Mobile device report. June Android Samsung. Android. other iphone. Android HTC Android. Dongle/ datacard. BlackBerry. Tablet.

Transcription

1 Android Samsung Android LG Android HTC Android other iphone Dongle/ datacard BlackBerry Tablet Mobile Wi-Fi ndows Phone Mobile device report June 215

2 Contents About this report... 3 Summary... 4 Key findings... 5 Network impact of mobile devices... 6 Network impact rankings... 7 Digging deeper: Inside the Android network impact... 9 Digging deeper: How network impact varies across individual networks... 1 A closer look at devices Radio inefficiency scores Individual device network cost rankings Digging deeper: Inside the Android cost bubble Digging deeper: Regional variations The LTE factor LTE versus 3G: Network impact scores LTE versus 3G: Device costs Signaling analysis: How Androids and iphones are different Top signaling applications of Androids and iphones Application signaling cost analysis for Androids and iphones Digging deeper: Google s power to impact network signaling Conclusion Motive mobile device report

3 About this report Mobile data growth continues at an incredible rate as mobile devices have evolved. No longer just tools for personal communication, they have become highperforming, multimedia platforms that enable consumers to stream high definition (HD) video, surf the web with high performance, engage in social media, participate in online gaming and do banking securely, to name just a few capabilities. The total number of active wireless connected devices is expected to exceed 4.9 billion in 22, up from 13 billion in With this projected growth in mind, this report examines how each mobile device category impacts the network it connects to. Table 1 provides a glossary of the device categories considered in this report. The findings in this report are derived from mobilenetwork and device analytics provided by the Motive Wireless Network Guardian (WNG) from Nokia. Motive WNG gives us a unique vantage point for measuring how mobile data traffic is used in live, commercial mobile data networks, because it sees all traffic used by cellular mobile devices, irrespective of application, device capability or corresponding traffic endpoint. This comprehensive view contrasts with similar industry device reports that are based on surveys, sales reports or traffic measurements at selected web server sites. All analytics from this report were taken in March 215 and are based on data from live 3G and LTE networks. The 3G analytics are drawn from more than 3 million subscribers, who generate over 1 petabyte of mobile data daily on 3G networks around the world. All results are aggregated and anonymized, and they are not representative of any specific network. Instead, they represent a composite, aggregated view of a single global network, which will be referred to as the global composite 3G network. This network will be the prime basis of study in this report. Table 1. Mobile device categories Device category Android OS smartphone (Android) ios smartphone (iphone) Tablet Mobile Wi-Fi Dongle/ Feature phone BlackBerry OS smartphone (BlackBerry) Windows Phone OS smartphone (Windows Phone) Machine-tomachine (M2M) Symbian OS smartphone (Symbian) Other Description Google Android-based smartphones across all Android OS versions and manufacturers Apple ios-based smartphones across all ios versions Tablet-sized (>6.9 in.), cellular-capable mobile devices across all OS vendors and manufacturers Cellular-capable wireless routers that act as a Wi-Fi hotspot for Wi-Fi aggregation All dongles and datacards that attach to a computer, TV or other electronic device to offer cellular access A general class of phones with limited capabilities, when compared to modern smartphones. Feature phones typically provide voice calling and text messaging functionality, as well as basic multimedia and Internet capabilities All BlackBerry phones running BlackBerry OS Windows Phone-based smartphones across all Windows Phone-based OS versions and manufacturers M2M-based mobile devices that are not associated with a specific consumer and geared toward commercial use Symbian-based smartphones across all Symbian-based OS versions and manufacturers An aggregate of devices that are not called out specifically in certain charts. This category includes Symbian, Windows Phone, laptops and PCs with embedded SIMs, and other less statistically significant smartphones 1 Source: 3 Motive mobile device report

4 Summary The Nokia Mobile Device Report examines the impact of mobile devices on service provider networks, in terms of data usage and signaling activity. Together, these two aspects provide the key to understanding the device s overall behavior and impact on the network, as well as the device s individual network cost. Data usage represents the actual amount of data packets delivered downstream and upstream to and from the mobile device as identified by Motive WNG. This consumption drives the service provider s bandwidth-related capital expenditures and the consumer s data usage fees. Signaling activity measures the network-to-device bidirectional exchanges that occur to set up a radio connection to a mobile device for data use. Signaling uses spectral, hardware and processing resources in service providers networks, and it is a significant cause of battery depletion on the mobile device. This report provides an aggregated view of each device category s overall network impact, in terms of data usage, signaling activity and subscriber share (that is, device popularity). Then it looks more closely at each device s individual data usage and signaling activity, which is also defined as the device s network cost. The influence of LTE on mobile devices is then examined, and the report concludes with analysis of the top smartphones signaling activity. The findings benefit three distinct, yet interconnected stakeholders: mobile service providers, mobile device owners and mobile device manufacturers. Mobile service providers gain a better understanding of the impact that each mobile device category has on their network and how they consume data delivery and signaling resources from their network infrastructure. For example, they get answers to the following questions: Which devices consume the most signaling resources? Which use the largest amounts of data? What is the most signaling-efficient device in the market? How does LTE impact the behavior of devices in their network? The answers and insights can help service providers find ways to maximize network efficiency, minimize network cost and increase subscriber satisfaction. In other words, they can optimize their networks to accommodate crucial device characteristics. Mobile device owners can see how their specific devices behave in the network. And this awareness may encourage changes in their own behavior for example, to minimize signaling to preserve their battery life. Or they may become more conscious of the bandwidth they use to lower their data costs. Furthermore, this new understanding may influence device selection, because certain device characteristics may be better suited to specific uses. Mobile device manufacturers will learn the impact that their devices have on the network, and they can compare their efficiency to other devices in the study. Although device behavior is due to many things, including consumer behavior and application use, inherent device design is also a factor. New insights may help these manufacturers optimize their designs, increasing efficiency in the network and make them more attractive to service providers for promotion. More efficient designs will also be more attractive to users, because their usage costs will be reduced, and their device s battery life may be extended. 4 Motive mobile device report

