A digital infrastructure design for physical internet connection in favelas

(1)

A digital infrastructure design for

physical internet connection in

favelas

In partial fulfillment of the requirements for the Master program of

Technology and Operations Management

Author name:

Student number:

Supervisor:

Second supervisor:

Gerke Julius Schaap

S2405261

Dr. N.B. Szirbik

Dr. J.A.C. Bokhorst

(2)

Abstract

The physical internet is proposed to integrate the fragmented logistical networks of today into a global hyperconnected network, which is said to improve efficiency by an order of magnitude. While it is evident that technological foundations are being laid that can eventually support the physical internet throughout the devel-oped world, these developments seem to be absent in parts of the developing world, most notably in the favelas of Brazil. This research proposes designs that show how internet connection and mobile phone adoption may provide a sufficient baseline to reap benefits of the physical internet, also for these marginalised societies. It is argued that this does not only improve the efficiency of distributing basic commodi-ties into these areas, but enhanced digital capabilicommodi-ties may also increase influence and improve transparency of food supply into these areas, resulting in better value for money for these marginalised societies.

Acknowledgements

The writing of this thesis was in part a collaborative effort with two fellow students. Bart Louwerse and Patricia Assen are thanked for their willingness to share ideas, and will be referred to by name for parts of their research that are incorporated in this thesis. Secondly, the thesis supervisor Dr. Szirbik is thanked for his suggestions that helped shape this thesis during brainstorming sessions with us students. Finally, Ana Clara Cassanti who has experienced the favelas first hand, is thanked for her insights given while attending brainstorming sessions, and for pointing out useful sources for this research.

(3)

List of Figures

1 single tier, single CDC city logistics system (Crainic and Montreuil, 2016) 11

2 Overview of last mile delivery models for nanostores (Fransoo et al., 2017) 14

3 Theoretical framework . . . 16

4 3 cycles of DSR (Hevner, 2007) . . . 17

5 Regulative cycle (Wieringa, 2009) . . . 18

6 Current practice of supplying nanostores . . . 19

7 Envisioned practice of supplying nanostores . . . 20

8 Context function A1 hierarchy . . . 31

9 Context function A1 interaction diagram . . . 32

10 Function 0 hierarchy . . . 33

11 Function 0 interaction diagram . . . 34

14 Function 12 hierarchy. . . 37

(6)

1 Introduction

Montreuil et al.(2012b)define the physical internet(PI) as ”An open global logistics sys-tem founded on physical, digital and operational interconnectivity through encapsulation, interfaces and protocols”. The PI is proposed to improve the economic, environmental and social efficiency of global logistics by an order of magnitude. Similar to how the digital internet has connected computers, the PI will connect logistical networks that are largely fragmented nowadays, into a single hyperconnected network. (Crainic and Montreuil, 2016).

The PI is envisioned to be fully operational by 2050 (Alice, 2015), but current research is already producing architectures with increasing levels of specificity that may bring this futuristic vision closer to realisation. Also in practice, enabling technologies are currently being deployed in logistics that can be considered predecessors for the physical internet. Some examples of these technologies are track and trace technologies that digitise trajectories of freight, planning and optimisation tools that determine optimal routes, and cloud based services that improve collaboration and integration of supply chains, to mention a few.

Paradoxically however, the regions that stand to gain the most from such innovations in logistics, are the regions that are at risk to be left behind. The favelas of Brazil are an example of such marginalised societies. In these areas, the basic digital and physical preconditions seem to be absent to take advantage of these innovations, especially for a connection to the physial internet in the future. As this research will point out, the current process of food distribution into favelas is also highly inefficient. This results in a higher price for food and basic commodities for these already marginalised societies. This research investigates the progress of internet and smartphone adoption in favelas, and shows that these technologies provide a basic digital infrastructure that can support innovation in the food distribution process, including a PI connection in the future. The favelas of brazil are by no means the only regions where uncontrolled urbanization has lead to adverse social, economic, and environmental effects (Vieira et al., 2015), but they are arguably the most problematic given the weak formal authority and lack of institutions (Fernandes et al.,2019). Nevertheless, elements of designs proposed by this research may still be of use for other areas.

(7)

2 Theoretical background

2.1 The Physical Internet

Seminal work conceptualising the physical internet (PI) states that ”[PI] exploits enabling concepts of standardized, modular and smart containers, as well as the universal inter-connectivity of logistics networks and services.” (Montreuil et al., 2010). The authors mention the physical elements of the PI to consist of PI containers, PI nodes and PI movers. The functioning of the PI draws upon the analogy of the digital internet, where PI containers behave like data packets in the digital internet, and PI nodes perform the functions of connection points in the digital internet (routers, modems, etc). However, the physical reality of moving PI containers by PI movers adds complexity compared to the digital internet. Nevertheless, by developing protocols and container standards, the goal of the PI is to shift from a fragmented, hard to optimize organization to an open, distributed organization (Montreuil et al., 2010), which should cut energy consumption and associated costs of global logistics by an order of magnitude (Montreuil,2011). Based on these seminal works, four characteristics are identified that capture the essence of the PI, which will be elaborated on in the remainder of this section.

Autonomy

Montreuil et al. (2012a) proposes the Open Logistics Interconnection (OLI) model, in-spired by the Open System Interconnection (OSI) model, a predecessor for the model that governs communication between computers in the internet of today. The OSI model allowed communication between heterogeneous devices because of a layered structure. ”Layering aims to ensure independence of each layer by defining services provided by a layer to the next higher layer, independent of how these services are performed” (( Zim-mermann, 1980) as cited in (Montreuil et al., 2012a)). Just like the OSI has achieved for the digital internet, a layered communication structure such as the OLI enables the PI to function in an autonomous way. As envisioned by the authors, users and providers of transport services interact with eachother in the 7th layer: the logistics web layer. Matches between offers and requests are then communicated across layers to provide the appropriate shipping services facilitated by a large variety of heterogeneous PI transporta-tion means(rail, truck, inner city transport networks, etc). Rules and protocols governing communication between layers should enable smooth and autonomous handling of ship-ments over and across different logistic networks.

Decentralisation

(8)

Openness

Related to the decentralised structure,Montreuil et al.(2012a) propose that the PI is open to be used by all providers and users of logistic services. An open structure stimulates shared infrastructure of logistics, that results in an interconnected, or hyper connected structure Crainic and Montreuil(2016). Shared logistic infrastructure and openness as-sures that logistic processes are optimised as a whole, rather than sub-optimised for separate networks.

Modularity

Containerised shipping has already radically improved logistic efficiency during the second half of the 20th century, because standardised sizes allowed to build infrastructure to efficiently handle containers at ports all around the world. The PI will further exploit opportunities stemming from standardisation and modularity. Design suggestions have already been put forward for different standardized sizes that can be stacked together efficiently such as the 3 tier system proposed by Montreuil et al.(2014).

