Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

1 Introduction

Recent national broadband initiatives have led to the construction of country-wide, fibre-based broadband networks. Countries such as Australia, NZ and Singapore have adopted a high-speed, FTTH network model where public funds are invested (with or without participation of private partners) and an open access operation is adopted. This paper models a high-speed, open access broadband network as a two-sided platform which comprises of two markets: an access market and a content market. The scope of the access market involves consumers achieving fibre connectivity from their network provider known as Local Fibre Company (LFC). The content market involves consumers subscribing to a Retail Service Provider (RSP) for actual retail services and products such as Internet access, voice service, and video (broadcast TV or on-demand). In this paper we focus on the access market and exclude issues of content market because the platform is still under construction. Focusing on consumer uptake helps understand salient aspects of FTTH growth when both, attractiveness of the platform to end-users and incentives for RSP participation are considered through the network effect approach. New high-speed broadband network build-up presents us with a timely unique opportunity to analyse the issues and drivers for uptake and growth as the markets take shape and evolve. Our approach builds upon [1] where both access market and content market are modelled as a developing two-sided market platform. We use an Agent-based Model (ABM) to simulate a range of scenarios that illustrate how the broadband uptake rate is affected by varying factors.

The rest of the paper develops as follows. In Sect. 2 we present the market structure of the FTTH platform and relate it with the theory of two-sided platforms, followed by highlighting key features of agent-based modelling. In Sect. 3 we present the access market model and provide details on configuration of simulation scenarios. We describe how the agents will interact in the model including consumer decision making process. Section 4 presents the key findings followed by conclusions in Sect. 5.

2 A High-Speed Broadband Market as a Two-Sided Platform

The new high-speed broadband structure will see a vertical separation of ownership and operation with LFCs being owners and RSPs being providers of services. LFCs will own and operate the lower layers of the network which will be geographically exclusive providing them a monopoly on the franchised area. This structure (Fig. 1) has introduced a vertical separation (lower layers structurally separated from upper layers), giving birth to an open-access platform. The main task of each LFC is to install fibre connections to their consumers in their regions and maintain fibre lines. Higher benefit will be achieved if LFCs manage to quickly implement the network, as it would satisfy the prerequisite for RSPs to start retail services. The retail services can only be sold by the RSPs to consumers. RSPs will purchase layer 2 services at regulated wholesale prices from the LFCs. The RSPs are expected to develop business models that include high-speed broadband, phone line, IPTV and video services over the platform.

Fig. 1
figure 1

High-speed broadband market structure

Full size image

2.1 The Broadband Platform as a Two-Sided Market

Founding work on the economics of two-sided platforms is found in Rochet and Tirole [2, 3] who introduced the term two-sided market. In a traditional value chain diagram, value moves from left to right: to the left of the company is cost; to the right is revenue whereas in two-sided networks, cost and revenue are both to the left and the right, because the platform has a distinct group of users on each side [4].

An open-access, high-speed broadband network represents a scenario of a two-sided platform [1]. A network provider implements a fibre-based network which can be regarded as a platform. On one hand the platform sells wholesale services to RSPs, whereas on the other it offers fibre connections to consumers. It is common in two-sided markets for one of the user groups to be subsidised. For example the Yellow Pages is usually free of cost to consumers, but advertisers pay to get a featured advertisement. Usually job seekers access job portals for free and employers need to pay to advertise. If these platforms reversed their approach, their network probably would not exist. In the case of FTTH network consumer access is being partially subsidised to encourage consumer participation.

2.2 Agent-Based Approach

The high-speed broadband market comprises multiple stakeholder groups, each with their own self-serving objectives. The stakeholder groups of LFCs, RSPs and consumers can be represented as heterogeneous agents in an ABM, each behaving according to their own preferences. Our ABM demonstrates how some market activities can be generated by the endogenously evolving interactions among such boundedly-rational stakeholders over time [5].