5 Key findings Most popular devices Androids and iphones dominate the global composite 3G network with a combined subscriber share of 86.2 percent of the total device population. Androids are the most popular with almost a 5 percent share of all devices, and iphones are second with 36.8 percent. Other mobile device categories are not even close, with the next highest being M2M at 3.3 percent. Devices with highest network impact Android and iphone device categories dominate the global composite 3G network with a combined data-usage share of more than 8 percent of total daily usage. They also represent an almost 9 percent share of the network s total daily average signaling activity. In large part, their dominance is due to the massive popularity of these devices. Androids have a larger network impact than iphones. Their share of signaling is more than 3 percent higher, and their share of data usage is nearly 15 percent higher. Variance across networks When considering each customer network independently, data usage and signaling activity vary significantly for Android and iphone device categories. These variations are primarily driven by the differences in their popularity among provider networks. For Androids, the subscriber share ranges from 3 percent to over 7 percent. For iphones, the range is from 9 percent to over 5 percent. Devices network costs Each device s network cost is measured with the daily average user traffic. Specifically, it is measured as the daily average data usage and the daily average signaling activity. By comparing network costs across Androids and iphones, we found that Androids use 56 percent more signaling than iphones. However, Androids and iphones use about the same amount of data. In the other device categories, the dongle and datacard and mobile Wi-Fi categories have the highest data usage and signaling activity by far. Specifically, the amount of signaling used by the dongle and datacard category is well over two times the amount used by Android and three times the amount used by iphone. Radio inefficiency scores Radio inefficiency scores can be calculated for each device category as a ratio of the average daily signaling activity to the average daily data usage. It measures how much signaling is used per unit of data or how chatty a device is. It was found that the M2M category is the most radio-inefficient device category, eclipsing all other categories in this measure. The iphone is a more radio-efficient device than the Android, using more than 5 percent less signaling for the same amount of data. Overall device cost rankings Mobile Wi-Fi and the dongle and datacard categories have the highest cost ranking, followed by Androids, tablets, and Windows Phones. The iphone category is in the bottom half of the cost ranking, placing sixth. M2M and feature phones are ranked the lowest. Within the Android category, the HTC Android is more costly, in terms of data usage and signaling activity, than Samsung and LG devices, with LG being the lightest of all. iphones cost the network less than any of the top three Android brands. Regional variations The study revealed significant trends across major regions of the world. Androids are the most popular device in all regions of this study. In North America, Android is still most popular, but iphone is almost as popular. African users use the most data across all categories, with the exception of the dongle and datacard. The dongle and datacard and mobile Wi-Fi categories rank highest in data usage, with the biggest users in the Middle East. Average daily data usage of their dongle and datacard users is almost 55 MB. In North America, the dongle and datacard category shows, by far, the most signaling activity of any device category and region. Impact of LTE networks on top devices When comparing device behavior in LTE networks to our findings for 3G networks, Androids have a 4 percent lower share of data usage, but they gain a 5 percent in share of signaling. iphones gain a significant 11 percent share of data usage and a 3 percent share of signaling. Androids gain a 1 percent share of subscribers, and iphones gain a 4 percent share of subscribers. In LTE networks, iphones have a higher share of data usage than Androids, and they are tied with Androids as the category with the highest overall network impact. Impact of LTE networks on other devices In LTE networks, the impact of the dongle and datacard category is significantly lower. Its share of data usage falls below 1 percent, and its share of signaling drops below.5 percent. This can be explained by a significant drop in its share of subscribers. The tablet category shows an increase in popularity in LTE networks, and its subscriber share almost doubles. Despite this popularity, it shows a decrease in its share of data usage, while its share of signaling activity remains about the same. The BlackBerry and M2M categories are less popular in LTE, with M2M almost disappearing. 5 Motive mobile device report

6 How LTE changes device costs There is a massive increase in data usage for devices on LTE networks, compared with devices on 3G networks. On LTE, a device s average daily data usage is almost four times greater than its 3G counterpart. Signaling activity also increases, but not as much as data usage. For Androids and iphones, data usage increases 3.5 times and 4.5 times, respectively, while signaling activity increases by 2.3 times and 2.1 times. Top signaling applications For Android-based smartphones, Facebook Messenger has the highest share at 17 percent, followed by Google Cloud Messaging (GCM) at 13 percent, Google at 12 percent, HTTPS at 11 percent, Facebook at 1 percent. For iphones, Apple Push Notification Service (APNS) has the highest share at 38 percent, followed by HTTPS at 12 percent, Facebook Messenger at 9 percent, Apple at 6 percent, and Facebook at 6 percent. Application signaling costs Applications running on Androids exhibit a larger signaling cost than the same applications running on iphones. Our data suggests that this is partially due to the effective and broad use of the APNS for most iphone applications. Network impact of mobile devices This section examines the overall impact of each major device category on the network. Data usage is measured by percent share of total average daily data usage, and signaling activity is measured by percent share of the total average daily connection requests. The popularity of each device category is also discussed. Figure 1 shows these three factors across all device categories. The device popularity or subscriber share bar in Figure 1 shows the dominance of Androids and iphones within the global composite 3G network. Combined, these devices make up over 86 percent of the total distribution of devices. This finding is consistent with those of other industry reports. 2,3 When looking at data usage, Androids represent an almost 5 percent share of total network data usage. Combined with iphones, they account for over 8 percent share of total network data usage. Figure 1. Network impact of devices in the global composite 3G network Percent Android iphone M2M Feature phone Dongle/ Tablet BlackBerry Mobile Wi-Fi Other Device Device popularity Percent share of data usage Percent share of signaling activity 2 Source: 3 Source: 6 Motive mobile device report