2.2 Digital technologies in logistics

The PI of the future requires large amount of data to be gathered, processed and trans-mitted. The presence of digital technologies to facilitate this are already increasing in modern logistics. This section provides an overview of digital technologies in logistics used today, that also serve as potential technologies to consider for designs later on.

2.2.1 Data gathering

Within the PI, a large variety of digitised data is required. Unlike data that originates in a digital setting such as online order placements, there is also data that first requires digitisation. Data gathering in this sense refers to data that does not originate in a digital form, but should first be captured from physical processes. This definition of data gathering limits the technologies involved in this category to sensors, and an initial connection to a digital data transmission technology.

(9)

over barcoding.(Probst et al., 2015). Near Field Communication (NFC) is another example of an advanced tracking system, better suited for situations where a physical touch is required for security purposes(Probst et al., 2015). In the envisioned PI, all PI elements (containers, transporters, handlers) are expected to be outfitted with track and tracing technology, along with the additional sensors needed to perform additional data gathering functions. Connecting items to the internet in this way, is referred to as the Internet of Things (IoT). Currently the presence of track and tracing technology is already increasing in supply chains, but main obstacles for implementation are a lack of standardization, and security and privacy concerns(Probst et al., 2015). Standardization refers to standards for data and communication across companies, sectors and countries, but also the range of frequencies in the radio spectrum that RFID is allowed to transmit data in, which differs across countries.

2.2.2 Data processing

Data processing here refers to a broad range of techniques used to create useful infor-mation from large amounts of digital data(Big data). Data analytics, Data mining and Data fusion are all terms referring to the practice of extracting usefull data from large amounts of unstructured data by filtering, reorganising or combining data from different sources. Within the PI, Intelligent algorithms should be able to perform this data processing, and easy to use software should be developed to facilitate rapid human interaction with the PI (Alice, 2015)

(10)

helps in understanding relationships and causality effects between variables. (Wang et al.,

2016). Simulation uses data as inputs for predefined concepts such as targets, boundaries, entities and constraints, among others, and returns outcomes as new data. Optimization techniques help improve the accuracy of demand forecasting and supply chain planning. Even though the use of data analytics is increasing in logistics and supply chain man-agement, current techniques are inadequate to attain the maximal value from big data. Problems include inability to deal with massive sample size, high dimensionality, noise accumulation, incidental homogeneity of data, to mention just a few problems. (Wang et al.,2016;Fan et al.,2014). This requires more robust techniques for data analytics for the future. This section concludes with a brief introduction into Machine Learning, which among other means of Artificial Intelligence (AI), may help further develop the field of data analytics in logistics.

Machine learning within supply chains relies on algorithms to quickly pinpoint influential factors to a supply networks success, while constantly learning in the process. ”These algorithms iteratively query data to find core set of factors with the greatest predictive value”(Columbus,2018). The full potential of machine learning and AI seems enormous, and by no means restricted to logistics and supply chains, but improved demand forecast-ing usforecast-ing diverse datasets, improved collaboration in supply chain networks, and more effective asset management are among the potential improvements machine learning can make in logistics (Columbus, 2018). The benefits of machine learning and AI seem to be limited only by the extend to which organisations collaborate and share data with eachother. For this reason, key providers of machine learning and AI solutions also stim-ulate collaboration in their offering of solutions. IBM is an example of such an provider of machine learning and AI solutions for supply chains. IBMs Watson Supply Chain offers transparent, intelligent and predictive supply chains solutions(IBM,2018b), enhanced by tools for B2B collaboration(IBM,2018a).

2.2.3 Data transmission

Related to the transmission of data, it is also important where data processing and decision making based on this data occurs. The OLI model allows different options for the degree of centralisation/decentralisation of this intelligence within the PI. Cloud computing is a technology that refers to centralised data processing and decision making using the internet, where Edge computing refers to technology where this intelligence occurs decentralised, at the source where data is gathered. The main advantage of cloud computing is avoiding the need for computational power at all PI means (Containers, transporter, handlers, etc.). However, the need for edge based systems in the envisioned PI is apparent (Alice, 2015) as it decreases the need for data transmission, eliminating lag time and thus improves real-time decision making (HP, 2018).

(11)

Positioning System) is currently a commonly used technology. GPS however only al-lows to receive information from satellites to determine locations, but not to transmit data back (Hoffman, 2016). In order to transmit location or any other data from mobile units to remote processing units (or to the cloud), cellular connections can be used (Li et al., 2010). Cellular connections use the same networks as mobile phones, and can refer to to different generations of connections (2g, GPRS, 3g, 4g, and upcomming 5g). Cellular connections provide transmission opportunities as long as there is a cell tower within a maximum range of 35 to 70 Km.(Markgraf, 2018). For connection at remote locations such as at sea, special satellite connections can be used, such as provided by Inmarsat(Inmarsat,2018).

Vodafone (2017) provides a broader overview of connection types, with key performance dimensions being power consumption and range. The authors argue that in order for In-ternet of Things (IoT) technology to truly gain momentum, so called Low Power Wide Area technologies(LPWA) should be developed. Current communication technologies such as cellular are considered as too power consuming for mobile devices. several com-peting (LPWA) technologies are currently under development, Sigfox, LoRa and NB-IoT are considered front runners in LPWA (Vodafone, 2017).

A final technology worth mentioning is Blockchain technology. Not technically a transmission technology by itself, but blockchain uses cryptography and decentralised storage of data which increases transparency and security of data. Access to data is only granted when there is a transactional relationship between parties, or permission has been actively granted. Furthermore, data quality is said to be improved because of decentralized storage, which avoids error propagation(Hofman et al.,2017). IBM provides an early use case of Blockchain technology in logistics, offering solutions for Maersk, a global leader in transport and logistics. The solution is said to achieve end to end visibility for all parties involved in the shipping process, based on their level of permission. This helps to reduce fraud and errors, reduce time products spend in transit, improve inventory management and ultimately reduce waste and costs. (IBM, 2017).

2.3 Current best practice of last mile delivery

(12)

Figure 1: single tier, single CDC city logistics system (Crainic and Montreuil, 2016)

for re-consolidation of freight. Adding tiers becomes increasingly attractive the more freight is required to access city centres, and the larger cities are. The value of adding tiers will be further enhanced by the PI, as navigation of freight through cross-docking points will be seamless within the PI (Crainic and Montreuil, 2016).

2.4 Current last mile delivery of food supply in favelas

Vieira et al.(2015) provides insights on the wide range of issues that logistics in megacities face because of disorganised population growth. These issues can first of all be explained as an infrastructural problem: narrow and heavily congested streets that are incapable of properly supporting freight distribution needs. However, the situation is made worse by the lack of collaboration for delivering and receiving goods, which is also associated with the lack of technology for data interchange. Also security risks from cargo theft and robbery add to the problems. These problems create a domino effect that push up the prices for goods and services (Vieira et al., 2015).