The use of ABM is not peculiar to the analysis of telecommunication markets. It is an effective methodology to tackle dynamics of telecommunication markets that involve: changing technologies, products, and regulatory reforms. Among the work conducted on telecommunication markets using ABM following is a sample of relevant references.

Beltrán and Roggendorf [6] used ABM to create an auction-based pricing scheme to facilitate network resource distribution negotiations for the analysis of bidding behaviours and in [7] they enhanced the simulation model by introducing richer strategies. Baryshnikov, Borger, Lee, and Saleh [8] created consumer and service provider agents assigned with utility based preference scores. The model included two types of RSP agents, RSPs providing bundled services vs. undiversified services. The simulation model showed RSPs providing bundled services outperformed the other type of RSP. Douglas, Lee, and Lee [9] presented a model in which the iPhone was introduced into the market, the model showed satisfactory reproduction of historical data but failed to predict exact market share in the future. Zheng, Jin, and Zhang [10] explored the effects of regulation with a duopoly mobile market and found that the duopoly market operated more efficiently with regulatory interventions. Diedrich and Beltrán [11] leveraged ABM to compare traffic discrimination policies. The model varied policy and competition scenarios. The results found that the content providers performed best when network neutrality is imposed; while network providers and consumers may benefit from traffic discrimination under certain circumstances. In the following text we present the ABM for the high-speed broadband platform access market.

3 Participation of RSPs and Consumers in the Access Market

Beltrán [1] used two-sided market theory to explain the presence of cross-network effects on an open access broadband platform. Cross-network effects embody the interdependencies between the two sides. One side, the RSP side, purchases wholesale services from the platform operator. The other side, the end-user side, is split in two groups: residential (R) customers and business (B) customers.

3.1 Access Market Model

\(C_{\mathit{RSP}}\) represents the wholesale cost to a RSP to get services from the platform. The number of residential and business consumers with active subscriptions from the RSP are given by: \(n_{\mathit{AR}}\) and \(n_{\mathit{AB}}\). \(P_{R}\) and \(P_{B}\) are the wholesale rates payable to the platform per individual residential or business consumer Therefore we can define the \(C_{RSP}\) as below:

$$\displaystyle{ C_{\mathit{RSP}} = n_{\mathit{AR}}P_{R} + n_{\mathit{BR}}P_{B} }$$
(1)

Regulated wholesale prices are charged based on individual connections (each of \(n_{\mathit{AR}}\) and \(n_{\mathit{AB}}\)). This means the LFC owned platform obtains revenue \(C_{\mathit{RSP}}\) from the RSP side based on the number of consumers it will manage to subscribe.

An RSP’s utility function would consider the positive effect of the presence of residential (\(n_{\mathit{FR}}\)) and business consumers (\(n_{\mathit{FB}}\)) with potential fibre connectivity (passing fibre line on the street) less the cost of purchasing wholesale services from the platform. In other words, the market becomes increasingly attractive to the RSP as passing fibre becomes available to homes and businesses. On a first approach business fibre consumers (\(n_{\mathit{FB}}\)) are excluded mainly because their associated connection costs are not clearly defined at this early stage of implementation (at least in the international cases inspiring the present model). The expenses that need to be paid to the platform by a RSP are represented by \(P_{\mathit{RSP}}\), which is equal to \(C_{\mathit{RSP}}\). Thus, if \(n_{\mathit{FR}}\) is the number of residential passing fibre consumers – who may potentially activate a fibre connection with a RSP, the utility of a RSP, \(U_{\mathit{RSP}}\) can be written as, Eq. (2). \(U_{\mathit{RSP}}\) increases as \(n_{\mathit{FR}}\) increases over time (t) and \(\alpha _{\mathit{RSP}}\) measures the effect of each consumer’s platform presence perceived by the RSP.

$$\displaystyle{ U_{\mathit{RSP}}(t) = n_{\mathit{FR}}(t) -\alpha _{\mathit{RSP}}n_{\mathit{FR}}\ }$$
(2)

The other side of the access market involves consumers. A residential consumer’s utility function is expressed as:

$$\displaystyle{ U_{R}(t) = n_{\mathit{RSP}}(t)\alpha _{\mathit{R}} }$$
(3)

where \(\alpha _{R}\) measures the effect of each of the \(n_{\mathit{RSP}}\) RSPs presence connected to the platform, which is perceived by a representative consumer. An additional assumption is that residential users may or may not pay for their fibre connection to home; this will depend on the subsidy terms defined by LFCs and the government.

