to deliver due to a lack of time. In this regard, the work by Pegado-Bardayo et al. (2023), which provides a pathway to identify these undelivered or uncollected parcels, should be highlighted.

With that in mind, it must be noted that the correlations of both pairs of variables, while statistically significant, are admittedly weak. The results regarding the distance and the daily attempts should therefore be interpreted with caution.

5 Conclusion

The findings of this study indicate significant direct correlations between last-mile delivery and first-mile collection incompletion rates and three operational variables: the workload assigned to the courier, the number of service attempts within the postal code, and the distance between the postal code and the warehouse. For the first pair of variables, the results show only a moderate correlation between the assigned workloads and the incompletion rates. Concerning the other two pairs of variables, the coefficients indicate weak correlations.

The most important conclusion and managerial implication drawn from this work has to address the correlation between the couriers’ workload and their incompletion rates, both for its higher statistical significance and correlation strength, and for its managerial relevance. Couriers can certainly be expected to incur in higher incompletion rates as the workloads assigned to them increase. Thus, managers should improve their efforts in the allocation of services to their workforce, mainly with two very different objectives: firstly, having to load and unload parcels into the vehicles that will be most likely not delivered is a non-trivial handling cost for the carrier. Of course, additional fuel expense due to unnecessary weight in the vehicles should also be accounted for, not only for its economic consequence, but also for its environmental impact. Secondly, as already discussed, Wang et al. (2022) state that excessive workloads impact couriers’ quality of life, which now becomes and even more harmful and unjustifiable malpractice given the unfruitfulness of such excess in allocated workload.

Of course, it must be kept in mind that the performed analyses are only able to identify correlations, rather than causality. Last-mile and first-mile reverse logistics truly involve a multitude of factors, which makes identifying isolated causal relationships quite difficult.

In terms of areas for future research, the results highlight the already discussed need to improve the assessment of the workload achievable by each courier. Isolating determining factors of courier effectiveness and creating more complex prediction models should play an important role in workload prediction. In time, these predictions can be expected to be integrated with most of the decision making process at the tactical and operational level in last-mile and first-mile reverse logistics. In this regard, more attention should be paid to analysing couriers’ workload capacity individually: of course, variables related to the predetermined conditions of couriers routes, such as the distance from the service points warehouse, the assigned workload or even the urban typology of the delivery area. However, with equal external conditions, different attainable workloads could be expected between two couriers. Their personal routing decisions (since route plans are very frequently not issued by the carrier) or even their contractual situation (being part of the stable workforce or, conversely, being paid on the basis of the parcels delivered/collected as an outsourced courier) can impact the number of services a courier is able to perform in their daily routes.

Finally, from an economic and cost perspective, the findings of this study accentuate the importance of calculating and accounting for the additional costs sustained when accepting and loading service orders with potential to be unattended. For example, based on the results, service points situated far from the warehouse could generate even greater costs than simply that of distance, as a lower completion rate on them is expected. In a competitive market with relatively low profit margins (Allen et al. 2018), last-mile and first-mile reverse logistics carriers should incorporate these cost considerations in their decision-making process.