- Email: [email protected]
Video quality vs. mobile data billing plans
Telematics and Informatics 33 (2016) 227–231
Contents lists available at ScienceDirect
Telematics and Informatics journal homepage: www.elsevier.com/locate/tele
Video quality vs. mobile data billing plans Andreea Molnar Portsmouth University, Winston Churchill Avenue, Portsmouth PO1 2UP, United Kingdom
a r t i c l e
i n f o
Article history: Received 25 June 2015 Received in revised form 24 July 2015 Accepted 24 July 2015 Available online 26 July 2015 Keywords: Consumer behaviour User choices Video quality Mobile data Bundle based billing Capped billing plans
a b s t r a c t Despite extensive research regarding the user willingness to pay under different billing plans and prices it is still unclear how users choose a preferred video quality under existing mobile data plans. This is especially important as the usage of adaptive video delivery servers that make use of different video quality levels is increasing. This study presents the results of assessing user preferences for video quality under three different mobile data billing plans. The results show that the user choice of different quality of video content depends on the billing plans used. It also shows that in the case when the mobile data plan is capped but the user has the possibility to pay extra for the exceeding quantity, s/he will choose higher video quality compared to the same plan in which the bandwidth is throttled when the bandwidth is exceeded. No correlation was found between the user choices and age or gender. Ó 2015 Elsevier Ltd. All rights reserved.
1. Introduction Mobile Internet traffic is characterised by an increase in number of subscribers (Pande et al., 2013) and an exponential increase in video content (Trestian et al., 2012) through an increase of both user generated content (Trestian et al., 2011) and access to other services, such as video through mobile networks and devices (Lee et al., 2015). Video content is usually heavier than other types of content and consumes more bandwidth. In an effort to impede congestion and increase revenues, mobile network operators that initially offered unlimited data access, realised that flat rate plans are no longer feasible and decided to cap them. Reviews of mobile prices in major European countries (Molnar and Muntean, 2013), and not only (Sen et al., 2013), have shown that most mobile data plans are cap based (aka bundle based, buckle based). Although capped billing plans are mostly present in mobile networks, broadband network operators also make use of them (Higginbotham, 2013). Capping the amount of data the user can consume is likely to continue even with the deployment of more powerful networks (Pande et al., 2013; Chih-Lin et al., 2014). Investing in new network technology is seen by mobile network operators as a way to impede congestion, and increase the revenues for network operators by removing unprofitable user behaviour (i.e. consumers that use a large amount of data). In this context, capped billing plans can be seen as a means of controlling network traffic and ensure a ‘‘fair-usage’’ policy for Internet resources. As another measure to avoid congestion, content providers adapt the video content to the network conditions, a process that leads to lower size data that will be delivered over the network at the expense of the video quality. For example, YouTube automatically selects a lower quality video when the request is performed from a mobile network rather than over a Wi-Fi or a wired network and Netflix allows the users to select a lower video quality in order to avoid reaching the cap limit (Newman, 2011).
E-mail address: [email protected] http://dx.doi.org/10.1016/j.tele.2015.07.009 0736-5853/Ó 2015 Elsevier Ltd. All rights reserved.
228
A. Molnar / Telematics and Informatics 33 (2016) 227–231
Despite the increasing use of capped billing plans on mobile networks and the increased usage of video over these types of networks, our understanding of how people would choose between different video qualities under different billing plans remains poor. The goal of this research is to partially fill this void by providing further understanding on the user choices for video quality under different prices. Without knowing their users, mobile network operators cannot optimise their plans for marketing purposes (Jung and Kwon, 2015). This would also help improve video adaptation algorithms leading to a better user quality of experience, which could help retain customers. Moreover, it could maximise the resources mobile operators have. With this aim, a quasi-experimental study (Reichardt, 2009) was designed to assess how users select different video qualities under billing plan constraints. Three different billing plans are considered for the study: a flat rate billing plan and two types of capped billing plans: in one billing plan, once the user reaches the data cap, the user has to pay extra for the exceeding quantity whereas in the other billing plan, once the user consumed the data from the bundle, the user bandwidth is throttled. The rest of this paper is organised as following. Section two introduces existing research on mobile Internet consumer behaviour. Section three presents the study set-up and the findings. Section four discusses the results and section five presents the study’s conclusions.