The cross network issues described above were setup as an ABM [12] for the NZ case in order to highlight the impact of these cross network effects between RSPs and consumer groups. In this article we introduce further improvements to the ABM of [12].

We conducted an empirical study in NZ by engaging with leading platform operator (Chorus), RSP (Snap Internet) and consumers using a dominant imperialist multi-methodological design – one method or methodology as the main approach with contribution(s) from other(s) [13]. In our case the dominant method is qualitative data from 15 consumer interviews. The contributions include: firstly research study data from Chorus that included a large sample of 132 qualitative interviews followed by an online quantitative survey that totalled 1,009 respondents. Secondly broadband customers sample data from Snap’s Customer relationship management (CRM) system – which included information including the type of broadband products consumers are using, costs, location, and why the consumer decided to use Snap. The aim of the interviews was to acquire detailed information about the consumer perceptions of participating in the access and content market of a high-speed FTTH platform. The results would complement the secondary data from Chorus and Snap.

In this article we are leveraging the results obtained from this empirical study, therefore only describing a brief summary of relevant details for the ABM. We found the main driver for consumers to participate in the FTTH access market was perceived platform’s reliability and consistently fast connection. The barrier for participation was consumer contentment with the inferior ADSL or alike alternatives, however in the Australian NBN initiative the plan is to eventually make the old technologies obsolete. The deciding factors which promotes or withdraws consumer participation in the access market were firstly, start up costs which include cost of connection. Secondly, awareness regarding the benefits of fibre technology and its products. We incorporate these deciding factors in our ABM.

The price of the connection to the street is often met by the government or the platform operators. However the connection into the house may or may not be subsidised. In NZ the connection from curb to home is free until 2015, however not a very small number of potential consumers have taken benefit of such an offer. Chorus revealed only a 1.7 % uptake of 80,299 end users able to connect [14]. This report echoes our findings related to awareness being an important driver for consumer participation.

3.2 Setting Up Simulation Scenarios

The model simulates consumer uptake of a newly developed open-access, high speed broadband platform. End-users find it attractive to connect to the network when they find that a large number of RSPs operate on the platform, a fact that enhances consumers’ expectations of service.

RSP and consumer agents will be assigned a unique utility score by the simulation model and this will influence how they participate in this two-sided platform, based on Eqs. (2) and (3). The values for the \(\alpha _{\mathit{RSP}}\) and \(\alpha _{R}\) are randomly generated from probability distributions. RSPs will slowly appear into the market and will eventually reach saturation. New consumers, on the other-hand, keep subscribing to platform services as the connection becomes available. For the RSPs the larger number of consumers adopting fibre, the more encouraged they will be to enter the market. RSP agents maintain a when-to-market attribute, this specifies when a RSP should enter the market. This is when the \(U_{\mathit{RSP}}\) is positive. The consumer on the other hand maintains a when-to-subscribe attribute. These attributes vary based on the scenario settings and the probability distribution of the network effect parameter. We set up the following scenario groups:

  1. 1.

    Strength of cross network effects are high for consumers and RSPs.

  2. 2.

    Strength of cross network effects are high for consumers and low for RSPs.

  3. 3.

    Strength of cross network effects are low for consumers and high for RSPs.

  4. 4.

    Strength of cross network effects are low for consumers and RSPs.