(13)

to avoid a tiered system is especially strong for the informal context of the favelas as ex-plained by Fransoo et al.(2017). This tendency to bypass distribution tiers implies that small carriers from the outskirt of the city cover large distances before reaching favelas, also worsening the traffic conditions on the periphery of favelas. For this reason, the remainder of this chapter is devoted to explaining the current reality of inefficient food distribution into favelas, and the mechanisms at play that cause these inefficiencies. Closely related to food distribution, are the dominant retail channels for food in favelas.

Fransoo et al. (2017) explains the different reality of retailing in emerging megacities compared to most develop countries. The dominance of supermarkets and hypermarkets was not achieved for most developing countries. These countries mostly rely on so called nanostores: very small stores typically serving 100 to 200 people, with limited assort-ments. Figures on market shares were not given for Brazil, but figures for Argentina show an initial increase in market share for supermarkets, increasing from 27% in 1984 to 57% in 1999. However, this trend had completely reversed and a decade later nanostores again represented 67% market share in retail of food products (Fransoo et al., 2017). Mobility constraints of people are an important factor mentioned for the dominance of small nanostores present at every corner, compared to large supermarkets that require cars to access. While Fransoo et al. (2017) make no mention of favelas explicitly, they mention poverty and informality to be contributing factors for the dominance of nanos-tores, therefore nanostores are assumed to be the exclusive channel for food retail in favelas.

(14)

and wholesalers, that have extra tiers in their distribution channels. Small vehicles drive large distances from the DC in the outskirts of the city to deliver relatively small deliveries, to few nanostores. Moreover, trips are frequently made only for order processing purposes, as communcation technology to process orders is often absent. this adds vehicles millage, without any distribution taking place. However, the most primitive direct service model is referred to as on-board sales. In this model, no orders are being processed in advance, but vehicles drive around carrying (limited variety) of stock that they intend to sell when they reach nanostores, frequently visiting nanostores where no sales are made.

Fransoo et al.(2017) provides an overview of all service types used to service nanostores, as illustrated in figure 2. These models will be briefly elaborated on, it is assumed that either on-board delivery or pre-sales + direct store delivery are currently dominant methods for supplying nanostores in favelas.

On-board sales: this sales model relies on sales being decided upon arrival of the courier at a nanostore. The advantage of this model is that all 5 sales functions can be fulfilled at the same time without any prior communication, even though demand generation might be limited to what the courier is carrying. From a logistics point of view this is the least efficient model as couriers take longer stops at each nanostore to fulfill all functions, and may end up visiting stores without making a sale.

Pre-sales + Direct store delivery: CPG manufacturers still use their own dedicated couriers but communicate with shop owners in advance about deals and order processing. This communication can be done by phone, but is still also performed through store visits by sales agents. Physical presence of sales agents may be required because of limited communication technology, but also because agents want to check stores to assure positive brand presentation.

Pre-sales + Distributor: Under this model, CPG manufacturers use a third party distributor to perform the physical distribution of products, but still contact nanostores themselves for order processing. A third party distributor may benefit from economies of scale which increases last mile efficiency.

(15)

(16)

3 Design issues and framework

This thesis is written from a pragmatic point of view that PI last mile delivery requires not only high level conceptual feasibility, but also requires feasible digital and physical infrastructures for specific areas such as the favelas in Brazil. The physical infrastructure is considered to be the mayor constraining factor for how close goods can get to their final point of consumption in favelas using the PI, due to security concerns for PI containers and transporters. Even though the PI is envisioned to be an open network where matchmaking occurs between transport requests and offers for logistic services, it is assumed that route safety and accessibility standards maintained by PI logistic providers may lead to favelas that are not covered by the future PI. Alternatively, this research along with two other master thesis’s written in parallel, intend to design solutions that enable PI transportation of goods close to the final destination, but with feasible options for the very last mile delivery occurring outside the PI.

The basis for designs will follow the design of the City Logistics Smart Rack Micro-Hub (CLSR micro-hub) proposed byVries(2018). The micro hub was designed as an extension on the smart locker bank (SLB) proposed by Faugere and Montreuil (2016) that can be considered a PI compatible last mile solution. The SLB functions as a drop of point for e-commerce packages that can be claimed by the recipient with an authorization code. The mirco hub proposed byVries(2018) extended this design by allowing larger shipments to be stored, that may be used by businesses to temporarily store incoming stock.

(17)

(18)

4 Methodology

4.1 Design Science Research

Design Science Research(DSR) is the methodology that guides this research. Hevner

(2007) describes DSR to consist of two activities: building designs, artifacts and processes, and the validation of these constructs. The author mentions the design cycle as the mechanism that iterates between using previous evaluations for new designs. Intertwined with the design cycle are the relevance cycle, and the rigor cycle. The relevance cycle provides motivation for the design process by the desire to make an improvement in a certain environment. The rigor cycle provides capabilities to the design process by providing knowledge, which can come in all forms from highly abstracted theories, to context specific knowledge/ designs. The design cycle can therefore be seen as the link that provides artifacts to improve certain contexts, and to evaluate on its performance.

Figure 4: 3 cycles of DSR (Hevner,2007)

Evaluation entails actual implementation of designed solutions, to study its effects. Since implementation is beyond the capabilities of this research, the regulative cycle ofWieringa

(2009) provides some more guidance on the solution design process that precedes imple-mentation and evaluation. The regulative cycle consists of 4 steps:

1. Problem investigation/ evaluation of how previous implemented designs have per-formed

2. Solution design 3. Design validation

4. Solution implementation

(19)

design, research questions are formulated and data is collected.

Figure 5: Regulative cycle (Wieringa, 2009)

4.2 Research questions and data collection for solution design

Research questions:

1. What digital technologies can be relevant to improve micro hub operations and fea-sibility in favelas?

2. What is the current state of adoption of these technologies in favelas?

RQ 1 is partly answered by brainstorming sessions with 2 fellow master thesis students performing parallel research, and the thesis supervisor. These sessions have resulted in an envisioned operational process for supplying nanostore, which provides leads for how relevant digital technologies may apply. Additionally, literature investigation as well as interviews with a third party logistical service provider (Fietskoerier) are used to scan for potentially relevant digital technologies. Lastly, a stakeholder analysis is conducted to capture additional digital functionality that may improve success of the micro hub implementation in favelas, focusing also on strategic aspects of technologies.

RQ 2 investigates current states of adoption of enabling technologies to support these digital functionalities, based on literature research.

4.3 Design validation

(20)

5 Analysis

5.1 Current operational process of supplying nanostores

In line with the direct service models discussed, an interpretation of the current oper-ational process of supplying nanostores is expressed in a business process model and notation scheme (BPMN), that has been constructed in parallel research. The BPMN schema can be found in appendix A, with courtesy of Bart Louwerse. The BPMN shows the operational process from the perspective of a single supplier. Figure 6 is added to show that different suppliers compete over the region with separate operations. Figure 6 shows the positioning of a favela in an highly abstracted fashion, appendix C is added to show the actual spread of favelas in the urban landscape of Rio. A brief summary of key points of the current operational process now follows.