2. Related work Several studies (Koenigstorfer and Groeppel-Klein, 2012; Kumar et al., 2015; Jung and Kwon, 2015; Molnar & Muntean, 2015) have assessed consumers’ usage of mobile data. The study presented in Koenigstorfer and Groeppel-Klein (2012) assessed how consumer personality affects the consumer acceptance of the mobile Internet, showing that innovativeness, low desire for social contact and technology optimism influence consumer acceptance. Molnar and Muntean (2015) looked at how the user risk attitude affects user’s preference of video quality under monetary cost constraints, showing that risk adverse people are more willing to trade off quality for price than the risk seekers. Kumar et al. (2005) and Jung and Kwon (2015) assess factors that affect customer satisfaction and preference with mobile network operators. Jung and Kwon (2015) have shown through a study performed in Korea that by improving the customer satisfaction, the mobile network operators could improve the loyalty of their 3G and LTE users. They have also shown that 3G subscribers are more sensitive to call services and that LTE users are more sensitive to quality of data service and pricing. Kumar et al. (2015) show that mobile network parameters and tariff schemes are affecting customer preference towards a mobile subscriber. Other studies have looked into how consumer behaviour is characterised under different billing plans. For example, it has been shown that people are optimising their time spent on the Internet if the billing plan is not unlimited (Roto et al., 2006) and people own past experiences or highly publicised stories that report high billing plans are likely to have an effect on how people use and interact with mobile data (Isomursu et al., 2007). Marcus and Godlovitch (2013) showed that consumers use more intensively mobile data when they are connected to a Wi-Fi network and speculate that the reason may be rationing the consumption on mobile phones. Previous research has also shown that users are also uncertain about how much they consume and dislike bill shocks preferring flat rate billing plans due to their inherent predictability (Odlyzko, 2001). Molnar and Muntean (2010) have shown that some people may prefer a lower quality video under capped billing plans than what they would choose under a flat rate billing plan, however the study did not further assess how the user choices differ among existing billing plans. This study aims to address this gap by assessing how people choose between different quality levels on currently used mobile data billing plans.
3. Study 3.1. Sample and data collection A quasi-experimental design was followed in which the participants had to select among videos of different qualities given the constraint of different billing plans. The study was advertised and the participants volunteered to participate in the study. A total of 60 people took part in the study, 75% males and 25% females. The ages varied from 20 to 57 years old, with an average age of 32 years old. 3.2. Procedure The billing plans used in the experimental study are selected based on the results of the survey containing the most common billing plans in Europe presented in (Molnar and Muntean, 2013). Three scenarios were designed based on existing mobile data billing plans. Two capped billing plans were selected: one in which the user has to pay extra for exceeding the threshold included in the bundle and another one in which after the bundle data is exceeded the bandwidth is throttled. Although a flat rate data plan is not common on the mobile Internet, one of the scenarios included a flat rate billing plan for comparison. The following billing plans were selected:
229
A. Molnar / Telematics and Informatics 33 (2016) 227–231
Billing plan 1: Flat rate billing plan that allows unlimited data consumption. Billing plan 2: Capped daily billing plan in which for €0.99 per day the user gets 50 MB data included. If the bundle is exceeded they have to pay €1/MB. Billing plan 3: Capped billing plan in which the user has 50 MB data per day. After the quantity is exceeded, their bandwidth would be throttled to 64 kbps. The users were asked to assume that they have one of the above billing plans and select among different quality levels for a video clip. The study took place in a controlled environment following a similar design as recommended by the International Telecommunication Union (ITU-T P.910, 2008). The participants were given the same smartphone, a Google Nexus, with the following characteristics: resolution of 480 800, 512 MB internal memory and 1 GHz CPU. The video clips were encoded at five different quality levels similar with the recommendations contained in (Molnar & Muntean, 2015) when the adaptation for this type of device is to be performed. Table 1 presents the encoding of the video clips and the Mean Opinion Score (MOS). MOS was obtained by computing Structural Similarity (SSIM) index (Dosselmann and Yang, 2011) and mapping it to the MOS as described in Zinner et al. (2010). The participants were allowed to watch the video clips as many times as they wanted to but they were asked to watch all versions of the same video at least once in order to make an informed decision. The participants were able to see how long the video was (each clip’s playtime was slightly less than 2 min). Four scenarios were designed to investigate how the subjects’ preferences towards the quality of the video content changes when monetary cost is involved. These scenarios were designed based on the three billing plans above: one scenario for billing plan 1 and billing plan 3 and two for billing plan 2 (one in which the