3.3 Consumer Awareness

The consumer can become aware by multiple means. For this model we setup the following three drivers. Firstly, the platform operating LFC may run marketing campaigns causing an increase in awareness in a given percentage subset of the population. For this model we configured these awareness campaigns to occur every 6 months reaching 5 % of consumers. Secondly the consumer becomes more aware when the number of RSPs in the market increases beyond the consumer’s own perceived utility. Lastly we found from the interviews and industry data that family, friends, and word of mouth advertising is a trusted way for consumers to become aware. Therefore we integrated a friend circle creator model [15] into this ABM. This allows the consumer agents to become informed via their friend circle; in the case of this ABM the awareness goes up for the consumer when a majority of his friends subscribe to RSPs. Awareness score is maintained for each consumer and a score of 5 is considered to be high awareness, and anything lower is low awareness.

The scenario groups above are further explored by varying connection subsidisation within the market and consumer awareness settings. As a result each of the scenario groups 1–4 specified are further explored by the scenarios A–D below, producing a total of 16 scenarios.

  1. A.

    Partial subsidy and low awareness

  2. B.

    Partial subsidy and high awareness

  3. C.

    Full subsidy and low awareness

  4. D.

    Full subsidy and high awareness

The agents are: a single broadband platform operator, a number of RSPs and a large number of residential consumers. The following text will describe each agents behaviour in the model.

The platform operator is tasked to implement fibre, which involves adjusting the consumer’s status from no fibre to passing fibre. The speed of implementation is scaled relatively to the actual implementation – which is around 5–7 years in the international FTTH cases. In the reported results (Fig. 4) we configured the simulation for 250 ticks for the platform operator achieving 2,500 homes with passing fibre status. In this term all of the consumers will have fibre installed at their premises (i.e. n FR  = 2, 500), making the implementation speed approximately ten houses per day. The RSPs behaviour is simple – which is to become active in the market when \(U_{\mathit{RSP}}\) becomes positive, as explained above.

3.4 Consumer Agent Decision Making Process

Consumer agents maintain a connection status attribute. The connection status is either – no fibre, passing fibre or subscribed to RSP. The model creates all consumer agents with a no fibre status. The fibre eventually reaches consumer’s curb based on the speed at which the platform agent is laying fibre, this changes the consumer status to passing fibre. This is when the consumer becomes eligible to subscribe to a RSP in order to benefit from the fibre based services. The decision making process for a consumer transitioning eventually to activating high-speed services is shown in Fig. 2.

Fig. 2
figure 2

Consumer decision making process for considering subscribing to a fibre retailer

Full size image

When the consumer’s street receives passing fibre, the LFC usually informs the consumer regarding their construction schedule and identify which RSPs are operating on their platform. This is when the process presented in Fig. 2 starts, whereby the consumer will check if \(U_{R}(t)\) is positive. If \(U_{R}(t)\) is positive, the consumer agent will activate their fibre connection with a RSP. Incases when the \(U_{R}(t)\) isn’t positive then the consumer agent defers its reconsideration till its awareness score increases over a configured threshold. This action is triggered endogenously at the time when each consumer’s awareness becomes high, in this ABM its set to a score of 5. Consumers are also rationally bounded as they may or may not become aware based either via their friends or through LFC’s marketing impact on the environment of the model.

While reconsidering, the consumer agent obtains a score from the matrix shown in Fig. 3, this value is based on a combination of present awareness and dissatisfaction with existing connection. The consumer agent upon creation is profiled to have a certain type of satisfaction score with the alternative broadband technology to fibre. The consumer with high awareness of fibre services and with an unsatisfactory alternative connection, will be most likely to transition to fibre. This score is further scaled – either higher or lower depending on the present subsidy conditions in the market. As a result, the consumer either activates fibre or decides to remain on the inferior alternative, such as ADSL.

Fig. 3
figure 3

Matrix for determining a score for consumer reconsidering fibre

Full size image
Fig. 4
figure 4

Simulation results of varying scenarios

Full size image

The access market model described above is a simplification of the many complex issues (political as well as economic) surrounding the build-up of a government-funded FTTH network. Fibre-access uptake in this kind of subsidised environment presents itself with issues not found in full private network expansion.