1. Nanostore orders are collected by face to face contact contact with agents of CPG companies either during delivery of a previous order, or during a round trip devoted to only taking orders.

2. Agents make round trips using small vehicles like auto rickshaws (tuk-tuk) between a DC on the outskirt of the city, and a handfull of nanostores in the favela. In case of on-board sales the sales representative does not carry orders for specific nanostores, but rather carries stock that it hopes to sell when they reach the nanostores. 3. Payments are exclusively in cash, and cash is passed on from nanostores to sales

representatives to the CPG manufacturer at the DC.

(21)

5.2 Envisioned operational process of supplying nanostores

The envisioned operational process is visualised in detail in another BPMN scheme, which can be found in appendix B, with courtesy of Patricia Assen. Key differences with the current process are briefly summarised below, and the process taking into account different companies is visualised in figure 7.

1. The micro hub is connected to the PI, which provides it with access to a large variety of food products from different suppliers.

2. The micro does not only serve as a passive distribution hub, it also serves as the communication channel that takes orders and payments from nanostores, and ex-ercises consolidated demand towards suppliers. Both order communication and payments are envisioned to be exclusively digital.

3. The micro hub arranges last mile delivery for a cluster of nanostores within the favela using local carriers.

(22)

5.3 Digital technology diffusion in favelas

Brazil like other Latin American countries has known high levels of inequality in its history, stemming from its labour intensive agricultural economy in colonial times, that resulted in high levels of slave labour. Slavery was officially abandoned in the late 19th century but institutional constructs remained in place to maintain high inequality. Fer-nandes et al. (2019) presents a timeline of events how favelas gradually developed in the expanding city of Rio. The formation of favelas was a result of poorly planned city remodelling initiatives by the city authorities, that led low income groups to cluster in neglected parts of the city in the early 20th century. The term favela was first used by the press in 1920, at the time when favelas grew rapidly. Initiatives have since then been taken by authorities to gain control over favelas but these were often ineffective. Criminal gangs associated with drug trade have spread throughout the favelas of Rio in the 1980 and 1990’s. These gangs have remained as the dominant authority in many of the favelas ever since. In 2010, approximately 11.4 million people lived in favelas, corresponding to roughly 6% of the population of Brazil(da Costa Bezerra,2017).

Given this history of weak influence and neglect by formal authorities, favelas have hardly developed any infrastructure for basic utilities such as sewage, power and running wa-ter. Illegal tapping these of basic utilities from outside of the favelas is common practice (Nemer, 2018). Limited access to these basic utilities, most notably electricity, poses a challenge to building a digital infrastructure. Relevant digital technologies in this con-text regard technologies that can facilitate the transition between the current and the envisioned operational process. These technologies include internet connection, mobile telephony, and digital financial services. Even though the functionality of these tech-nologies are not fully separated (internet connection enhances smartphone functionality, and also enables online financial services), their diffusion levels in favelas will be briefly discussed separately as each technology also holds individual capability.

5.3.1 Internet and smartphones

Policies aimed to promote social and economic inclusion by providing access to infor-mation and communication technologies (ICT) have led to establishment of community technology centers (CTCs) in favelas. These CTCs, or LAN houses, or telecenters, have provided favela residents with the first access to internet, starting around 20 year ago (Nemer,2018). Despite internet gaining some initial traction in favelas because of CTCs, the success of CTCs in achieving its goal of social and economic inclusion through ICT has been limited by numerous factors. These factors include unstable power supply limit-ing and even damaglimit-ing CTC computers (Nemer,2018), but also a lack of initiative from residents to go online. The lack of initiative from residents in turn can be explained by a complete absence of initial digital literacy, low education levels, and unawareness of benefits that internet and other ICT may bring (Prado, 2016). The limited success of CTCs has in turn led to decreased interest for policies to support CTCs.

(23)

development literature often excludes Brazil because it is not classified as a developing country. Nevertheless, (Nemer, 2018) provides a snapshot of the status of smartphone adoption based on field work conducted in 2012 and 2013 in favelas located in the city of Vitoria, Brazil. Findings conclude that some young adults in favelas managed to ob-tain smartphones by the time when interviews were conducted, and that authorities had invested in public Wi-Fi hotspots, using funds previously destined for CTCs. However, smartphone adoption was still low, and the author mentions the practice of sharing a smartphone with groups of 3-4 friends. The only type of smartphone mentioned was the Xingling. The Xingling refers to different models of replicas for phones such as the Iphone. Xinglings were not bought in stores, but smuggled in from China by gangsters and sold on the streets.

Nemer (2018) makes no mention of smartphones connected to the internet via cellular (3g, 4g) networks. Similar fieldwork was conducted by de Souza e Silva et al. (2011) regarding mobile phone use in favelas, but prior to smartphones. The authors confirm that mobile telephone subscriptions are often too expensive for favela residents, and also using credit on a pre-paid basis of often omitted. Phones are said to be frequently used for practices such as ’beeping’, which entails ringing and hanging up to send out a signal to the recipient, without spending credit.

(24)

Population group/ Indica-tor

Accessed the in-ternet in lifetime

Accessed the in-ternet by phone in the last 3 months

Used 3G/4G on phone Used WI-FI on phone 2017 Illiterate 13% 14% 56% 80% Fundamental ed-ucation 62% 59% 60% 84% Lowest income class 60% 55% 61% 83% 2013 Illiterate 3% 2% 81% 21% Fundamental ed-ucation 40% 18% 69% 55% Lowest income class 32% 15% 71% 40%

Table 1: Internet and smartphone adoption for favela communities proxy groups

5.3.2 Digital financial services

Financial services have also been considered as means to improve inclusion and empow-erment for communities in favelas, which can be stimulated by policies. Regulators are motivated to improve availibility of financial services, as electronic payments helps reduce the black market, but also since access to loans provide opportunities for local businesses to expand PWC (2016). Helena and M¨uller (2018) describes the process of a few daring banks that have established offices in favelas, partnering with an NGO that knows these areas well. While these banks have succesfully build some clientele in favelas, the favelas under scope had been subject to a wider range of so called ”pacification” policies, in-cluding armed forces occupation and instalment of police stations, among other actions. The apparent inability of formal authorities to keep up with with such pacification ef-forts as described by Brooks (2016), makes it unlikely that traditional banking services with physical bank offices will continue to gain traction in favelas. Apart from a lack of bank offices and ATM machines in favelas, an additional structural problem with tradi-tional banking services is that costs associated with formal bank accounts often makes it unattractive for the lowest income groups (PWC,2016).