7 Androids and iphones also dominate when looking at signaling activity, with Androids representing an incredible 59.7 percent of signaling. Combined with iphones, they account for almost 9 percent of total signaling activity. These extremely high percentages of total data usage and signaling activity no doubt correlate with the popularity of these devices. Androids have a larger impact on the global composite network than iphones. Their share of signaling activity is 59.7 percent, compared to 28 percent for iphones, and their share of data usage is 47.9 percent, with 34 percent for iphones. These differences represent a 31.7 percent higher share of signaling activity and a 13.9 percent higher share of data usage for the Android category over the iphone category. Figure 1 also shows that, despite only a 2 percent subscriber share, the dongle and datacard category has a 9.4 percent share of data usage. This may be because these devices are typically attached to PCs or laptops, which have larger screens and are less mobile than smartphones. As a result, these devices tend to consume proportionally larger amounts of data than other categories by streaming video, playing online video games, downloading and uploading high-resolution pictures, and so forth. The mobile Wi-Fi category shows a similar trend. With only.5 percent of subscriber share, this category still manages to consume a 4.1 percent share of data usage the largest ratio of data usage to subscriber share across all device categories. To understand this trend, keep in mind that each mobile Wi-Fi device can aggregate many mobile Wi-Fi devices behind it. Thus, it collectively consumes a large amount of data for a relatively small subscriber share. The M2M category represents non-personal mobile devices that are used commercially for monitoring and control purposes. For example, they re often deployed in industrial automation, healthcare imaging, banking and finance, smart homes, logistics, security and more. In Figure 1, this category has a small subscriber share, only 3.3 percent, which tells us that M2M may not yet have penetrated service provider networks in a really significant way. The data also reveals that M2M devices signaling activity is relatively much greater than their data usage. In the global composite 3G network, they consume.2 percent share of data usage and 1.4 percent share of signaling activity. In other words, these devices are signaling a lot more than they are using data. This makes sense, because many M2M applications establish mobile connections frequently, then send very little data. For example, home smart meters send automated updates several times a day, generating multiple signaling messages to establish network connectivity, with very little data to send each time. Network impact rankings To provide another perspective on each device category s impact on the global composite network, we have established an overall network impact score between 1 and 1 for each device category. This score is calculated by first computing a network impact score between 1 and 1 for both data usage and for signaling activity. The overall network impact score is then an average of both of those individual scores. Device categories are then ranked. The device category with the highest score is ranked Number 1, which means it has the highest network impact. Table 2 shows these rankings, and as expected, Androids and iphones are at the top. To offer a deeper, more visual understanding of the network impact rankings, Figure 2 plots the data usage score and the signaling activity score for each device. The size of the bubble on the chart reflects the device popularity of that category. Table 2. Network impact rankings Rank Device category Overall score 1 Android 1 2 iphone 9 3 Dongle/ 8 4 BlackBerry 6 5 Tablet Mobile Wi-Fi M2M Windows Phone 3 9 Feature phone Symbian 1 7 Motive mobile device report

8 Figure 2. Network impact scores plotted for the global composite 3G network More popular Less popular Android Rank by increasing signaling activity Feature phone Symbian BlackBerry M2M Windows Phone Tablet Mobile Wi-Fi Dongle/ iphone With this plotting it is obvious that Android and iphone are the dominant categories Rank by increasing data usage Figure 2 offers service providers and device manufacturers a quick snapshot of the impact that various devices have on the network, while also showing which devices are most popular. With this plotting, it is obvious that Android and iphone are the dominant categories. But it also makes clear that dongles and datacards have a relatively significant impact on the network, even though they re not as popular. The mobile Wi-Fi, tablet and BlackBerry categories come next in terms of network impact, although their bubbles in Figure 2 are quite small, indicating a small subscriber share value. This overall network impact study is a good starting point for understanding the impact of various devices. Later, this report establishes individual network costs for each device category, independent of the influence of popularity. 8 Motive mobile device report

9 Digging deeper: Inside the Android network impact Within the Android category, several device manufacturers implement the Android OS. This section of the report examines the major manufacturers and provides their individual network impact scores. Specifically, they include Samsung, HTC, and LG, along with a category called other that includes approximately 6 more Android-based device manufacturers. Table 3 adds these new categories to the network impact scores and ranking analysis. In this new ranking, the Android Samsung category is the most popular manufacturer of Android, with almost 3 percent share of all 3G mobile devices. This is not surprising as Samsung is a marketing juggernaut, dominating social video marketing, and was ranked as one of the top two shared brands in and In terms of data usage, the iphone category has the greatest impact which contributes to its being tied with Android Samsung as the device with the greatest network impact overall. However, Android Samsung remains most impactful with respect to signaling activity, despite being 8.3 percent less popular than the iphone. Android HTC and LG are the next most popular Android device manufacturers, with subscriber shares of 4.6 percent and 5.5 percent, respectively. HTC and LG rank fourth and sixth in their network impact, respectively, and Android other ranks third in network impact, with a 1.3 percent subscriber share. Table 3. Network impact rankings, including Android manufacturers Rank Device category Overall score (1-1) 1 iphone Android Samsung Android other Android HTC Dongle/ Android LG BlackBerry Tablet Mobile Wi-Fi M2M Windows Phone 12 Feature phone Symbian.77 Figure 3 provides a scatter diagram showing the Android bubble of Figure 2 broken into its representative manufacturers. It makes clear that the Android Samsung category is the most dominant Android category. It is also tied for greatest overall network impact with the iphone category and has the greatest signaling impact. The other Android categories all remain in the upper right quadrant of the graph, representing their high impact in both data usage and signaling activity. Figure 3. Network impact scores plotted, including more specific Android categories More popular Less popular Android Samsung 1 Rank by signaling activity Feature Phone M2M BlackBerry Tablet Mobile Wi-Fi Android LG Android HTC Dongle/ datacard Android other iphone 2 1 Windows Phone Symbian Rank by increasing data usage 4 Source: blog/6464-how-samsung-owns-socialvideo-with-youtube-and-vine/ 5 Source: news/214/12/3/activia-samsungand-nike-most-shared-social-videobrands Motive mobile device report

10 Digging deeper: How network impact varies across individual networks The previous section analyzed aggregated data from all the networks in this study. This approach provides a macro view of the devices and their overall behavior. However, a device s impact within each provider network can vary significantly, because they are influenced by a variety of factors, ranging from service provider s device promotion strategy and data plans to cultural differences that influence usage patterns and application use. Figure 4. Range of data usage across all service provider networks Percent share of data usage Android iphone Mobile Wi-Fi Dongle/ Tablet BlackBerry M2M Feature phone Other -5 th percentile 5 th - 1 th percentile Mean Median 1 Motive mobile device report

