1. © 2009 TKK & NOKIA 1 Mobile Service Usage and Business Models in Wireless Local Area Networks Beyene Abebe 17.08.2009 Supervisor Prof. Heikki Hämmäinen Instructors Timo Smura, Thomas Casey M.Sc.(Tech.) Projects MOMI & IMCOS

2. © 2009 TKK & NOKIA 2 Outline Definition Research questions and scope Research methods and tools Results from handset-based measurement analysis System dynamics modeling of local area access value networks Discussion

3. © 2009 TKK & NOKIA 3 Definition Mobile Service Usage and Business Models in Wireless Local Area Networks - the service that an end-user with a mobile device receives from the network operator or a 3 rd party service provider. e.g. voice/video calls, SMS and MMS messages, access to content on web, email, ticketing or booking and etc. Mobile Application - “either network or handset-based pieces of software that run services” e.g. Music Players, Logs, Calendars, Contacts and Handset Clocks

4. © 2009 TKK & NOKIA 4 Definition (2) “ Application is a more technical term referring to the technical solution, whereas service is the whole entity as seen by end-users.” Sometimes the terms service and application are used interchangeably, for example, in case of voice calling and SMS messaging applications. Mobile Service Usage and Business Models in Wireless Local Area Networks -“ Description of how a company or a set of companies intend to create and capture value with a product or service by linking new technological environments to business strategies ” Mobile Service Usage and Business Models in Wireless Local Area Networks - wireless networks which provide local area access with high data-transfer rates. E.g. Wi-Fi

5. © 2009 TKK & NOKIA 5 Research questions and scope (1/2) 1)How do people use mobile devices (especially indoor located devices) with different access technologies? (extension of the work reported in Smura (2008)) Scope Devices Handsets Markets Finnish market

6. © 2009 TKK & NOKIA 6 Research questions and scope (2/2) 2)How is network connectivity to indoor located devices provided in the future? What are the possible evolution paths? The concept of system dynamics modeling is applied Scope Devices All mobile and portable devices included But more focus on handsets and laptop devices Markets Finnish market Traffic type Data traffic Time Frame 2009-2015

7. © 2009 TKK & NOKIA 7 Research methods and tools Handset-based mobile service usage measurement Data was collected from two panels of Finnish smartphone users during October – December 2008 Finnish2008 panel OtaSizzle panel A handset-based measurement platform (developed for Nokia S60 class of mobile devices) was used to collect the data The platform consists of a Symbian application monitoring software client which is installed to the mobile devices 223 panelists successfully installed the software client 70 were excluded due to too few active days in the panel 4 were excluded because they were only testing the software client and used unclear foreign language settings. Therefore, 149 panelists with S60 3rd edition devices were included in this study (105 and 44 from Finnish2008 panel and OtaSizzle pannel,respectively) System Dynamics Vensim PLE software Brainstorming session Conducted with a group of experts from Nokia Research Centre (NRC) on 14th of May 2009

8. Results from handset-based measurements analysis © 2009 TKK & NOKIA 8

9. 9 Usage of different applications (1/3) When and for which application do people use their mobile device? Figure 1:Distribution of active smartphone usage time between application categories and hours of day, average over Monday-Sunday throughout the panel period. N=149 Voice calling application is dominant on average. Voice calls accounted for 30.7% of the total average daily active usage seconds per panelist. Browsing, messaging and business and productivity accounted for 21.9%, 20.1% and 8.5%, respectively while multimedia accounted for 6.5%.

10. © 2009 TKK & NOKIA 10 Usage of different applications (3/3) The most actively used applications under different application categories Application categoriesApplication namesAverage usage_seconds/day/panelist Web405.53 BrowsingServices36.36 Opera Mini89.3 Text message272.82 MessagingMMS7.21 Music player55.61 MultimediaGallery26.28 Camera34.27 Calendar57.36 Business and productivityContacts39.9 Log29.84 Web, text message, music player and calendar are the most widely used applications under their category. Table 1: Actively used mobile applications under different application categories. N=149

11. © 2009 TKK & NOKIA 11 Usage of different access networks (1/3) When and which bearer technology do people use to start data session? Figure 3:Distribution of daily data usage between bearers and hours of day, average over Monday-Sunday. N=130 On average the daily data usage in kB for WCDMA is higher than other bearer technologies.

12. © 2009 TKK & NOKIA 12 Usage of different access networks (2/3) Figure 4: Distribution of active usage time in seconds between bearers and hours of day, average over Monday-Sunday for N=130 The active usage (in seconds) for WCDMA is also greater than other bearer technologies on average  Figure 3 and Figure 4 show that WCDMA is the most widely used bearer technology.

13. © 2009 TKK & NOKIA 13 Usage of different access networks (3/3) Users’ choice between access networks while launching various applications Table 3: Usage of application categories during WCDMA, WLAN, GPRS and EDGE connections. N=130 More users for browsing applications For multimedia applications, less number of sessions per user However,high amount of data transmitted per session A smartphone data session during WLAN connections was higher in terms of data volumes than during wide area network connections (for browsing and messaging) More than 85% of WLAN users used WLAN connections for web browsing, while 32% used it for communications services such as messaging and 15% for streaming multimedia content  WLAN is mainly used by those users who in general use large amounts of data services.