4 Simulation Results

Cumulative results are collected to appreciate how uptake rate is affected by the combination of cross-network effects, consumer awareness and connection pricing as shown in Fig. 4. Each plot displays the number of consumers connecting to the platform and subscribing to a RSP as a function of time. The inclining straight dashed line shows the number of households the fibre is passing. The solid black line with varying values shows the number of consumers who subscribed to a RSP. Underneath each plot the percentage of consumers that subscribed to a RSP is given along with number of active RSPs in the market at the end of each simulation run. The plots shown in Fig. 4 are averages of running the simulation a number of times.

The platform best outcome from the cross network effects perspective is found in scenario group 1 (high utilities) and the worst outcome in scenario group 4 (low utilities). Scenario 1D is the best platform outcome as expected because the consumer awareness is high, connection costs are subsidised, and the strength of cross network effects is high for RSPs and consumer sides of the platform. The opposite applies for scenario 4A. The conditions in scenario 1D manage to subscribe 46 % of consumers to a RSP and 11 RSPs become active in the market. The worst conditions of 4A could only subscribe 9 % of consumers to a RSP and 4 RSPs became active in the market. The results of the best and worse outcomes are as expected.

Scenario group 2 is closest to the best outcome scenario group 1 and scenario 3 is closer to the worst outcome scenario group 4, emphasising effects of high \(U_{R}\) having a greater effect on the platform than \(U_{\mathit{RSP}}\). For example if we compare the percentages of consumers that subscribed to a RSP between scenario 2D (41 %) and scenario 3D (30 %), we can see that the subscribers 2D score significantly higher despite the active number of RSPs were four. This shows that the consumers took advantage of free connections because they were aware of the benefits of the high-speed broadband platform.

In scenario groups C and D – the connections were fully subsidised; whereas the scenario groups A and B provided partial subsidy, which expired in mid-implementation term. The curvature in the plots show how removing the subsidy limits the consumer uptake in the platform. It is common in two-sided markets for one of the user groups to be subsidised. The simulation shows the negative effects of not providing the subsidy.

If we compare the end of run values for percentage of subscribed consumers in scenario groups 3 and 4, we can see that there is a little difference. For example scenario 3B (15.32 %) versus scenario 4B (14.92 %). The advantage of scenario group 3 is such that the rate of consumer subscriptions were higher than scenario group 4 from the start, which is much more beneficial as the platform would be generating revenue from an early phase of implementation. This is because more RSPs became active in the early stages of implementation.

The friend recommendations helped uplifting the uptake percentages especially when cross network effects were low and subsidisation was partial. Friend recommendations increased the uptake ranging between 2–4 %.

5 Conclusion

The platform embodies an architecture that separates network services provision from end-user (retail) service provision. It operates under a set of rules which are in place by design (and regulatory intervention) whereby the LFC is prohibited from selling services directly to consumers. RSPs deal with the platform to acquire wholesale services which are used as inputs to their end-user services. We postulate the existence of cross network effects on the platform whereby end-users represented by consumer agents and RSPs represented by provider agents find themselves mutually attracted with an agent type’s attraction level increasing, the larger the number of agents of the other agent type in the platform.

A high consumer utility assures the access market’s performance, regardless of RSPs having a high or a low utility. Additionally the results show that having large number of first mover RSPs allows the platform to keep generating revenue, even though the consumer utility may remain low. The deciding factors for consumers in transitioning to the high-speed platform include awareness, subsidised connection to home, dissatisfaction with their existing broadband connection, and friends’ influence. The simulation scenarios display a wide variation of these factors to demonstrate how they affect consumer uptake. Critical success factors for a successful FTTH market establishment includes evaluating subsidy considerations along with upliftment of consumer awareness. This will prevent delays in consumer uptake which will benefit the overall market by attracting revenue from early stages.

Our work contributes to the increasing literature on agent-based modelling of market performance whose main features include: modelling the FTTH access market structure, understanding the access market as a two-sided platform, and the computational ABM for a scenario-based analysis.