11 This section examines the variance across networks by showing data usage for each device category within each mobile network studied. In this specific analysis, data from each network has equal weight, so that exceptionally large networks do not dominate smaller networks when the results are analyzed. With this information, a range of percentage share values, from high to low, can be established, along with the mean and the median. (The mean is the average of the shares of data usage across each network. The median indicates the exact middle across all of the share values.) The vertical bars in Figure 4 show the varying percentage share of data usage across each network for each device category. These findings indicate that the Android and iphone device categories collectively dominate the networks where they re deployed, as shown by the mean values of percent and percent, respectively. However, these data usage figures range widely across individual networks for both these device categories. For Android, the percentage share range extends from percent to percent. For iphones, it ranges from 7.78 percent to 5.81 percent. these observations.) The results are very similar when considering signaling activity. In fact, the network impact for both data usage and signaling activity correlates strongly with device popularity across each of the networks. For Androids-based devices, the range of popularity varies from percent to percent. For iphones, the range varies from 9.15 percent to 51.1 percent. In general, networks showing larger ranges of device popularity generally had larger ranges of data usage and signaling activity for that device. Likewise, when networks have smaller ranges of device popularity, they generally had smaller ranges of data usage and signaling activity for that device. Comparing the mean with the median reveals more about the distribution of percentage share values across the networks. For Android, the median of percent is much lower than the mean. This difference suggests that there are more networks that have a percentage share value below the mean than above the mean. This shows that there is a small number of networks that have very high share values that pull the mean value well above the median. The network at the top end of the range, with percent share, is an example of one. For the iphone, the median of percent is closer to the mean, indicating that the percentage shares are more evenly distributed across the range. Figure 4 also makes clear how significant the mobile Wi-Fi, dongle and datacard, and even tablet device categories can be in some networks in terms of data usage. In networks where these devices have the largest impact, the highest percent share is percent for the mobile Wi-Fi device category, along with a whopping percent for the dongle and datacard device category. BlackBerrys, M2M and feature phones make up a very small share of data usage across all networks. (This report s section on regional variations presents some reasons for 11 Motive mobile device report

12 A closer look at devices The analysis presented in the previous section offers a great way to understand the impact that each device category has on the network. However, these results are heavily weighted by the impact of device popularity. This limits the analysis to a more general understanding of the characteristics and impact that devices have as an aggregated group. A specific device may initially appear quite innocuous when it is unpopular and not widely deployed, but what happens when it is actively promoted and its popularity skyrockets? Some devices may appear quite costly, but they are really quite efficient in terms of network cost, on a per-device basis. To really understand how each device behaves in the network, it is important to consider each device separately and determine its individual network cost. This cost is defined and measured across two dimensions, the average daily data usage and the average daily signaling activity. With this type of information, service providers can predict how shifts in popularity and usage trends of a specific device will impact their networks. Figure 5 reveals the individual network costs of each device category. Figure 5 shows that Androids use 56 percent more signaling on average, on a daily basis, than iphones do. (In a later section, we will examine some reasons for this difference.) Both categories consume about the same amount of data. The dongle and datacard and mobile Wi-Fi categories use by far the most data and generate the most signaling activity. In fact, the amount of per-device signaling activity exhibited by the dongle and datacard category is well over two times and three times the amounts for Android and iphone devices, respectively. M2M, BlackBerrys and feature phones exhibit very little data usage with respect to their signaling activity. That s because unlike smartphones these devices are not used as data-intensive multimedia platforms. Figure 5. Individual network costs across all device categories MB or setups Android iphone BlackBerry Dongle/ M2M 52 Feature phone Mobile Wi-Fi Symbian Tablet Windows Phone Data usage cost (MB) Signaling activity cost (setups) Device 12 Motive mobile device report

13 Radio inefficiency scores The network costs just described are used to establish radio inefficiency scores for each device. The amount of daily signaling activity is simply divided by the amount of daily data usage. This score measures the amount of signaling per unit of data usage and demonstrates how chatty certain devices are on the network. Figure 6 shows these inefficiency scores across each device category. The M2M category immediately stands out in Figure 6, because its radio inefficiency score of 33 makes it, by far, the most inefficient or chatty. This may be explained by the nature of certain M2M services. In some cases, these services establish connections while having relatively little data to transmit. For example, a home monitoring appliance may send an update many times per day to a centralized server, transmitting small bits of information on home temperature, natural gas use and so forth. BlackBerrys and feature phones are also relatively inefficient, with scores of 2 and 14, respectively. These devices do not signal more than other categories. Their high scores reflect the fact that they do not use a lot of data in an average day. That is, these devices are not used like the more data-intensive multimedia platforms that Androids and iphones have become. Androids and iphones are relatively efficient with scores of 7 and 5, respectively. These scores also indicate that the iphone is a more radio-efficient device, using over 5 percent less signaling than Androids for the same amount of data usage. The inverse of this score, a device s radio efficiency, is measured by the relative amount of data delivered per unit of signaling. Radio inefficiency and efficiency scores are a quick way to understand what the network impact will be relative to signaling activity when rolling out specific mobile devices in new markets. Figure 6. Radio inefficiency scores across devices Radio inefficiency (setups/mb) Android iphone BlackBerry Dongle/ M2M Device Feature phone Mobile Wi-Fi Symbian Tablet Windows Phone 13 Motive mobile device report

14 Individual device network cost rankings In this section, an individual network cost score from 1 to 1 is established for each device. This score represents the individual cost that the device has on the global composite 3G network, and it reflects both data usage and signaling activity by taking the average of the individual cost scores for these dimensions. Similar to network impact rankings, device categories are ranked from 1 to 1, and the device category with the highest network cost score has the highest rank. Table 4 clearly shows that the mobile Wi-Fi and the dongle and datacard categories are most costly, followed by Androids, tablets and Windows Phones. The iphone category is ranked sixth, in the bottom half of network cost scores. M2M and feature phones exhibit the smallest cost. Figure 7 takes data usage and signaling activity cost scores and plots them on a scatter diagram. This view offers a deeper, more visual understanding of a device s network cost rankings. It also further demonstrates the enormous cost of the dongle and datacard and mobile Wi-Fi categories, compared with other categories. The reason for the extremely high cost for dongles and datacards is twofold. First, these devices are naturally data intensive, because their larger screens promote video use, and their lower propensity for mobility also encourages data usage. Second, in the North American market these devices are used by business road warriors who have been shown to be heavy on signaling. (See this report s section on regional variations for more detail.) Mobile Wi-Fi will naturally consume a large amount of data and generate a lot of signaling activity as it effectively represents many Wi-Fi devices that are aggregated behind it. Table 4. Individual device network cost scores and rankings Rank Device category Overall score (1-1) 1 Mobile Wi-Fi Dongle/ Android Tablet Windows Phone 6. 6 iphone 5. 7 BlackBerry Symbian M2M Feature phone 1.5 These individual costs can help service providers determine the potential impact to the network, when a new device is promoted and expected to increase in popularity. Of course, device costs are determined by many things, including mobile application use, user behavior, individual traffic patterns, and the inherent design of the device and its OS. As a result, mobile device manufacturers do have some degree of control over the individual network cost of their devices, and insights like these may be leveraged to influence their designs. Figure 7. Individual device costs: Data usage and signaling activity 1 Dongle/ Rank by increasing signaling activity M2M Feature phone BlackBerry Symbian Windows Phone iphone Android Tablet Mobile Wi-Fi Rank by increasing data usage 14 Motive mobile device report