user has data in the acquired bundle and another one in which s/he is exceeding the bundle). Each participant was involved in all of the four scenarios and during all the scenarios they were encouraged to watch the video versions as many times as they needed to. Scenario 1: Participants were asked to assume that they have a billing plan that allows them to use unlimited data and were asked to select one of the video clip versions. Scenario 2: Participants are asked to assume that they have a daily billing plan in which for €0.99 per day they get 50 MB data included. They were told that they have data in the bundle to see any of the versions of the videos regardless of the size. Scenario 3: Participants are asked to assume that they have exceed the data in the bundle presented in Scenario 2 and as a result they will have to pay €1/MB if they consume data. The prices for each video version were presented to them (see Table 2 for the prices of each video clip). Also in this case they were asked to select one of the five versions of the video clip. Scenario 4: Participants are asked to assume that they have a capped billing plan but in the case when the cap is exceeded the bandwidth will be throttled. They were told that they have data in the bundle to see any of the versions of the videos but as opposed to Scenario 2 they were told to assume that once the data is exceeded their bandwidth will be reduced to 64 kbps. To make sure that the participants understand what that means in terms of downloading a video clip, they were given the approximate download time for each video. However they were told as in Scenario 2 that they have data in the bundle to watch any of the videos. 3.3. Findings A multiple paired t-test (Black, 2011) (N = 60) was used to compare the data across different billing plans. A 99% confidence interval was considered for statistical significance. Table 3 presents a detailed description of the findings. The results show that there is a statistically significant difference between the flat rate billing plan (Scenario 1) and both capped billing plans regardless if the user still has data in the bundle or paying for the data outside the bundle (Scenario 1 & Scenario 2 – p = 0.001, Scenario 1 & Scenario 3 – p < 0.001, Scenario 1 & Scenario 4 – p < 0.001). There is also a statistically significant difference between the scenarios in which the user has data in the bundle but s/he will have to pay extra when the data in the bundle is exceeded and the scenario in which the user has the bandwidth throttled (Scenario 2 & Scenario 4 – p < 0.001). However no statistically significant difference was obtained between the scenario in which the user is paying for the exceeding data and when the bandwidth is throttled but the user still has data in the bundle (Scenario 3 & Scenario 4, p = 0.306). To better understand the participants’ choices and how they differ based on the given scenario, user choices for each scenario are presented in Fig. 1. With 1 we have coded the video with the highest quality and with 5 the video with the lowest quality. It can be seen that the users selected higher quality when the mobile data billing plan is flat rate (Scenario 1) and the lowest when they are exceeding the quantity of data in the bundle and they have to pay extra for it (Scenario 3). Although in Scenario 2 and Scenario 4 the participants still have data in the bundle to watch any of the videos, the users in Scenario 4 Table 1 Video clips encoding and quality. Resolution
Bit rate (kbps)
Size (MB)
Frame rate (fps)
Coding format
SSIM index
MOS
Video versions 800 400 800 400 480 320 480 320 480 320
1000 600 550 350 150
14 9.15 8.46 5.99 3.65
29.97 29.97 29.97 29.97 29.97
MPEG-4 MPEG-4 MPEG-4 MPEG-4 MPEG-4
0.95 0.93 0.91 0.84
5 4 3 3 2
– – – – –
Excellent Good Fair Fair Poor
230
A. Molnar / Telematics and Informatics 33 (2016) 227–231 Table 2 Size and price of the video clips. Size (MB)
Price – Scenario 3
Approximative download time for Scenario 4
Video versions 14 €14 9.15 €9.15 8.46 €8.46 5.99 €5.99 3.65 €3.65
3 min 2 min 2 min 1 min 57 s
and and and and
39 s 23 s 12 s 34 s
Table 3 Paired t-test results.
r
Pairs Scenario Scenario Scenario Scenario Scenario Scenario
1 1 1 2 2 3
& & & & & &
Scenario Scenario Scenario Scenario Scenario Scenario
2 3 4 3 4 4
t
1.5567 1.4046 1.5112 1.319 1.525 1.251
p 3.400 8.824 7.347 5.385 3.809 1.032
.001 <.001 <.001 <.001 <.001 .306
70% 60% 50% Scenario 1
40%
Scenario 2
30%
Scenario 3
20%
Scenario 4
10% 0% 1- Highest Quality
2
3
4
5- Lowest Quality
Fig. 1. Participants video choice in each scenario. Table 4 Pearson correlation (Lawrence and Lin, 1989): results. Scenario 1
Scenario 2
Age r r (2-tailed)
Scenario 3
Scenario 4
.009 .944
.083 .526
.027 .840
.035 .791
Gender r r (2-tailed)
.084 .523
.111 .399
.030 .817
.225 .084
selected a lower quality than the ones in Scenario 2. Therefore people that have their bandwidth throttled (Scenario 4) seem savvier in saving the bundle data than the ones who have to pay extra once the quality is exceeded. We also assessed whether age or gender had an influence on responses. The results showed no statistically significant correlation (at 0.01 level) and a weak correlation between age and user choices regardless of the scenario. The same results were obtained for gender. Table 4 presents the statistical results in more details. 4. General discussions This article has examined how people’s preferences for video quality differ under different mobile data billing plans. The results have shown that the participants’ preferences for different video quality differ across the billing plans and the differences between their choices are statistically significant. A single exception was found in this study. No statistically significant difference was found between Scenario 3 (under the scenario in which the participants assumed that they exceeded the quantity of data in the bundle and they have to pay extra for the data consumed) and Scenario 4 (under the scenario in which the participants were asked to assume that their bandwidth will be throttled when the cap is reached, however they still have data in the bundle to watch any of the videos). The results also show that the participants selected a higher video