14. © 2009 TKK & NOKIA 14 Daily data usage of WLAN vs. non-WLAN users Figure 5: Daily data usage per bearer for panelists either using or not using WLAN during the panel. N=130 The daily data usage of WLAN users was higher than that of non-WLAN users in all types of bearer technologies. The share of WLAN access increased from 12% of total network data traffic in 2007 (Smura, 2008) to 39% in 2008 The share of WCDMA access dropped from 79% among non-WLAN users to 50% among WLAN users  WLAN can be considered as a potential substitute to 3G networks for usage of data services.

15. © 2009 TKK & NOKIA 15 Users’ choice between alternative access methods Figure 6: Distribution of WLAN usage events between users and access points. N=53 On average, for each user, 80% of WLAN connections were made to the first access point, 12% to a second one, and 8% to other access points. 91% of all WLAN connections were made from private access points while the rest was from public ones  Most of WLAN connections are made from private access points located at home or office.

16. © 2009 TKK & NOKIA 16 Conclusions Voice call is still a dominant application category used by mobile users followed by browsing, messaging, business and productivity, multimedia and others in order. WLAN is a potential alternative access to 3G networks for usage of data services. WLAN is mainly used by those users who in general use large amounts of data services. 91% of all WLAN connections were made from private access points while the rest was from public ones The share of WLAN access increased from 12% of total network data traffic in 2007 (Smura, 2008) to 39% in 2008 Limitation: limited number of panelists

17. © 2009 TKK & NOKIA 17 System Dynamic Modeling of Local Area Access Value Networks

18. © 2009 TKK & NOKIA 18 Basics of System Dynamics Causal loops diagrams Stock and flow

19. © 2009 TKK & NOKIA 19 Background Goal to continue the Scenario planning work conducted by Smura & Sorri 2009 Research question How is network connectivity to indoor located devices provided in the future? What are the possible evolution paths? Time Frame: 2009-2015 Scope Devices All mobile and portable devices included But more focus on mobile and laptop devices Markets → Focus on Finnish market Traffic type → Data traffic studied The overall goal is to understand dynamic relationships between forces not to model exact numerical values (quantitative modeling based on literature and expert opinion) System dynamic modeling is an iterative process and the results presented here should serve as the first iteration round

20. Scenarios Horizontal industry structure Vertical industry structure WA-LA divorce WA-LA marriage 1. Pick-n-mix - Internet rules 2. Complete bundles - Operator rules 3. Operators as bitpipes 4. Internet giants Source: Smura & Sorri (2009)

21. © 2009 TKK & NOKIA 21 Cause for disruptions? (”muljahdus”) Vertical Integration (services + NW access?) U 1 Level of access techology fragmentation (Wide Area (WA)+ Local Area (LA)?) U 2, U 3, U 4 Source: Smura & Sorri (2009)

22. © 2009 TKK & NOKIA 22 Uncertainties Trends T31: Num of non-3GPP LA APs T33: Number of 3GPP WA BSs T32: Number of 3GPP LA APs Main Forces used in SD model U2: Competition between technology substitutes T1: Devices’ capabilities and performance improve U1: Industry structure T2: Wireless traffic will increase T4: Importance of indoor wireless access increases U3: Spectrum policy and regulation U4: Role of unlicensed spectrum T6: Operational costs will dominate over hardware costs T3: Number of base stations / access points increases U5: Number of connected devices U6: Role of emerging markets in affecting technology choices T5: Role of developing countries increasing T7: Wireless emissions scare people U22: Share of LA out of all indoor traffic U23:market share of 3GPP operator of all indoor traffic U21: Share of 3GPP LA out of all LA T12: AP self optimizing capability T11: Device access selection capabilities Source: Smura & Sorri (2009)

23. © 2009 TKK & NOKIA 23 Conceptual Model Background assumptions: Unserved demand drives the supply The capacity of backhaul network is not a bottleneck (since a major part of the traffic is local)

24. © 2009 TKK & NOKIA 24 The model has four domains: 1. User (Demand) 2. Infrastructure (Supply) 3. Spectrum Regulation and Technology 4. Market Share (defined in terms of traffic volume)

25. © 2009 TKK & NOKIA 25 1.User (Demand) Demand based on two components: Primary devices (includes laptops, smartphones) Secondary devices (household appliances etc.)