(25)

accounts. In the case of M-pesa, users can exchange cash for mobile credit at dedicated vendors. While the previous source suggests that mobile wallet adoption is low among favela residents, the Central Union of Favelas has recently introduced a similar concept to provide banking services especially for favela residents: The Cufa card. The Cufa card is linked to users prepaid mobile credit account, but is accompanied by a card that can be used for digital payments like any other debit card. The Cufa card subscription charges 8 Brazilian real per month but returns 10 real in credit, to stimulate adoption. The cufa card is free of charge and does not require a bank account (Sorosini,2017).

Another financial innovation enabled by internet and mobile devices, are so called mobile Points of Sale (mPOS). mPOS refer to small electronic devices that can be linked to a mobile device to turn it into a payment terminal for shop owners. Fransoo et al. (2017) mention SrPago as a provider of such mPOS devices active in Latin America. The authors mention that also shops using mPOS are not necessarily required to have a formal bank account, as they can have it linked to a digital account to receive funds, which they can use as a mobile wallet. A final innovation in digital payments that should be mentioned is NFT technology. While this technology has existed for some time as a wireless solution for card payments secured by a PIN, the latest smart phones of today are also equipped with NFT technology. NFT, mobile wallets and mPOS together can completely facilitate payments digitally at stores, without the need for bank accounts.

5.4 stakeholder analysis

A stakeholder analysis is performed to extract critical succes factors (CSFs) for the de-signs.

Favela residents

This group of stakeholders includes all of the Favela residents that rely on nanostores for their access to food. Residents are said to be very loyal to a specific nanostore due to their bond with the owners, and the proximity of the store. Restricted mobility and sometimes also safety concerns reinforce the habit of visiting the same nanostore that is in close proximity each time. On an individual level, the goal of residents is considered to be access to high quality affordable food. On a collective level, it is assumed that residents want to feel empowered. This includes having influence over developments in their local food supply chain, so that changes are to their benefit.

Nanostore owners

(26)

Intra favela transporters

The extra distribution tier added by the micro hub divides the transport between DC’s and nanostores into DC to micro hub (inter favela) and micro hub to nanostore (intra favela) transport. The intra favela transporters therefore only cover routes within the favela. They can be either employed by the micro hub or independent couriers paid for their services. It is assumed that transporters lose their order processing tasks as is the case in the current situation, because of improved digital ordering capabilities for the nanostores. Depending on the area that one micro hub covers, the transporter may have preferences or restrictions for which routes his services are available, due to preferred nanostores to visit, or neighborhoods to avoid due to safety concerns. The goal of in-tra favela in-transporters is considered to be earning an income by offering in-transportation services, and to have an influence over his assigned routes.

Micro hub owners/ operators

The micro hub adds value to the logistical process of supplying nanostores by allowing larger vehicles to carry products in (semi) bulk closer towards nanostores, after which products can be reconsolidated into orders destined for individual nanostores, which can then be transported along optimized routes in the last mile. Since the micro hub is a novel concept, its ownership structure and the exact goals of the owners are not definite. Ideally micro hubs should be owned and operated by local entrepreneurs. Assuming the micro hub operates as an independent business, it generates an income simply by buying products in bulk and selling it to nanostores against a higher price. The main goal of the micro hub is then to make a profit by doing so.

Inter favela transporters

In the envisioned situation, DC’s and micro hubs are connected to the PI. Based on PI principles, DC’s may request transportation, and logistic providers may offer their ser-vices on an open network. Matchmaking ensures that the best possible match is made based on characteristics such as delivery time, costs, and possibly additional requests such as cooling of produce during deliveries in case of perishable products. In this sense, inter favela transporters refers to an unknown group of potential transportation options, maybe even including autonomous (driverless) vehicles. However, to ensure transporta-tion optransporta-tions are available between DC’s and micro hubs, micro hubs should be positransporta-tioned appropriately so that they are accessible for vehicles in a safe manner.

Suppliers

(27)

CPG manufacturers

Consumer packaged goods refer to a variety of preservable goods including liquor, beer and tobacco, cookies and coffee, juices and drinks, dried meat and canned food, candy, and bread ingredients. Due to the unperishable nature of these goods, distribution and transportation of these goods is relatively simple and can benefit from economies of scale. This has contributed to the formation of large companies such as Coca Cola and Unilever. These large CPG manufacturers compete over market share. This competition has lead to direct service models as the dominant method for supplying nanostores in the current situation, despite the inefficiency from a logistical point of view. In the envisioned situation the micro hub should be positioned in between the CPG manufacturer and the nanostore, therefore the CPG manufacturer should be motivated to use this channel. The goal of CPG manufacturers is considered to be the option to effectively compete for nanostore shelf space.

Fresh food suppliers

Farmers outside the city are considered for this archetype of suppliers. It is assumed that these farmers currently only reach favelas through markets, in or outside of favelas. Markets are selling directly to residents and also nanostore owners buy their stock of fresh produce on markets, to sell to residents later. Problems with this method of distribution to inner cities is that markets are only possible at few locations, and are challenging to reach for farmers or wholesalers of fresh food. Secondly, the restricted mobility of favela residents makes that there is a lack of reliable channels for fresh food suppliers to reach favela residents. The goal of these suppliers is therefore considered to be a reliable channel to reach favela residents.

Formal authorities

Formal authorities consist of city regulators, it is expected that they are interested in supporting cost effective initiatives to improve conditions for favela residents, and to improve integration of favelas into the formal economy. Also, formal authorities may be interested in the micro hub as means to improve taxation processes. Fransoo et al.

(28)

6 Design considerations

6.1 Critical success factors

Stakeholder goals Operational CSF Strategic CSF Favela residents

Access to high quality affordable food

• Access to payment methods accepted by nanostores

Empowerment • Influence on nanonstore as-sortment and prices

Nanostore owners

Earn an income

• Online ordering from micro hub

• Offer payment methods to residents

• Access to payment methods accepted by micro hub • Ability to buy per consumer

unit

Provide service to community

• Access to high quality af-fordable food from micro hub

Intra favela transporters

Earn an income • Receive assigned delivery route

Influence on assigned routes

(29)

Stakeholder goals Operational CSF Strategic CSF Micro hub operators

Make a profit

• Online ordering from DC • Online nanostore order

pro-cessing

• Offer payment methods to nanostores

• Communicate delivery routes to intra Favela transporters

• Sufficiently large as-sortment for attracting nanostore orders

• Manageable assortment size for effective inventory man-agement

Inter favela (PI) transporters

Ability to access micro hub safely

• Digital verification of micro hub location and accessibil-ity

CPG manufacturers

Effectively compete for nanostore

shelfspace • Access to PI transportation

• Communication channels with micro hubs to offer deals

• Incentive for micro hub to push suppliers products to nanostores

Fresh food suppliers

Reach favela residents • Access to PI transportation

Formal authorities - regulators

Support favela integration in for-mal economy

• Establish micro hubs as for-mal enterprises

• Truce with informal author-ities

(30)

6.2 Balancing strategic CSFs

As explained in previous chapters, CPG manufacturers currently avoid shared logistical infrastructures because of competitive reasons, despite higher logistical efficiency of such shared infrastructures. To this end, this section explains a scenario how the micro hub might be a feasible alternative because strategic benefits of key stakeholders are taken into account.