15 Digging deeper: Inside the Android cost bubble This section examines the network cost scores of the top device manufacturers within the overall Android category. These subgroups are Android HTC, Android Samsung and Android LG. Table 5 shows the individual device scores and rankings, while Figure 8 plots data usage and signaling activity cost scores. Table 5 shows that, in terms of network cost, the mobile Wi-Fi and dongle and datacards categories are still ranked at the top, while Android HTC remains the third most costly category. That makes Android HTC the most costly Android-based device, with Android LG being the least costly. Figure 8. Android in detail: Individual cost scores plotted Rank by increasing signaling activity M2M Feature phone BlackBerry Symbian Android LG Windows Phone Android Samsung iphone Android HTC Tablet Dongle/ Mobile Wi-Fi Table 5. Android in detail: Individual network cost scores and ranking Rank Device category Overall score (1-1) 1 Mobile Wi-Fi Dongle/ Android HTC Tablet Windows Phone 6 Android Samsung Android LG iphone BlackBerry Symbian Feature phone M2M Rank by increasing data usage 15 Motive mobile device report

16 Digging deeper: Regional variations This section organizes our analysis across major regions of the world by creating separate regional composite 3G networks. The regions included in this study include Africa and North America, along with a grouping that represents the other major regions of our study including Asia, the Middle East and Europe (AMEE). Figure 9 shows the device popularity across these regions. As expected, the Android and iphone categories are the most popular across all regions. Androids are by far the most popular category in Africa and in AMEE with percentage shares of 75 percent and 59 percent, respectively. In North America, the Android is still the most popular category but only slightly more than the iphone. It has a 48 percent subscriber share compared with iphone s 42 percent. The other device categories are not very popular, with the exception of the dongle and datacard category in the African region. Figure 9. Device popularity across major regions Subscriber share (popularity) Tablet Mobile Wi-Fi M2M Dongle/ BlackBerry iphone Android Device AMEE NA Africa 16 Motive mobile device report

17 Some of the regional variations in popularity between iphones and Androids can be attributed to how each device is marketed and promoted. In North America, smartphones are usually sold with a yearly data plan attached to the device. In addition, the iphone has a very small range of phone models and cost points, and Apple typically targets users who are willing to pay more for a phone that has more features and capabilities and who are also willing to spend more on applications at the istore. This approach, embraced by Apple and their iphone marketing strategy, is well received in North America, as reflected by the iphone s popularity in this region. In other regions of the world, the concept of pay as you go with prepaid data is more popular, because flexibility and cost effectiveness are paramount. Androids have embraced this approach and offer a very large range of devices from different manufacturers with a broad spectrum of capabilities and cost points. This may help explain why Androids are significantly more popular than iphones in regions outside of North America. Figure 5 provides the daily averages for data usage and signaling activity, calculated across the entire global composite 3G network. In this section, the same calculation is applied to each major region of our study: Africa, North America and AMEE. Figures 1 shows the results of this analysis. Figure 1. Daily average data usage (left) and signaling activity (right) across regions Tablet Mobile Wi-Fi M2M Dongle/ BlackBerry iphone Android Tablet Mobile Wi-Fi M2M Dongle/ BlackBerry iphone Android AMEE NA Africa 17 Motive mobile device report

18 One point that immediately stands out is how much more data users in Africa and AMEE use each day than users in North America. This can be explained by examining some of the cultural usage patterns within these regions. In AMEE and, especially, within the Middle East, users consume a very large amount of video. Delving deeper into this trend, we found that, within Middle Eastern networks, the top applications all involved video use, such as YouTube, Apple QuickTime and video downloads. The dongle and datacard category was the top device used for video, resulting in an average daily data usage of 541 MB. African users consume the most data across all device categories, with the exception of dongle and datacard. Video viewing still contributes to this consumption more than all other forms of data. In addition, this heavy use of mobile data supports descriptions of Africa as the mobile continent, 6 where many people first connect to the Internet through mobile devices. Lack of fixed infrastructure, unreliable electricity, and increasingly cheaper smartphones are likely reasons that mobile data usage is much higher in certain parts of Africa and preferred over wireline connections. 7 Daily signaling activity is more evenly distributed across regions and device categories than daily data usage. However, in North America, the dongle and datacard category shows the most signaling activity, far more than any other device category and region. Closer examination of the data points to the large number of road warriors in the North American market who regularly use their laptops on the go for business. The applications they use are very signaling-intensive, like chatty mobile VPNs that typically send a constant keep alive signaling heartbeat, VoIP, and messaging applications like Google Talk, as well as lots of web surfing that generates significant HTTPS and HTTP traffic. The LTE factor Up to this point, the findings we ve discussed have been restricted to 3G technology, which is deployed by service providers worldwide in almost all countries. LTE, however, is not widespread enough to enable comparisons across all the regions within this study. Nevertheless, it is important and interesting to understand how different technologies can impact the behavior of mobile devices. So in this section, we compare our baseline 3G analysis with an LTE network consisting of a smaller group of LTE networks. The study uses actual data from more than 24 million subscribers, generating over 3 petabytes of mobile data daily on live LTE networks across North America, the Middle East and Asia. All results are aggregated and anonymized and are not representative of any specific network. Instead, they represent a composite, aggregated view of a single global LTE network, which we refer to as the global composite LTE network. 6 Source: 7 Source: 18 Motive mobile device report