A. Molnar / Telematics and Informatics 33 (2016) 227–231
231
quality when they have a flat rate plan but they preferred a lower quality when their mobile data plan is capped. Among the capped billing plans, if the user still has data in the bundle, the results show that s/he is more likely to access a higher quality video content once the cap is reached if s/he is able to pay extra than when his bandwidth is throttled. The findings of this research provide insights into user video quality preference under different billing plans. The results could help network providers and operators to prioritise their needs and user needs. If the network providers’ have limited resources and their aim is to avoid congestion these results indicate that using a billing plan that throttles the bandwidth once the cap limit is exceeded may be the best option. However if increasing revenue was the main aim, capping the billing plan but leaving the user the option to pay extra when the quantity is exceeded would provide the compromise between congestion avoidance and increasing revenue. For the content providers the results of this study provide a way of reducing traffic to their servers but also a way to determine whether their customers would prefer a lower quality due to the restrictions of their billing plans. This would lead to a ‘‘win–win’’ situation in which the users’ satisfaction will improve as they would be able to consume more content through their mobile data plans and the content providers as a result will increase their revenues. Moreover, this research could be used to improve existing adaptation algorithms through a greater understanding of decisions that people make under different billing plans. For the users this could provide a better way of addressing their data consumption and making the best of their data billing plans and therefore increasing the customer experience. Although some of the information provided to the user in our study is not always provided to the user in real life, this study presents the results of user choices when given the chance to make an informed decision. 5. Conclusion This research presents the results of assessing user preferences for video quality under three mobile data billing plans: (1) flat rate billing plan; (2) capped billing plan in which the user pays extra when exceeding the data included in the bundle; and (3) capped billing plan in which the user’s bandwidth is throttled if s/he exceeds the quantity of data included in the bundle. The results show that the users preferences for video quality under different billing plan differ and these difference are statistically significant, with the exception of the following billing plan scenarios: (a) when the user has a capped billing plan which allows him to pay extra for the exceeding data and (b) when the user’s bandwidth is throttled. In the before mentioned scenarios no statistically significant difference was obtained. Moreover, the results show that the user would select a higher quality video when they have a flat rate than when they have a capped billing plan. Among the two capped billing plans, when the participants had a capped billing plan that allows them to pay extra if the cap is exceeded, they preferred a higher video quality than when they were asked to assume that they have a billing plan in which the bandwidth is throttled, as long as they still have data left in the bundle to consume. Further studies are required to confirm or infirm the results by using different video categories, quality levels and billing plans. References Black, T.R., 2011. Doing Quantitative Research in the Social Sciences: An Integrated Approach to Research Design, Measurement and Statistics. Sage. Chih-Lin, I., Rowell, C., Han, S., Xu, Z., Li, G., Pan, Z., 2014. Toward green and soft: a 5G perspective. IEEE Commun. Mag. 52 (2), 66–73. Dosselmann, R., Yang, X.D., 2011. A comprehensive assessment of the structural similarity index. SIViP 5 (1), 81–91. Higginbotham, S., 2013. Want to Know if Your ISP is Capping Data? Check our updated chart (Online). Available:. ITU-T P.910, 2008. Series P: Telephone Transmission Quality, Telephone Installations, Local Line Networks (Online). Available: . Isomursu, P., Hinman, R., Isomursu, M., Spasojevic, M., 2007. Metaphors for the mobile Internet. Knowledge Technol. Policy 20 (4), 259–268. Jung, W., Kwon, Y., 2015. Differences between LTE and 3G service customers: business and policy implications. Telematics Inform. 32 (4), 667–680. Koenigstorfer, J., Groeppel-Klein, A., 2012. Consumer acceptance of the mobile Internet. Marketing Lett. 23 (4), 917–928. Kumar, A., Shankar, R., Debnath, R.M., 2015. Analyzing customer preference and measuring relative efficiency in telecom sector: a hybrid fuzzy AHP/DEA study. Telematics Inform. 32 (3), 447–462. Lawrence, I., Lin, K., 1989. A concordance correlation coefficient to evaluate reproducibility. Biometrics, 255–268. Lee, S., Park, E.A., Lee, S., Brown, J., 2015. Determinants of IPTV diffusion. Telematics Inform. 