1. Each nanostore is registered to a single micro hub that is able to fully supply the nanostore.

2. Micro hubs serve a cluster of nanostores in the area, performing deliveries to these nanostores using optimized routes.

3. Efficient inventory management is important for profitable operations of the micro hub. It is assumed that the bond between favela residents and their nanostore is stronger then the preference of residents for brands, therefore also the nanostore will not have a significant preference for a specific brand of comparable products. For this reason, this scenario assumes that the micro hub preferably sells only one or a few brands of comparable products. This allows micro hubs to buy products in larger quantities and improve efficiency of inventory management.

4. The preferred exclusivity of brands to have in stock for micro hubs provides a window for CPG manufacturers to compete for this exclusivity. Micro hubs will posses a stronger position of power to negotiate contracts with CPG manufacturers compared to individual nanostores, but CPG manufacturers also benefit from deal-ing with micro hubs that represent clusters of nanostores, instead of dealdeal-ing with individual nanostores. The competition among CPG manufacturers to get their products into micro hubs exclusively is expected to result in better prices for micro hubs, and nanostores and residents consequently.

5. While the stocked items provide best value for money to nanostores and residents, access to variety should not be limited to these stocked items. In addition to ordering items from stock, also ordering items from catalogues will be considered in designs. This option is further discussed below.

6.3 Digital solutions for operational CSFs

6.3.1 Online order processing

• It is assumed that mobile devices and internet are sufficiently present now or in the near future, for nanostore owners to make orders online. Therefore it is considered that nanostores will make orders on an online platform provided by the micro hub, in a similar fashion to e-commerce orders.

(31)

has access to from the central DC that supplies the micro hub, or any other food supplier able to get their products to the mirco hub using the PI.

• Unlike stocked items that can be ordered by the nanostore per consumer unit, catalogue items should be purchased per full casepack, to avoid partial casepacks remaining at the micro hub. However, a crowdsourcing functionality will be consid-ered where nanostores can subscribe for single consumer units of a catalogue item, and the micro hub will only order a full casepack after all units are accounted for by different nanostores. This functionality also provides feedback to the micro hub regarding new items to consider to keep as stock.

6.3.2 Digital payments

• It is expected that mobile wallets, and either a linked card like the Cufa card, or a smartphone equipped with NFT for wireless payment, can be expected for most favela residents in the near future.

• Nanostores are considered to be equipped with mobile Point of Sales(mPOS), turn-ing their mobile devices into payment terminals. Additionally, nanostores should have either a formal bank account, or a mobile wallet to receive the funds.

• Micro hubs should have a payment option in their online portal, that supports popular mobile wallets used by residents. The micro hub as a legal entity is expected to have a bank account the receive funds.

• Transactions between micro hubs and DCs are only considered to be using regular bank accounts.

6.3.3 Inventory management

• Upon placing an order for new goods from the DC, the DC should be able to provide an estimated time to arrival (ETA) for the new goods. Furthermore, PI containers should enable to keep updating ETA while goods are in transit.

• Arrival of goods at the micro hub can be automatically recognised when the RFID tag of the PI container comes in close proximity to a tag reader in the micro hub, and update the digital inventory.

• An inventory management system can monitor stock levels and provide reorder suggestions.

(32)

7 Functional architectures

7.1 Context function A1: Supply food to favelas using PI

Figure 8 and 9 show the hierarchy and the interaction diagram of the context function A1 respectively. This context function intends to capture all relevant functions con-taining inputs, outputs and controls required for the central function 0 (Supply favelas through final DC). Inputs and outputs refer to horizontal lines going into and coming from functions, and controls, or triggers, refer to vertical lines into functions.

The inner working of sub-functions A11 to A13 are considered outside the scope of this research. Function Dash.11 (Supply final DC using PI) provides the micro hub with a catalogue of items to order that can be ordered from the DC, and the items per casepack that serves as minimal order quantity per product. Dash.12 (Supplier competition) is triggered by the consolidated demand exercised by the micro hub, and results in Micro hub-Supplier negotiations. Dash.13(Manage empty PI containers) regulates the flow of empty PI containers needed for PI transport of goods to the micro hub, and empty containers coming from the micro hub.

The remainder of this chapter is devoted to explaining function 0 and its sub-functions. At the lowest level of each branch of the functional hierarchy, digital technologies and applications that support the functions are marked bold.

(33)

(34)

7.2 Function 0: Supply favelas through final DC

Function 0 is considered the central function of the designs proposed in this chapter. figure 10 shows how function 0 is broken down into four sub-functions, and figure 11 shows their interactions. Inner working of function 0 is better explained by individual subsystems in the upcomming sections but figure 11 can be consulted to check origins of inputs, outputs and controls.

(35)

(36)

7.3 Function 1: Consolidate orders

Function 1 consolidates orders from the nanostore level to a consolidated level per micro hub, for a cluster of nanostores. Function 12 consolidates orders for stocked items, which will be further detailed in the next subsection. Function 13 consolidates orders for non stocked, catalogue items. Function 13 is controlled (limited) to products available in the final DC, and by the minimum order size of products which should be at least bought per casepack. Function 13 features a feedback loop towards function 12 regarding popular catalogue goods that may be considered to keep as stock. Function 14 places combined orders for catalogue and stock goods, and provides feedback for estimated time to arrival (ETA) of the products which helps function 12.

Function 11 and function 14 will be supported by an online cloud environment for receiving nanostore orders and making orders to DC.

(37)

(38)

7.3.1 Function 12: Manage micro hub inventory

The functionality described in this function should be incorporated in the micro hubs inventory management software. Figure 15 shows how popular catalogue goods are controlled by outcomes from micro hub- supplier negotiations, before they are chosen to be adopted as stock. Function 122 determines the amount of stock based on historic de-mand. Not included in the designs but the inventory management software may quantify overage and underage cost parameters for stock and provide suggestions based on a cost minimising function.

Function 123 monitors the stock level of goods by removing stock that has been ordered, and adding incoming stock to its database. Real time monitoring can be applied to include the time it takes until arrival of new, and withdrawal of requested stock. The micro hub can be equiped with RFID readers to detect and automatically register incomming stock when the RFID readers detect a tag associated with an incomming (PI) delivery. A feature not incorporated in the design but that should belong in the inventory management software is the option to manually adjust the digital stock level for possible corrections due to breakage, perishing of goods, etc. Function 124 consequently makes a reorder suggestion when the stock level drops below the reorder threshold.

(39)

(40)

7.4 Function 2: Arrange distribution of goods

Payment requests initiated at all levels (function 24 for residents at nanostores, function 11 for nanostores to micro hubs, and function 14 for micro hubs to DCs), are verified under function 3, which serve as crucial controls to initiate distribution steps in function 2.