19 LTE versus 3G: Network impact scores Before any comparisons are made with our 3G results, Table 6 reveals the network impact scores (from 1 to 8) and rankings for devices in the global composite LTE network. As in Table 2, these scores reflect both data usage and signaling activity. The score is calculated by first computing a network impact score between 1 and 8 for both data usage and for signaling activity. The overall network impact score is then an average of both of those individual scores. As Table 6 shows, there are no Symbian or feature phones in this network. Figure 11 plots data usage and signaling activity for each device, with the size of the plotting point reflecting the popularity of the device category. Table 6 and Figure 11 show that Androids and iphones are tied, when measuring which devices have the highest overall impact on the global composite LTE network. The Android category has the highest signaling impact, and the iphone category has the highest data usage impact. The impact of M2M and dongle and datacard categories is noticeably smaller in LTE than on the 3G network. Table 6. Network impact rankings for the global composite LTE network Rank Device category Overall score (1-8) Subscriber share 1 Android % 2 iphone % 3 Tablet % 4 Mobile Wi-Fi % 5 Windows Phone 6 Dongle/ 3..38% 3..15% 7 BlackBerry 3..8% 8 M2M 1..7% Figure 11. Network impact scores plotted for the global composite LTE network More popular Less popular 1 9 Android Rank by increasing signaling activity BlackBerry Windows Phone Dongle/ Tablet Mobile Wi-Fi iphone 1 M2M Rank by increasing data usage 19 Motive mobile device report

20 Table 7. Network impact across 3G and LTE networks a comparison Device category Data usage share 3G Data usage share LTE Signaling activity share 3G Signaling activity share LTE Subscriber share 3G Subscriber share LTE Android 5.% 46.12% 6.4% 64.92% 51.33% 52.31% iphone 37.9% 47.98% 29.27% 31.76% 38.89% 42.69% Tablet 1.91%.86% 1.44% 1.46% 1.86% 3.9% Mobile Wi-Fi 3.67% 4.4%.94%.7%.49%.52% Windows Phone.29%.2%.4%.31%.39%.38% Dongle/ 6.31%.64% 4.51%.23% 1.77%.15% BlackBerry.46%.15% 1.52%.62% 1.74%.8% Using this network impact baseline, our study made a direct comparison of device network impact across 3G and LTE networks. Table 7 shows the percentage share values for both 3G and LTE networks, providing side-byside comparisons of data usage, signaling activity and device popularity. Please note that because there are no Symbians, feature phones, and other devices studied within the global composite LTE network, the percentage shares calculated for 3G in Table 7 are calculated across a smaller number of devices and thus will differ slightly from those presented earlier. As shown in Table 7, Androids lost 4 percent in its share of data usage but gained 5 percent in its share of signaling activity. iphones gained a significant 11 percent in its share of data usage and also gained 3 percent in its share of signaling activity. Androids gained 1 percent in its share of subscribers, and iphones gained 4 percent in its share of subscribers. In general, we found no dramatic changes in percentage share values, except that iphones increased their network impact, driven primarily by their larger data usage. Among other device categories, the dongle and datacard category declined significantly in the global composite LTE network. Its share of data usage decreased below 1 percent, and its share of signaling dropped below.5 percent. This decrease can be explained by the sizable drop in its share of subscribers. This drop probably results from a slower transition to LTE, and the fact that many of these devices are provided by users employers who have a mandate to maximize the life of the device. The BlackBerry and M2M categories are also less popular in LTE, with M2M almost disappearing, as it drops from 3.4 percent to.7 percent. The reduced popularity of M2M devices is easy to explain as it is about economics and coverage. M2M applications and services usually don t need a lot of bandwidth and performance, but they certainly need coverage. Economically, 3G networks are best suited for both cost and coverage for these types of services. The tablet category actually increased in popularity in LTE networks, with subscriber share growing from 1.8 percent to 3.9 percent a 66 percent increase. Despite this increase in popularity, however, the category shows a decrease in its share of data usage, while its share of signaling activity remains about the same. These findings reflect that iphones claimed a greater share of data usage from the tablet category. The mobile Wi-Fi and Windows Phone categories both remain about the same from 3G to LTE. The tablet category actually increased in popularity in LTE networks, with subscriber share growing from 1.8 percent to 3.9 percent a 66 percent increase. 2 Motive mobile device report

21 Figure 12. Network impact comparison percentage share ratios of 3G to LTE Ratio of LTE to 3G network impact Android iphone BlackBerry Dongle/ Device M2M Mobile Wi-Fi Tablet Windows Phone LTE/3G share of data usage LTE/3G share of signaling activity LTE/3G share of subscribers Figure 12 shows the ratio of percentage shares between LTE and 3G for each device category. Any value above 1 represents an increase in network impact, while any value below 1 represents a decrease in network impact. Clearly, the dongle and datacard, M2M, and BlackBerry categories all decreased their network impact dramatically across all dimensions. However, the Android and iphone categories remain relatively stable from 3G to LTE, with respect to their network impact, except for the iphone s increased share of data usage. The most noticeable item on this chart may be the increase in popularity of the tablet category from 3G to LTE. But even with this increase, its share of data usage has decreased. 21 Motive mobile device report

22 LTE versus 3G: Device costs In January 214, a Nokia blog 8 projected that the growth of data usage for devices on LTE networks would be three times that of devices operating on 3G networks. This section of our report examines that projection. Figure 12 already compared device impact on 3G and LTE networks, finding that, in general, there were no major shifts in network impact across the top device categories. However, the network impact of BlackBerrys, dongle and datacards, and M2M devices was significantly lower on LTE. In this section, a similar comparison is made, but this time comparing the devices network cost. Figure 5 shows device network costs established from the global composite 3G network. Figure 13 now compares those costs with device network costs from the global composite LTE network. The data in Figure 13 was calculated for each device category by establishing the ratio of its average daily data usage and its average daily signaling activity on the global composite LTE network to the global composite 3G network. Values greater than one represent a cost increase on the LTE network, and values smaller than one represent a decrease. Figure 13 shows a massive increase on the global composite LTE network for both data usage and signaling activity across almost all categories. An average Android-based device, for example, will use 3.5 times more data and generate 2.3 times more signaling activity when on an LTE network, rather than a 3G network. An average iphone will use 4.5 times more data and generate 2.1 times more signaling when on an LTE network. Figure 13. Device costs across 3G and LTE a comparison Ratio of LTE to 3G costs Android iphone BlackBerry Dongle/ M2M Mobile Wi-Fi Tablet Windows Phone LTE/3G data usage cost LTE/3G signaling activity cost Device 8 Source: 22 Motive mobile device report