32 (3), 439–446. Marcus, J.S., Godlovitch, I., 2013. Mobile Traffic Off-Koad and Fixed-Mobile Competition. Available at SSRN 2342639. Molnar, A., Muntean, C.H., 2010. Educational content delivery: an experimental study assessing student preference for multimedia content when monetary cost is involved. In: 10th International Conference on Intelligent Systems Design and Applications, pp. 871–876. Molnar, A., Muntean, C.H., 2013. Cost-oriented adaptive multimedia delivery. IEEE Trans. Broadcast. 59 (3), 484–499. Molnar, A., Muntean, C. Hava, 2015. What can people risk attitude tell us about people preference for multimedia quality? Inf. Technol. People 28 (2), 383– 404. Newman, J., 2011. Netflix Has Bandwidth Cap Sufferers Covered. PC World (Online). Available: . Odlyzko, A., 2001. Internet pricing and the history of communications. Comput. Netw. 36 (5), 493–517. Pande, A., Ahuja, V., Sivaraj, R., Baik, E., Mohapatra, P., 2013. Video delivery challenges and opportunities in 4G networks. IEEE Multimedia 20 (3), 88–94. Reichardt, C.S., 2009. Quasi-experimental design. In: The SAGE Handbook of Quantitative Methods in Psychology, SAGE, pp. 46–71. Roto, V., Geisler, R., Kaikkonen, A., Popescu, A., Vartiainen, E., 2006. Data traffic costs and mobile browsing user experience. in: Proceedings of the 4th MobEA Workshop on Empowering the Mobile Web, pp. 1–6. Sen, S., Joe-Wong, C., Ha, S., Chiang, M., 2013. A survey of smart data pricing: past proposals, current plans, and future trends. ACM Comput. Surv. (CSUR) 46 (2), 15. Trestian, I., Ranjan, S., Kuzmanovic, A., Nucci, A., 2011. Taming user-generated content in mobile networks via drop zones. Proc. IEEE INFOCOM, 2840–2848. Trestian, I., Ranjan, S., Kuzmanovic, A., Nucci, A., 2012. Taming the mobile data deluge with drop zones. IEEE/ACM Trans. Netw. 20 (4), 1010–1023. Zinner, T., Abboud, O., Hohlfeld, O., Hossfeld, T., Tran-Gia, P., 2010. Towards QoE management for scalable video streaming. In: ITC Specialist Seminar on Multimedia Applications-Traffic, Performance and QoE, pp. 64–69.
Contents lists available at ScienceDirect
Telematics and Informatics journal homepage: www.elsevier.com/locate/tele
Video quality vs. mobile data billing plans Andreea Molnar Portsmouth University, Winston Churchill Avenue, Portsmouth PO1 2UP, United Kingdom
a r t i c l e
i n f o
Article history: Received 25 June 2015 Received in revised form 24 July 2015 Accepted 24 July 2015 Available online 26 July 2015 Keywords: Consumer behaviour User choices Video quality Mobile data Bundle based billing Capped billing plans
a b s t r a c t Despite extensive research regarding the user willingness to pay under different billing plans and prices it is still unclear how users choose a preferred video quality under existing mobile data plans. This is especially important as the usage of adaptive video delivery servers that make use of different video quality levels is increasing. This study presents the results of assessing user preferences for video quality under three different mobile data billing plans. The results show that the user choice of different quality of video content depends on the billing plans used. It also shows that in the case when the mobile data plan is capped but the user has the possibility to pay extra for the exceeding quantity, s/he will choose higher video quality compared to the same plan in which the bandwidth is throttled when the bandwidth is exceeded. No correlation was found between the user choices and age or gender. Ó 2015 Elsevier Ltd. All rights reserved.
1. Introduction Mobile Internet traffic is characterised by an increase in number of subscribers (Pande et al., 2013) and an exponential increase in video content (Trestian et al., 2012) through an increase of both user generated content (Trestian et al., 2011) and access to other services, such as video through mobile networks and devices (Lee et al., 2015). Video content is usually heavier than other types of content and consumes more bandwidth. In an effort to impede congestion and increase revenues, mobile network operators that initially offered unlimited data access, realised that flat rate plans are no longer feasible and decided to cap them. Reviews of mobile prices in major European countries (Molnar and Muntean, 2013), and not only (Sen et al., 2013), have shown that most mobile data plans are cap based (aka bundle based, buckle based). Although capped billing plans are mostly present in mobile networks, broadband network operators also make use of them (Higginbotham, 2013). Capping the amount of data the user can consume is likely to continue even with the deployment of more powerful networks (Pande et al., 2013; Chih-Lin et al., 2014). Investing in new network technology is seen by mobile network operators as a way to impede congestion, and increase the revenues for network operators by removing unprofitable user behaviour (i.e. consumers that use a large amount of data). In this context, capped billing plans can be seen as a means of controlling network traffic and ensure a ‘‘fair-usage’’ policy for Internet resources. As another measure to avoid congestion, content providers adapt the video content to the network conditions, a process that leads to lower size data that will be delivered over the network at the expense of the video quality. For example, YouTube automatically selects a lower quality video when the request is performed from a mobile network rather than over a Wi-Fi or a wired network and Netflix allows the users to select a lower video quality in order to avoid reaching the cap limit (Newman, 2011).
E-mail address: [email protected] http://dx.doi.org/10.1016/j.tele.2015.07.009 0736-5853/Ó 2015 Elsevier Ltd. All rights reserved.