Function 21 arranges secure PI transport from DCs to Micro hubs and is further detailed below. Function 22 uses the availible goods (stock and catalogue) and prepares orders for individual nanostores. Function 23 determines efficient nanostore delivery routes based on orders that are assigned to similar timeframes of delivery, and the availability of transporters. Digital mapping services such as the free google maps platform, or RouteXL as used by Fietskoerier can be used for generating optimal routes. Finally, function 24 performs the actual routes to nanostores and residents consequently. Function 24 is also detailed below by another layer of sub-functions.

(41)

(42)

7.4.1 Function 21: Transport food to micro hub using PI

It is assumed that the DC arranges transport for an order after micro hub payment has been verified, in an online PI cloud environment. The PI transport request includes a request for PI containers, and for an appropriate level of security that is determined in function 4. function 212 starts a shipment after a PI transport request has been matched to PI containers and secure transport. Shipment details include destination and specifications per containers, such as cooling requirements.

Function 213 provides real time updates on a shipment status. This is enabled by RFID tags placed on PI containers that can communicate with transporters and nodes that the containers pass. IoT is mentioned as an umbrella term for options how RFID information is further transmitted via LPWA technologies as discussed in chapter 2, and processing technologies. This provides micro hubs with an ETA for incoming stock.

Function 214 uses RFID tags of PI containers and RFID readers in the micro hub to detect RFID tags of incomming containers. The RFID signals are able to exchange information on the containers content and therefore arriving containers can automatically update the micro hub inventory in its digital database. It is assumed that containers remain at the hub as storage units until emptied, especially relevant for cooling units. The arrival of new containers should ideally be ’just in time’, matched to the timing of slots opening up because of empty containers. This may bring additional synergies by combining stock arrival with return function of empty containers using the same transportation means.

(43)

(44)

7.4.2 Function 24: Transport food to nanostores and sell to residents

Function 24 shows that mobile point of sales (mPOS) are considered to be used to provide payment options to residents. Smartphones and tablets can be turned into payment terminals using additional hardware from for example SrPago. This requires residents to have either a mobile wallet on their smart phone, or a credit or debit card. Cash is not considered in these designs to be an accepted method of payment at nanos-tores, which may be a strong assumption to make. However, mobile wallets like M-pesa allow to charge the mobile wallets using cash at dedicated vendors. Moreover, Fransoo et al. (2017) state that informal credit(ability to pay later) is frequently given to known customers at nanostores. This may mitigate the problems of inability to pay because of empty smart phone batteries. Finally, nanostores may decide to tolerate cash payments from few customers, and charge their own mobile wallets using this cash at vendors. These options to avoid lost sales because of mPOS are not taken into considerations in the designs, and may require additional infrastructure of vendors, in or outside of favelas. Function 243 expresses how interest for exotic catalogue goods is created by conversations between nanostore owners and residents. Finally, function 244 regarding the nanostore order placement is expected to be done in the online cloud environment as mentioned.

(45)

(46)

7.5 Function 3: Facilitate payments

This function shows how payments should be facilitated on three levels. Also mentioned in the previous function, Function 31 is expected to be done using mobile wallets for residents and either mobile wallets or bank accounts for the nanostores. Function 32 in turn expects bank accounts for micro hubs, likewise for DCs in function 33.

(47)

(48)

7.6 Function 4: Provide security

Function 4 shows how security can be provided to PI transport between DC and the micro hub, on a basic level. Function 4 can also be considered a subset of the matchmaking function for arranging PI transport. Again, digital mapping services such as google maps are considered for checking routes, but a plugin such as crimeradar is considered to check the safety of the routes.

(49)

(50)

8 Design validation

Since contributions proposed by this research have emerged in a late stage, and writing occurred far from the focus area, validation capability was limited. Nevertheless, a valida-tion workshop was conducted at Fietskoerier headquarters in Zwolle, which has provided some valuable insights regarding explicit and implicit assumptions made during the de-sign process. The workshop was conducted in combination with a student conducting parallel research, in a two hour interview session with a business developer from Fietsko-erier. The workshop was conducted in Dutch, relevant statements have been translated to English. Audio recordings can be shared upon request.

Fietskoerier is a third party logistics provider for first and last mile deliveries, currently operating in 33 cities in the Netherlands. Fietskoerier differentiates its self through their network of national coverage for deliveries per bike, performing distribution services for fulfilment centers (e-commerce), postal services, aswell as deliveries to pharmacies. It was evident that Fietskoerier seeks to enhance performance using digital technologies, as they are currently working together with an expert from the University of Groningen to adopt a system of Dynamic Planning.

”Eventually we want to combine all those services... resulting in dynamic planning, and the computer determining the optimal route”, ”Some day we will have so many packages to move within the city, and we enter the number of bike couriers, the capacity that they can carry... and [bike couriers] will see on an app where they have to go”

The services here refer to routes that are currently separated (e-commerce, postal, phar-macies) for both the first mile (picking up an order) and the last mile delivery of orders, that are currently also separated. Integration of such routes can be considered as a step towards the PI. The remainder of this validation report now shows opinions on subfunc-tions 1, 2 and 3 of the central function zero.

Function 1: Consolidate orders. Regarding the consolidation of orders with a distinc-tion between stocked items and crowdsourced exotic items, The expert could not confirm its feasibility and desirability. When thinking out loud about how this would work in practice:

(51)

relation is characterised.

”The most efficient way will be to pay them per package delivered, and let them figure it out themselves how to do it... That is what DHL does... They use freelance drivers... Under this model, the courier bleeds but the consumer wins the most”

Fietskoerier pays their couriers by the hour, but also states:

”It is really not-done to just tour around like a recreationist, we still have enough couriers [to hire], so when someone cycles too slow then they really will get fired... I assume that also in those countries there is no lack [of people to hire]... That is our tension field, we demand them to cycle fast”

The second simplification is that in designs, the interplay between various micro hubs was ignored. Regarding spread of their own hubs and logistical efficiency, the expert had this to say:

”[In Amsterdam] we make a loss on those routes...In due time we really should open a second [hub]...Some zip codes we do not provide services, we first need more volume there” Function 3: Facilitate payments. It was confirmed by the expert that digital means should be used to take responsibility for payments out of the hands of couriers.

”As soon as the delivery is done, and is confirmed by the nanostore owner, and [the nanostore owner] can make the payment on his own phone directly to the micro hub, then the micro hub can inquire to the nanostore if a payment is not fulfilled. You should not burdon the courier with this”

After suggesting that payments (and orders) between nanostores and micro hubs could be done like in today’s e-commerce platforms, paying upfront through an online application, the expert confirmed that in his opinion this seemed most suitable.

”...In our case [of transport services], you pay afterwards, but for goods you don’t do this... I do think that for reliability of payments, I would do it that way”

The latter answer does reveal a simplification in the design of payment facilitation, namely that transport costs are incorporated in the price for goods for each layer of payments, but no guidance is offered for who ultimately pays the price.