23 Dongle and datacard, M2M, and mobile Wi-Fi devices will also use 4.8 times, 5.8 times, and 4. times more data, respectively, on LTE networks. The only decrease on LTE networks is for the M2M category, where there is a 4 percent reduction in signaling activity. In general, all categories, except tablets, use more data, with increases ranging from 2.7 times to 5.8 times more on LTE networks. All categories except M2M use more signaling, with increases ranging from 1.3 times to 2.3 times more. The primary driver for this massive increase in data usage on LTE networks is the performance capabilities of LTE, which promote greater use of video. 3G and LTE performance were compared in the same blog from Analytics Beat, which found that LTE networks deliver more than four times the speed of 3G, on average (that is, 3.7 to 6 times faster, depending on the network). Because of this performance improvement, LTE networks can deliver data-intensive experiences, such as video streaming, on mobile devices. This same blog projected that LTE users would consume three times more data than 3G users by the end of 214. Another blog from Analytics Beat 9 reported finding that on LTE networks, video use represents the highest share of traffic of all applications categories and generates over a third of all daily traffic usage. The aforementioned blog projection seems to be validated by the analysis described in this section of our report. Specifically, if the data from Figure 13 is aggregated across all device categories for March 215 (the month the data was based on), and the ratio of the average daily data usage for devices on LTE networks is compared with that of 3G networks, the result is 3.74 times more data usage per user in LTE. When this result is compared with the blog s projection (3 times more data usage), it is clear that the growth of data usage for devices in LTE networks is even larger than projected. 9 Source: 23 Motive mobile device report

24 Signaling analysis: How Androids and iphones are different This section is dedicated to the analysis of signaling activity and the top applications that contribute to it on Androids and iphones. Because Androids and iphones behave very differently with regard to signaling, we are focusing our analysis on this topic, which is more revealing than examining data usage. For example, the top ten applications by data usage for Androids and iphones are almost the same on every network, indicating that users of both device categories have very similar application choice preferences. These top ten applications include YouTube, HTTPS, Facebook, Google, and several video applications from Facebook and Instagram. In addition, the average daily data usage per user for Androids and iphones is very similar, as observed previously in Figure 5. On the other hand, the top ten applications by signaling activity have only five applications in common across Androids and iphones: WhatsApp, HTTPS, HTTP, Facebook and Facebook Messenger. In addition, the daily signaling activity of Androids is notably higher than on iphones in every single network that was examined. The question we re asking is this: Why is the signaling behavior so different across Androids and iphones? Many factors influence the amount of signaling exhibited by a device, including the nature of the applications used, the networking efficiency of the application client implemented on the device and how the device is configured to interact with the radio network. (For example, when and how does it release radio channels?) It is difficult to pinpoint how all these factors weigh in to make Androidbased devices exhibit higher signaling as configuration and design aspects can vary across the implementation of smartphones. For instance, 3GPP s network-controlled fast dormancy feature was endorsed by Apple in 21 1 and adopted by many other smartphone manufacturers. This feature was designed to reduce the chattiness of smartphones by setting parameters on how, and how often, a smartphone switches between idle and active modes while also preserving device battery life. Although endorsed by Apple, its implementation and configuration can vary across smartphone manufacturers and OS versions thus creating variance on how it behaves in the network with respect to signaling....the top ten applications by data usage for Androids and iphones are almost the same on every network, indicating that users of both device categories have very similar application choice preferences. In this section, we take a closer look at the top signaling applications to reveal important differences in how applications on these device categories interact with the network. Our traffic measurements suggest that the push-notification infrastructure used by the applications on these devices is likely an important contributing factor to the amount of signaling each device generates. 1 Source: 24 Motive mobile device report

25 Top signaling applications of Androids and iphones Figures 14 and 15 show the top ten applications that account for the largest amount of daily signaling seen on iphones and Androids. These are the applications that generate the largest amount of signaling activity in the network over any given day for each respective device category. The Y axis shows the percentage share of signaling activity each application is responsible for, from among the top 9 heavy-signaling applications on that particular smartphone category. Figure 14 shows which applications have the highest percentage share of signaling activity for Android. Facebook Messenger has the highest share at 17 percent, following by Google Cloud Messaging (GCM) at 13 percent, Google at 12 percent, HTTPS at 11 percent, Facebook at 1 percent. Rounding out the top ten is HTTP, WhatsApp, Google Play, Extensible Messaging and Presence Protocol (XMPP), and Viber. XMPP is a protocol that is primarily used by Google Talk and Viber. Figure 14. Applications with highest percentage share of signaling activity for Android 2 Percent share of signaling activity Facebook Messenger GCM Google HTTPS Facebook HTTP WhatsApp Google Play XMPP Viber Figure 15 shows applications with the highest percentage share of signaling activity for iphone. Apple Push Notification Service (APNS) has the highest share at 38 percent, followed by HTTPS at 12 percent, Facebook Messenger at 9 percent, Apple at 6 percent, and Facebook at 6 percent. Rounding out the top ten is Hotmail, HTTP, Apple Maps, Microsoft, and WhatsApp. Figure 15. Applications with highest percentage share of signaling activity for iphone 45 Percent share of signaling activity APNS HTTPS Facebook Messenger Apple Facebook Hotmail HTTP Apple Maps Microsoft WhatsApp It is important to note that these are not all applications that users recognize. Some run in the background to provide supporting services. The two background applications that show up in Figures 14 and 15 are APNS 11 and GCM 12, respectively. These applications provide notification services to applications running on iphones and Android-based smartphones. 11 Source: 12 Source: 25 Motive mobile device report