228
A. Molnar / Telematics and Informatics 33 (2016) 227–231
Despite the increasing use of capped billing plans on mobile networks and the increased usage of video over these types of networks, our understanding of how people would choose between different video qualities under different billing plans remains poor. The goal of this research is to partially fill this void by providing further understanding on the user choices for video quality under different prices. Without knowing their users, mobile network operators cannot optimise their plans for marketing purposes (Jung and Kwon, 2015). This would also help improve video adaptation algorithms leading to a better user quality of experience, which could help retain customers. Moreover, it could maximise the resources mobile operators have. With this aim, a quasi-experimental study (Reichardt, 2009) was designed to assess how users select different video qualities under billing plan constraints. Three different billing plans are considered for the study: a flat rate billing plan and two types of capped billing plans: in one billing plan, once the user reaches the data cap, the user has to pay extra for the exceeding quantity whereas in the other billing plan, once the user consumed the data from the bundle, the user bandwidth is throttled. The rest of this paper is organised as following. Section two introduces existing research on mobile Internet consumer behaviour. Section three presents the study set-up and the findings. Section four discusses the results and section five presents the study’s conclusions.
2. Related work Several studies (Koenigstorfer and Groeppel-Klein, 2012; Kumar et al., 2015; Jung and Kwon, 2015; Molnar & Muntean, 2015) have assessed consumers’ usage of mobile data. The study presented in Koenigstorfer and Groeppel-Klein (2012) assessed how consumer personality affects the consumer acceptance of the mobile Internet, showing that innovativeness, low desire for social contact and technology optimism influence consumer acceptance. Molnar and Muntean (2015) looked at how the user risk attitude affects user’s preference of video quality under monetary cost constraints, showing that risk adverse people are more willing to trade off quality for price than the risk seekers. Kumar et al. (2005) and Jung and Kwon (2015) assess factors that affect customer satisfaction and preference with mobile network operators. Jung and Kwon (2015) have shown through a study performed in Korea that by improving the customer satisfaction, the mobile network operators could improve the loyalty of their 3G and LTE users. They have also shown that 3G subscribers are more sensitive to call services and that LTE users are more sensitive to quality of data service and pricing. Kumar et al. (2015) show that mobile network parameters and tariff schemes are affecting customer preference towards a mobile subscriber. Other studies have looked into how consumer behaviour is characterised under different billing plans. For example, it has been shown that people are optimising their time spent on the Internet if the billing plan is not unlimited (Roto et al., 2006) and people own past experiences or highly publicised stories that report high billing plans are likely to have an effect on how people use and interact with mobile data (Isomursu et al., 2007). Marcus and Godlovitch (2013) showed that consumers use more intensively mobile data when they are connected to a Wi-Fi network and speculate that the reason may be rationing the consumption on mobile phones. Previous research has also shown that users are also uncertain about how much they consume and dislike bill shocks preferring flat rate billing plans due to their inherent predictability (Odlyzko, 2001). Molnar and Muntean (2010) have shown that some people may prefer a lower quality video under capped billing plans than what they would choose under a flat rate billing plan, however the study did not further assess how the user choices differ among existing billing plans. This study aims to address this gap by assessing how people choose between different quality levels on currently used mobile data billing plans.
3. Study 3.1. Sample and data collection A quasi-experimental design was followed in which the participants had to select among videos of different qualities given the constraint of different billing plans. The study was advertised and the participants volunteered to participate in the study. A total of 60 people took part in the study, 75% males and 25% females. The ages varied from 20 to 57 years old, with an average age of 32 years old. 3.2. Procedure The billing plans used in the experimental study are selected based on the results of the survey containing the most common billing plans in Europe presented in (Molnar and Muntean, 2013). Three scenarios were designed based on existing mobile data billing plans. Two capped billing plans were selected: one in which the user has to pay extra for exceeding the threshold included in the bundle and another one in which after the bundle data is exceeded the bandwidth is throttled. Although a flat rate data plan is not common on the mobile Internet, one of the scenarios included a flat rate billing plan for comparison. The following billing plans were selected:
229
A. Molnar / Telematics and Informatics 33 (2016) 227–231