Finally, the assumption in current designs that payments between micro hub and the DC will also be upfront was challenged, as the micro hub is assumed to a legal entity part of the formal economy.

(52)

9 Discussion

9.1 Contributions to literature

This research has shown that digital capabilities stemming from internet and smartphones are improving for favela communities, and has argued how these capabilities can be used to improve efficiency of food supply to these areas. Designs were based on the micro hub as proposed by Vries (2018), but the digital infrastructure proposed in this research attempted to show that the micro hub may also pose benefits for food suppliers, that currently bypass shared logistic infrastructures for competitive reasons. Judging from the adoption rates of digital capabilities, a micro hub may already be considered for favelas even before implementation of the physical internet. However, this research has also pointed out the additional benefits that a PI connection to the micro hub may bring. The PI enables the mirco hub to source products from more places, with greater efficiency. Combined with online portal where nanostore can place orders, this improves transparency and influence in local food offerings. By taking into account strategic goals of stakeholders, this research can also be considered as a partial business case for PI hubs, building on Oktaei et al. (2014), but more specific for problematic areas.

Secondly, even though this research has only scratched the surface of development lit-erature, this research holds elements of four directions of development literature; ICT improvement, Financial inclusion, Logistical infrastructure, and health promotion (by access to better food). This research and resulting designs may be considered as an inte-grative effort that shows that these four developments have synergies, potentially yielding better achievements when combined instead of pursuing single purpose policies.

9.2 Contributions for practitioners

(53)

9.3 Limitations and future research

This research leaves numerous directions for future research. First, many elements pro-posed in designs have not been properly validated. Future research may investigate whether favela communities may truly start to adopt digital capabilities such as mobile payments and online ordering, and whether gangs will accept these developments in their area, especially the abandonment of cash payments. Also whether the suggested distinc-tion between stocked and crowdsourced goods will work to improve transparency and value for money, is not validated.

Second, this research has not attempted to quantify the benefits that a micro hub may bring, from improved logistical efficiency or increased negotiation power towards food suppliers, or any other potential benefits that have been claimed.

Third, as mentioned, the issue of ownership of a micro hub or a network of micro hubs, is still an open question. A possible option to further investigate is whether an en-forced structure of shared ownership might be appropriate to protect interests. Potential shareholders can be local entrepreneurs, nanostore owners, food suppliers and local trans-porters. this suggestion however is speculative and requires further investigation.

(54)

10 Conclusions

(55)

References

Alice (2015). InformatIon SyStemS for Interconnected LogIStIcS. Technical report.

Brooks, B. (2016). Rio’s slum ’pacification’ effort stalls as killings tick up.

CETIC.br (2013). TIC Domic´ılios - 2013 Indiv´ıduos.

CETIC.br (2017). TIC Domic´ılios - 2017 Indiv´ıduos.

Columbus, L. (2018). 10 Ways Machine Learning Is Revolutionizing Supply Chain Man-agement.

Crainic, T. G. and Montreuil, B. (2016). Physical Internet Enabled Hyperconnected City Logistics. In Transportation Research Procedia.

da Costa Bezerra, K. (2017). Postcards from Rio: Favelas and the Contested Geographies of Citizenship. Fordham University.

de Souza e Silva, A., Sutko, D. M., Salis, F. A., and de Souza e Silva, C. (2011). Mobile phone appropriation in the favelas of rio de Janeiro, Brazil. New Media and Society.

Fan, J., Han, F., and Liu, H. (2014). Challenges of Big Data analysis.

Faugere, L. and Montreuil, B. (2016). HYPERCONNECTED CITY LOGISTICS: SMART LOCKERS TERMINALS & LAST MILE DELIVERY NETWORKS. Technical report.

Fernandes, J., Mason, K., and Chakrabarti, R. (2019). Managing to make market agence-ments: The temporally bound elements of stigma in favelas. Journal of Business Re-search, 95(September 2018):128–142.

Fosso Wamba, S., Gunasekaran, A., Bhattacharya, M., and Dubey, R. (2016). Determi-nants of RFID adoption intention by SMEs: an empirical investigation. Production Planning and Control.

Fosso Wamba, S., Gunasekaran, A., Papadopoulos, T., and Ngai, E. (2018). Big Data Analytics and IoT in logistics: a case study. The International Journal of Logistics Management, 29(2):739–766.

Fransoo, J., Blanco, E., and Argueta, C. (2017). Reaching 50 million nanostores: retail distribution in emerging megacities. Cambridge: CreateSpace Independent Publishing Platform.

(56)

Hofman, W., Spek, J., and Brewster, C. (2017). Applying blockchain technology for hyperconnected logistics. pages 59–70.

HP (2018). What is Edge Computing?

IBM (2017). Maersk and IBM Unveil First Industry-Wide Cross-Border Supply Chain Solution on Blockchain.

IBM (2018a). B2B Collaboration Point of View.

IBM (2018b). Watson Supply Chain Point of View.

Inmarsat (2018). Inmarsat Maritime.

Kim, S. J., Deng, G., Gupta, S., and Murphy-Hoye, M. (2008). Intelligent Networked Containers for Enhancing Global Supply Chain Security. The Third International Conference on COMmunication Systems softWAre and middlewaRE.

Li, H., Zhou, Y., Tian, L., and Wan, C. (2010). Design of A Hybrid RFID / GPS-Based Terminal System in Vehicular Communications. (2008):0–3.

Markgraf, B. (2018). How Far Can a Cell Tower Be for a Cellphone to Pick Up the Signal?

Montreuil, B. (2011). Toward a Physical Internet: meeting the global logistics sustain-ability grand challenge. Logistics Research.

Montreuil, B., Ballot, E., and Fontane, F. (2012a). An open logistics interconnection model for the physical internet. In IFAC Proceedings Volumes (IFAC-PapersOnline).

Montreuil, B., Ballot, E., and Tremblay, W. (2014). Modular Design of Physical Internet Transport, Handling and Packaging Containers. Technical report.

Montreuil, B., Meller, R. D., and Ballot, E. (2012b). Physical Internet Foundations. IFAC Proceedings Volumes, 45:26–30.

Montreuil, B., Meller, R. D., Ballot, E., and Paristech, M. (2010). TOWARDS A PHYS-ICAL INTERNET: THE IMPACT ON LOGISTICS FACILITIES AND MATERIAL HANDLING SYSTEMS DESIGN AND INNOVATION. Technical report.

Nemer, D. (2018). Wired mobile phones: the case of community technology centers in favelas of Brazil*. Information Technology for Development, 24(3):461–481.

Nemzer, E. (2016). The Deep Inequality of Brazil’s Tax Code.

Oktaei, P., Lehoux, N., and Montreuil, B. (2014). Designing Business Models for Physical Internet Transit Centers. Proceedings of 1st International Physical Internet Conference (IPIC 2014), pages 1–17.

A digital infrastructure design for physical internet connection in favelas