26. © 2009 TKK & NOKIA 26 2.Infrastructure (Supply) Infrastructure is expanded based on Unserved demand Relative market shares of each technology Yearly traffic volume supply of each technology

27. © 2009 TKK & NOKIA 27 3.Spectrum Regulation and Technology Yearly traffic volume supply calculated for each technology based on: Capacity of the technology (spectrum and technology development) Simple traffic model (active yearly usage time)

28. © 2009 TKK & NOKIA 28 4.Market Share U23 = 0 In the model variable U23 (Market share of 3GPP operator) indicates the level of fragmentation in local access provisioning U23 = 1 WA-LA Divorce/ fragmented access WA-LA Marriage/intege rated access

29. Results and Sensitivity Analysis Base Case - refers to our system dynamics model simulated with the initial values Base Case Configurations (Initial values) Typical Finnish suburban area with 2300 users/km² Indoor access provisioning is fragmented (U23 ≈ 0.02) T31: Num of non-3GPP LA APs (WiFis) = 300 APs/km² T32: Number of 3GPP LA APs (Femtocells) = 10 APs/km² T33: Number of 3GPP WA BSs = 10 sectors/km² © 2009 TKK & NOKIA 29

30. Base Case Results © 2009 TKK & NOKIA 30 The capacity of the infrastructure is sufficient until 2011 when demand surpasses supply MB/(Km2*year) = MB/Km2/year (Vensim feature)

31. Base Case Results (2) © 2009 TKK & NOKIA 31 The maximum WA BS density is quickly reached after demand surpasses supply 150% growth 500% growth

32. Base Case Results (3) © 2009 TKK & NOKIA 32 After some delay unserved demand drives spectrum policy and regulation to become slightly more flexible

33. Base Case Results (4) © 2009 TKK & NOKIA 33 The growth in the number of WA BSs increases the market share of3GPP operators but the installed base of WiFi access points dominates (role of femtocells marginal) Note that when U23 = 0 → fragmented case And when U23 = 1 → integrated case

34. Base Case Conclusion Indoor access provisioning is fragmented (U23 ≈ 0.03) © 2009 TKK & NOKIA 34

35. Sensitivity cases © 2009 TKK & NOKIA 35

36. Case 1: High device access selection capabilities © 2009 TKK & NOKIA 36

37. Case 1 Results © 2009 TKK & NOKIA 37 Higher device selection capabilities increase demand and supply Less unserved demand

38. Case 1 Results (2) © 2009 TKK & NOKIA 38 Favors the growth of Wi-Fi access points (non-3GPP LA APs)  The market share of 3GPP operators declines which reflects the possible evolution path towards fragmented access in the local area access provisioning

39. Case 1 Results (3) © 2009 TKK & NOKIA 39 Better device access selection capability leads to less unserved demand and hence less pressure for flexibility spectrum regulation

40. Case 2: High AP self optimizing capability and high willingness of operators to subsidize femtocells © 2009 TKK & NOKIA 40

41. Case 2 Results © 2009 TKK & NOKIA 41 A better chance for femtocell technology to get off the ground  The market share of 3GPP operators grows which reflects the possible evolution path towards integrated access in the local area access provisioning

42. © 2009 TKK & NOKIA 42 Discussion The simulated evolution paths led to a rather fragmented indoor access provisioning scenario The large installed base of WiFi (T31: Num of non-3GPP LA APs) seems to dominate (loop B_non_GPP catered to most of the unserved demand) Only heavy subsidization and high AP self-optimizing ability (case 2) led to a significant market share for femtocells within the study period Femtocells were however on a positive growth path (e.g. in the base case the number of femtocells grew 500 % during the study period) System dynamic modeling is an iterative process and the results presented here should serve as the first iteration round are a good basis for continuing the iterative SD modeling process One should focus on the usefulness of a model rather than on validation and verification Modeler’s skills limited The first time the concept of system dynamics is being applied in the topic area

43. © 2009 TKK & NOKIA 43 Discussion (2) Recommendations for further work Quantitative modeling could be improved Market share currently defined based on data volumes → WiFi dominates → should market share be defined differently (e.g. based on access points)? The traffic model is very simple at the moment (based on static active usage times) The spectrum and technology domain needs more work Capacity of a technology should be based on spectral efficiency (i.e. Mbps/MHz) and the available blocks of frequencies Further calibration of the model based on more detailed data Modeling the y-axis (industry structure) In the current model only operator femtocell subsidization could be seen as industry structure related modeling Willingness to pay, cost and revenue of transmission for each technology Actors (end-users, 3GPP/local operators)

44. References T. Smura and A. Sorri, Future Scenarios for Local Area Access: Industry Structure and Access Fragmentation, in Proceedings of the Eighth International Conference on Mobile Business (ICMB 2009), Dalian, China, June 27-28, 2009, 2009 T. Smura, Access alternatives to mobile services and content: analysis of handset-based smartphone usage data, in ITS 17th Biennial Conference, Montreal, Canada, June 24-27, 2008, 2008 Sterman, J. D., 2000. Business Dynamics: Systems Thinking and Modeling for a Complex World. The McGraw-Hill Companies, Inc. © 2009 MOMI & IMCOS 44

45. Back-up Slides Figure 5.1 Market share of the three technologies U21: represents the ratio of Wi-Fi traffic to the sum of Wi-Fi and femtocell traffic. U22: represents the ratio of data traffic from Wi-Fi and femtocell access points to all indoor traffic. U23: represents the ratio of data traffic from femtocell and 3G/4G base stations to all indoor traffic. U23 can be seen as the combination of U2 and U4