Billing plan 1: Flat rate billing plan that allows unlimited data consumption. Billing plan 2: Capped daily billing plan in which for €0.99 per day the user gets 50 MB data included. If the bundle is exceeded they have to pay €1/MB. Billing plan 3: Capped billing plan in which the user has 50 MB data per day. After the quantity is exceeded, their bandwidth would be throttled to 64 kbps. The users were asked to assume that they have one of the above billing plans and select among different quality levels for a video clip. The study took place in a controlled environment following a similar design as recommended by the International Telecommunication Union (ITU-T P.910, 2008). The participants were given the same smartphone, a Google Nexus, with the following characteristics: resolution of 480 800, 512 MB internal memory and 1 GHz CPU. The video clips were encoded at five different quality levels similar with the recommendations contained in (Molnar & Muntean, 2015) when the adaptation for this type of device is to be performed. Table 1 presents the encoding of the video clips and the Mean Opinion Score (MOS). MOS was obtained by computing Structural Similarity (SSIM) index (Dosselmann and Yang, 2011) and mapping it to the MOS as described in Zinner et al. (2010). The participants were allowed to watch the video clips as many times as they wanted to but they were asked to watch all versions of the same video at least once in order to make an informed decision. The participants were able to see how long the video was (each clip’s playtime was slightly less than 2 min). Four scenarios were designed to investigate how the subjects’ preferences towards the quality of the video content changes when monetary cost is involved. These scenarios were designed based on the three billing plans above: one scenario for billing plan 1 and billing plan 3 and two for billing plan 2 (one in which the user has data in the acquired bundle and another one in which s/he is exceeding the bundle). Each participant was involved in all of the four scenarios and during all the scenarios they were encouraged to watch the video versions as many times as they needed to. Scenario 1: Participants were asked to assume that they have a billing plan that allows them to use unlimited data and were asked to select one of the video clip versions. Scenario 2: Participants are asked to assume that they have a daily billing plan in which for €0.99 per day they get 50 MB data included. They were told that they have data in the bundle to see any of the versions of the videos regardless of the size. Scenario 3: Participants are asked to assume that they have exceed the data in the bundle presented in Scenario 2 and as a result they will have to pay €1/MB if they consume data. The prices for each video version were presented to them (see Table 2 for the prices of each video clip). Also in this case they were asked to select one of the five versions of the video clip. Scenario 4: Participants are asked to assume that they have a capped billing plan but in the case when the cap is exceeded the bandwidth will be throttled. They were told that they have data in the bundle to see any of the versions of the videos but as opposed to Scenario 2 they were told to assume that once the data is exceeded their bandwidth will be reduced to 64 kbps. To make sure that the participants understand what that means in terms of downloading a video clip, they were given the approximate download time for each video. However they were told as in Scenario 2 that they have data in the bundle to watch any of the videos. 3.3. Findings A multiple paired t-test (Black, 2011) (N = 60) was used to compare the data across different billing plans. A 99% confidence interval was considered for statistical significance. Table 3 presents a detailed description of the findings. The results show that there is a statistically significant difference between the flat rate billing plan (Scenario 1) and both capped billing plans regardless if the user still has data in the bundle or paying for the data outside the bundle (Scenario 1 & Scenario 2 – p = 0.001, Scenario 1 & Scenario 3 – p < 0.001, Scenario 1 & Scenario 4 – p < 0.001). There is also a statistically significant difference between the scenarios in which the user has data in the bundle but s/he will have to pay extra when the data in the bundle is exceeded and the scenario in which the user has the bandwidth throttled (Scenario 2 & Scenario 4 – p < 0.001). However no statistically significant difference was obtained between the scenario in which the user is paying for the exceeding data and when the bandwidth is throttled but the user still has data in the bundle (Scenario 3 & Scenario 4, p = 0.306). To better understand the participants’ choices and how they differ based on the given scenario, user choices for each scenario are presented in Fig. 1. With 1 we have coded the video with the highest quality and with 5 the video with the lowest quality. It can be seen that the users selected higher quality when the mobile data billing plan is flat rate (Scenario 1) and the lowest when they are exceeding the quantity of data in the bundle and they have to pay extra for it (Scenario 3). Although in Scenario 2 and Scenario 4 the participants still have data in the bundle to watch any of the videos, the users in Scenario 4 Table 1 Video clips encoding and quality. Resolution
Bit rate (kbps)
Size (MB)
Frame rate (fps)
Coding format
SSIM index
MOS
Video versions 800 400 800 400 480 320 480 320 480 320
1000 600 550 350 150
14 9.15 8.46 5.99 3.65
29.97 29.97 29.97 29.97 29.97
MPEG-4 MPEG-4 MPEG-4 MPEG-4 MPEG-4
0.95 0.93 0.91 0.84
5 4 3 3 2
– – – – –
Excellent Good Fair Fair Poor
230
A. Molnar / Telematics and Informatics 33 (2016) 227–231 Table 2 Size and price of the video clips. Size (MB)
Price – Scenario 3
Approximative download time for Scenario 4
Video versions 14 €14 9.15 €9.15 8.46 €8.46 5.99 €5.99 3.65 €3.65
3 min 2 min 2 min 1 min 57 s
and and and and
39 s 23 s 12 s 34 s
Table 3 Paired t-test results.