26 When comparing Figure 14 and Figure 15, it s clear that the distribution of the share of signaling across the top ten applications is very different. For iphones, APNS dominates and accounts for a 38 percent share of daily signaling activity, while the share of signaling activity drops significantly across all other applications. For Androids, Facebook Messenger and GCM are at the top with 17 percent share and 13 percent share, respectively, while the share of signaling activity drops gradually for the other applications. This trend reveals that, on iphones, a good portion of the signaling is due to the delivery of push notifications from APNS. On Androids, the effect of GCM is not as great, and the signaling impact is more evenly spread across a larger set of apps. In fact, GCM on Androids is the second top signaling application, accounting for less than half as much signaling share as APNS does on iphones. Does this mean that Androids handle fewer push notifications than iphones? To answer this, we need to examine how both push notifications mechanism are handled. Apple 13,14 was first to develop a push notification feature into smartphones, recognizing that it was critical for applications to have a reliable, scalable and efficient mechanism for delivering notifications to devices. The core design principle behind Apple s solution is its centralized server, which coordinates the delivery of notifications to applications on a phone. As a result, there s no need for each application to develop and support its own notification mechanism. A large base of iphone applications came to rely on this centralized mechanism. After the Android entered the smartphone landscape, Google developed its own push notification infrastructure, called GCM, which also centralizes how notifications are managed. In addition, a number of third-party notification applications emerged, such as Xtify and Urbanairship. In the usual spirit of openness and flexibility of the Android community, this has led to a fragmented base of developer preferences for how to handle push notification services. The signaling impact of the APNS is quite large because it accounts for the signaling done on behalf of a large number of iphone apps, whereas Google s GCM appears to be serving a smaller set of applications. Xtify, for example, handles a lot of signaling traffic on Androids even though it does not appear in Figure 14. Apple was first to develop a push notification feature into smartphones, recognizing that it was critical for applications to have a reliable, scalable and efficient mechanism for delivering notifications to devices. 13 Source: 14 Source: 26 Motive mobile device report

27 Application signaling cost analysis for Androids and iphones To further contrast Android and iphone signaling behavior, Figure 16 shows the per-application signaling activity cost, measured by the average number of setups per day, for the top signaling applications already identified. For every application that is common across both devices, signaling activity is heavier on Androids than on iphones. For example, WhatsApp has 33 connection setups a day on Androids, compared to 19 on iphones. Because WhatsApp on iphone uses APNS, 15 much of the signaling required for notifications is being accounted for within APNS, thus lowering the overall signaling of the application. This is similar with Facebook Messenger, HTTPS, HTTP, Facebook and other common applications. The net result appears to be that the aggregation of notifications performed by APNS, coupled with well known architectural limitations of APNS (which limit the number of connections that are available for handling notifications), 16 steer iphone application developers to become more network friendly and place a cap on the aggregate signaling load across the registered applications. Conversely, when push notifications are distributed across different notification components, as appears to be the case with Android s approach, multiple applications are likely to compete for network resources and hit the network with more frequent connection setup requests. It remains to be investigated whether or not the responsiveness of applications regarding notifications is compromised on iphones in a way that affects the user s perceived quality of experience. However, from a network perspective, the centralization of push notifications under APNS seems to have a net effect of lowering the overall signaling activity on the iphone while enabling radio efficiencies that would not be viable in a more distributed solution, such as the Android s GCM. Figure 16. Daily per-application signaling cost for Android (left) and iphones (right) Signaling activity (setups) Facebook Messenger WhatsApp Facebook HTTPS GSM Google Viber XMPP HTTP Google Play APNS Facebook Messenger Hotmail HTTPS WhatsApp Facebook Microsoft HTTP Apple Apple Maps 15 Source: 16 Source: 27 Motive mobile device report

28 Digging deeper: Google s power to impact network signaling From January 12 to February 19, 215, a dramatic increase in signaling for the GCM application was observed across many networks. Figure 17 shows a representative signature of this phenomenon. Figure 17. Percentage subscriber and signaling shares for Google Cloud Messaging Percent share % erosion of network signaling capacity 15 January 12 February 4 February 19 Signaling share Subscriber share The bottom line on the chart reflects the percentage share of signaling activity for the GCM application over time. On January 12, GCM experienced a significant increase in signaling, as shown by its increase in signaling share from 17 percent to 2 percent. On February 4, GCM experienced another signaling increase, as its signaling share went from 21 percent to a peak of 23 percent. This increase in signaling resolved itself on February 19, when GCM s signaling share went back down to expected levels. The top line on Figure 17 shows the percentage subscriber share of the GCM application over the same time frame. Clearly, there is no increase in subscriber share during the time the signaling increase occurred. This indicates that the increase in signaling activity for GCM was not due to an increase in active subscribers. Although a rise in signaling share from 17 percent to 23 percent on a single application may appear rather innocuous at first, it does have a significant impact on networks. During this period of signaling increase, an average erosion of 6 percent in overall signaling capacity was experienced across the networks that were analyzed. This is a costly loss that can place a large strain on radio resources, and it can even cause outages in locations that were already operating close to capacity or where there was a dominant proportion of Android users. This signaling increase also impacts users, as individual signaling activity costs increased anywhere from 6 percent to 51 percent, with an average 32 percent increase across all networks. This is important because signaling activity is a significant contributor to battery drain, which is of primary concern for mobile users. The incident shown in this case study highlights the great vulnerability of carrier networks to sudden changes in the signaling behavior of popular applications. A similar incident, featured in a previous blog 17 in Analytics Beat, occurred when Facebook released a chattier version of its popular application. Developers of widely used applications need to be aware of their responsibility to ensure that software updates do not adversely affect how their apps interact with networks. 17 Source: 28 Motive mobile device report

29 Conclusion As mobile devices continue to grow exponentially around the world, they place greater demands on service providers data and signaling infrastructures. A detailed understanding of the behavior and impact of these devices can benefit consumers and service providers alike. Consumers can use device cost and efficiency information to adjust their usage behavior and their application and device choices, so they can optimize their experience, while getting the most from their personal investment. Service providers can benefit in several key ways. For example, they can anticipate the impact of device growth and popularity shifts as consumer trends shift. They will also be in a position to predict the impact of device proliferation while optimally planning network growth and the promotion of new devices. They can also use these insights to engage with device manufacturers, discussing how to optimize the behavior of devices on the network. This study highlights the power of knowledge with respect to the impact of mobile devices on the global composite 3G and LTE networks. But each individual network is truly unique, and device behavior and associated network impact is heavily influenced by market coverage, data plan variety, population demographics and cultural preferences. To fully harness the possibilities offered by insights like those contained within this report, service providers need to conduct their own studies using data derived from their own networks. Service providers can gain more powerful insights about their networks with the Motive Wireless Network Guardian, a network analytics solution that can correlate the six key dimensions of mobile intelligence. Then they can put that information to work across all parts of their organization using Motive Big Network Analytics. As mobile devices continue to grow exponentially around the world, they place greater demands on service providers data and signaling infrastructures. 29 Motive mobile device report

30 Nokia is a registered trademark of Nokia Corporation. Other product and company names mentioned herein may be trademarks or trade names of their respective owners. Product code: PR EN Nokia 216