r
Pairs Scenario Scenario Scenario Scenario Scenario Scenario
1 1 1 2 2 3
& & & & & &
Scenario Scenario Scenario Scenario Scenario Scenario
2 3 4 3 4 4
t
1.5567 1.4046 1.5112 1.319 1.525 1.251
p 3.400 8.824 7.347 5.385 3.809 1.032
.001 <.001 <.001 <.001 <.001 .306
70% 60% 50% Scenario 1
40%
Scenario 2
30%
Scenario 3
20%
Scenario 4
10% 0% 1- Highest Quality
2
3
4
5- Lowest Quality
Fig. 1. Participants video choice in each scenario. Table 4 Pearson correlation (Lawrence and Lin, 1989): results. Scenario 1
Scenario 2
Age r r (2-tailed)
Scenario 3
Scenario 4
.009 .944
.083 .526
.027 .840
.035 .791
Gender r r (2-tailed)
.084 .523
.111 .399
.030 .817
.225 .084
selected a lower quality than the ones in Scenario 2. Therefore people that have their bandwidth throttled (Scenario 4) seem savvier in saving the bundle data than the ones who have to pay extra once the quality is exceeded. We also assessed whether age or gender had an influence on responses. The results showed no statistically significant correlation (at 0.01 level) and a weak correlation between age and user choices regardless of the scenario. The same results were obtained for gender. Table 4 presents the statistical results in more details. 4. General discussions This article has examined how people’s preferences for video quality differ under different mobile data billing plans. The results have shown that the participants’ preferences for different video quality differ across the billing plans and the differences between their choices are statistically significant. A single exception was found in this study. No statistically significant difference was found between Scenario 3 (under the scenario in which the participants assumed that they exceeded the quantity of data in the bundle and they have to pay extra for the data consumed) and Scenario 4 (under the scenario in which the participants were asked to assume that their bandwidth will be throttled when the cap is reached, however they still have data in the bundle to watch any of the videos). The results also show that the participants selected a higher video
A. Molnar / Telematics and Informatics 33 (2016) 227–231
231
quality when they have a flat rate plan but they preferred a lower quality when their mobile data plan is capped. Among the capped billing plans, if the user still has data in the bundle, the results show that s/he is more likely to access a higher quality video content once the cap is reached if s/he is able to pay extra than when his bandwidth is throttled. The findings of this research provide insights into user video quality preference under different billing plans. The results could help network providers and operators to prioritise their needs and user needs. If the network providers’ have limited resources and their aim is to avoid congestion these results indicate that using a billing plan that throttles the bandwidth once the cap limit is exceeded may be the best option. However if increasing revenue was the main aim, capping the billing plan but leaving the user the option to pay extra when the quantity is exceeded would provide the compromise between congestion avoidance and increasing revenue. For the content providers the results of this study provide a way of reducing traffic to their servers but also a way to determine whether their customers would prefer a lower quality due to the restrictions of their billing plans. This would lead to a ‘‘win–win’’ situation in which the users’ satisfaction will improve as they would be able to consume more content through their mobile data plans and the content providers as a result will increase their revenues. Moreover, this research could be used to improve existing adaptation algorithms through a greater understanding of decisions that people make under different billing plans. For the users this could provide a better way of addressing their data consumption and making the best of their data billing plans and therefore increasing the customer experience. Although some of the information provided to the user in our study is not always provided to the user in real life, this study presents the results of user choices when given the chance to make an informed decision. 5. Conclusion This research presents the results of assessing user preferences for video quality under three mobile data billing plans: (1) flat rate billing plan; (2) capped billing plan in which the user pays extra when exceeding the data included in the bundle; and (3) capped billing plan in which the user’s bandwidth is throttled if s/he exceeds the quantity of data included in the bundle. The results show that the users preferences for video quality under different billing plan differ and these difference are statistically significant, with the exception of the following billing plan scenarios: (a) when the user has a capped billing plan which allows him to pay extra for the exceeding data and (b) when the user’s bandwidth is throttled. In the before mentioned scenarios no statistically significant difference was obtained. Moreover, the results show that the user would select a higher quality video when they have a flat rate than when they have a capped billing plan. Among the two capped billing plans, when the participants had a capped billing plan that allows them to pay extra if the cap is exceeded, they preferred a higher video quality than when they were asked to assume that they have a billing plan in which the bandwidth is throttled, as long as they still have data left in the bundle to consume. Further studies are required to confirm or infirm the results by using different video categories, quality levels and billing plans. References Black, T.R., 2011. Doing Quantitative Research in the Social Sciences: An Integrated Approach to Research Design, Measurement and Statistics. Sage. Chih-Lin, I., Rowell, C., Han, S., Xu, Z., Li, G., Pan, Z., 2014. Toward green and soft: a 5G perspective. IEEE Commun. Mag. 52 (2), 66–73. Dosselmann, R., Yang, X.D., 2011. A comprehensive assessment of the structural similarity index. SIViP 5 (1), 81–91. Higginbotham, S., 2013. Want to Know if Your ISP is Capping Data? Check our updated chart (Online). Available: