OpenSignal comments on Draft Annual Plan 2015/16 Consultation Document
OpenSignal1 is creating a comprehensive database of cell phone towers, cell phone signal strength readings, and Wi-‐Fi access points around the world. We create this database to provide insight on connectivity, adopting the philosophy that only a data driven examination of the true performance of these networks can lead to active improvements of electronic communications infrastructure. We achieve this by crowdsourcing data on mobile devices through our Android and iOS apps, which have had over 11 million downloads around the world, and have approximately 50,000 active users in the UK contributing data. We are providing comments on the Ofcom Draft Annual Plan 2015/16 (hereafter referred to as ‘Annual Plan’) because many of the strategic priorities are aligned with OpenSignal’s mission in terms of improving connectivity for mobile users through data analysis, provision of information to allow consumers to make informed choices, and promoting competition to encourage infrastructure improvements to benefit everyone. Based on OpenSignal as the market leader on mobile data collection and analysis with respect to the mobile user experience, we are providing comments relating to how a crowdsourced methodology can contribute to achieving the aims laid out in the Annual Plan. A recent report released in draft form in early 2015 by the CEPT, ECC Report 231 on Mobile Coverage Obligations2, listed various methodologies for monitoring mobile coverage. The methodology options include crowdsourcing, (mentioning OpenSignal as an example by name) although the document states that ‘This method needs further investigation’. The report also recommends that ‘describing the different practices on how coverage is defined and assessed among the CEPT countries could help administrations in supervision’. OpenSignal uses a crowdsourced methodology and we very much welcome further investigation, dialogue and education on this mechanism, and provide comments on the consultation document for Ofcom through the lens of offering details on how this methodology works, as recommended by the CEPT. About OpenSignal The OpenSignal data is collected from real world consumer smartphones, and is recorded under conditions of normal usage. Rather than approximate the user experience, we directly measure it from the users of our smartphone application. Our application can be freely downloaded on either iOS or Android devices and constantly monitors the true network experience that users are getting on those devices. Since our data is gathered from consumer smartphones, we are able to observe the network exactly as the end user experiences it. This customer-‐centric approach allows us to measure the true end-‐to-‐end experience of the mobile network. We do not run models, simulations or make assumptions – our goal is to directly measure user experience through the eyes of the users themselves. Although operators monitor how their networks perform, there remains a disconnect
1 http://opensignal.com/ 2http://www.cept.org/files/1051/Tools%20and%20Services/Public%20Consultations/2014/Draft%20ECC%20Report%20231%2 0for%20PC.docx
between the standard network KPIs and what surveys say about customer experience. We believe that the only way to bridge this gap is to measure the network using the customer experience as our starting point. The OpenSignal website provides consumer-‐focused visualizations and analysis, based exclusively on data collected from the users of the OpenSignal application. Our website’s features include coverage maps that show cellular signal strength for a given geographical area, ranking all of the networks by performance in that location, and the locations of all cell towers within that particular region. User experience Under the strategic purpose of “Promote opportunities to participate” the annual plan lists the priority “Promote better coverage of fixed and mobile services for residential and business consumers”. Within this priority, the Annual Plan states “We will develop a consistent approach to reporting on coverage and quality of experience of mobile services that reflects real consumer experiences, and publish operator-‐specific metrics to inform consumer choices.” This directly aligns with OpenSignal’s methodology and mission of reporting on and providing data on real consumer experience, and we support Ofcom’s focus in this area. Our data tests, run in our app, are designed to measure a user’s average experience as closely as possible. Unlike other testing methodologies, such as models or drive testing, we do not optimize anything related to the testing scenario; we instead keep conditions as close to normal usage as possible. This allows insights to be gathered on network performance that are not otherwise possible. For example, we provided data to CSL, an operator in Hong Kong, who wanted to • Gain a deeper insight into the experience on their network from a customer-‐centric point of view. • Gain an understanding of metrics they were not currently able to track such as Wifi usage behavior or the proportion of time users have no network coverage. • Benchmark the performance of their network against their competitors whilst reducing the large capital expenditure on drive testing. Using our data in Hong Kong, our analysis allowed for the identification of an issue that was causing a large proportion of their users to experience poor latencies, enabling the operator to roll out a fix to their core network within 10 days. This particular problem had only been affecting users on lower end devices, which had not been included in the device subset used in the drive and lab testing, and so the existing testing methods had completely missed the problem. The data also provided insight into certain metrics that were previously opaque such as: • Seeing the proportion of time the average user was connected to LTE, in order to assess the success of their current LTE network rollout. • Understanding how their unlimited data plans was impacting Wifi usage behavior. • Seeing the proportion of their customers and their competitor’s customers that were experiencing throttled data speeds.
The Annual plan references Ofcom’s Consumer experience report3, which states “We recognise that the planning tools are subject to a margin of error and local factors, such as tall buildings or trees, can affect the signal at different locations. In addition, the quality of mobile services are affected by factors other than signal strength, such as network capacity, number of simultaneous users and quality of handset. In 2014 we will be undertaking work to measure the actual consumer experience” OpenSignal supports this goal and recommends the use of a crowdsoured methodology as the only way to successfully measure actual consumer experience. The Ofcom report goes on to site OpenSignal as an example tool for using consumer-‐end devices to measure mobile coverage. The Annual Plan also references Ofcom’s most recent Infrastructure Report4 that discusses the particular methodologies for measuring coverage and user experience. The report states: “The consumer experience of mobile services depends very substantially on the quality of mobile coverage. There are however two practical issues which complicate measurement of mobile coverage: The maps of mobile coverage produced by operators are based on theoretical models, calibrated using measurements of actual performance that are broadly accurate overall but can never be absolutely accurate in predicting coverage at a specific location. We will continue to carry out work in this area. Consumers use mobile phones in many different situations – indoors, outdoors, on the move, in cars, as pedestrians along roads in built-‐up areas and in wide open spaces. No single measure of coverage can capture all these use cases. For this report, we have assessed mobile coverage against a range of different measures in order to gauge the overall consumer experience.” At OpenSignal we would argue that a crowdsourced methodology a) is independent and empirical, and therefore removes the first issue stated of models that are produced by the operators themselves, and b) does in fact take into account the various use cases described above indicative of user behaviour. Since data is gathered while these activities are going on, the datasets and are representative of the amount of time users actually spend on these activities. Moreover, since data is collected both actively and passively, it is possible to detect and account for any potential bias of actively submitted data. Competition and informed choices for consumers Strategic Purpose 1 of the Annual Plan is ‘promote effective competition and informed choice’, with the work area: “Promote effective choice for consumers by ensuring that clear and relevant information is readily available”. The Annual Plan goes on to state in section A2.15 “Consumer information plays a critical role in ensuring that competitive
3 http://stakeholders.ofcom.org.uk/market-data-research/market-data/consumer-experiencereports/?a=0 4 http://stakeholders.ofcom.org.uk/market-data-research/market-data/infrastructure/infrastructure2014/
communications markets work for consumers. We will continue to publish information in a range of areas to ensure that consumers can access clear and relevant information about the performance of different providers.” OpenSignal embraces this approach and benchmarks network operators’ performance against each other, providing this information publically so consumers can make informed choices. Accurate and useful information to aid consumer choice leads to healthy competition and encourages companies to innovate, improve their service offering and lower costs, which can in turn be passed onto consumers. When network operators self-‐ report coverage, it is often hard to compare on a consistent basis. Drive testing methodologies differ, as do testing devices, making apples-‐to-‐apples comparison across networks effectively impossible. A crowdsourced methodology, independent of network and device type allows for consistent reporting of performance of the competition. OpenSignal’s crowdsourced data compares network operators’ performance at the user level. One way in which this data is presented is through our coverage map, available on our website and in the app itself. This map tool allows the user to compare all network types in a given region, and filter to view the coverage of both individual operators and network types (e.g. 2G, 3G or 4G). Regular analysis of the performance of networks within a country can highlight opportunities for competition for infrastructure improvements. An example is OpenSignal’s study on Denmark’s Notspots in 20135. The report found 21,098 'notspots' -‐ areas where multiple users reported have no 3G access on a particular network, demonstrating a tangible opportunity for Danish networks to expand their service into these areas. OpenSignal’s NetworkRank algorithm uses a series of Quality of Service (QoS) parameters (listed below), all of which are measured empirically from users of the OpenSignal mobile app. To calculate the ranking, the score assigned to an operator for a given parameter is calculated according to the best performing operator for that parameter in a given region. In other words, it is a measure of the performance of an operator in relation to the local best operator. Therefore the network rank is not based on an absolute scale and rankings of operators from different regions cannot be compared directly. However, OpenSignal’s coverage maps on the website and in the app also provide average values of network speeds and latency, which can be compared across markets. Network Rank Parameters • Signal Coverage: This is an average of the signal strength reported. For LTE this is the RSRP, but for all other network types this is the RSSI (as defined by 3GPP specifications TS 27.007 8.5). • Download Speed: This is an average of the download speeds reported by any data network tests (see data network test section or more detail) carried out in the region. • Upload Speed: This is an average of the upload speeds reported by any data network tests carried out in the region. • Latency: This is an average of the latency times reported by any data network tests carried out in the region. • Data Reliability: This is based on a combination of the success rates of the latency, download and upload tests. It is the proportion of data requests that are
5 http://opensignal.com/reports/denmark-notspots-jan-2013.php
successfully completed. OpenSignal has partnered with consumer groups to ensure that end-‐users are informed of the performance of the network options they are considering. For example, we have partnered with Which6, the UK based consumer group, to embed our coverage map on their website and to publish a number of reports7 examining the status of coverage in several UK cities. We are also in discussions with several other consumer groups across Europe and Latin America to explore the same type of partnership. OpenSignal has also partnered will cell phone retail websites to allow consumers to be fully informed of their options, such as mobiles.co.uk8, which also has an embedded coverage map on their website. Since Opensignal empowers users to report on the quality of their cell phone coverage, both actively and passively through the application, this creates an opportunity for network operators to compete to offer better services to their customers. This in turn encourages and incentivizes further investment in network infrastructure. Switching For the Priority “Improve the process of switching providers for consumers,” Section A1.15 of the Annual plan states “In order to take advantage of competition in the communications sector, it is vital for consumers to be able to exercise choice by switching providers easily. It remains important that we make further progress in ensuring the effectiveness of switching procedures for consumers.” At OpenSignal we also support consumers having the option to switch between mobile networks as this promotes efficient markets and infrastructure improvements. We are working with consumer groups to provide our coverage maps for group customer tariff switching between networks based on reverse auctions by the networks.
Reporting on mobile coverage and speeds The Annual Plan explains that Ofcom will “publish mobile broadband speeds and mobile coverage information to enable consumers to compare the mobile coverage that each MNO offers across the UK’s nations.” Further, it is an interim outcome of the work plan that Ofcom will indeed, specifically “publish mobile broadband speeds research in Q1 2015/16 and 2G, 3G and 4G coverage research in Q3 2015/16.”
6 http://www.which.co.uk/technology/phones/reviews-ns/best-mobile-phone-networks/best-mobilenetworks-overview/ 7 http://www.which.co.uk/technology/phones/reviews-ns/best-mobile-phone-networks/mobile-phonenetwork-state-of-the-nation-report/ 8 http://www.mobiles.co.uk/4g/coverage-checker-map.html
At OpenSignal we strongly support Ofcom’s plans to publish speed and coverage data. The Annual Plan provides a link to a recent report 9 by Ofcom on measuring broadband performance in the UK, on 3G and 4G services offered by the UK’s MNOs. While OpenSignal supports making information of this nature public, there are limitations to the methodologies used by Ofcom to calculate the results, which could be ameliorated with consideration of crowdsourced data on mobile user experience. OpenSignal published a response to this Ofcom report, available at http://opensignal.com/blog/2014/11/14/the-‐ ofcom-‐mobile-‐broadband-‐report-‐a-‐response/. An excerpt from this response is included below, and the full response is included as an annex to this document. Excerpt of OpenSignal’s response to Ofcom’s report ‘Measuring mobile broadband performance in the UK’ “We feel very strongly that these kind of reports are exactly what OfCom should be doing, and that any and all independent data made available to consumers is a positive step leading to a more efficient market. We have made clear in the past that we want OfCom to do more independent testing, and make use of different datasets (such as OpenSignal) – so that better consumer information on network performance will force the operators to compete more in terms of the actual cellular service they provide. We are supporters of all independent information that helps to achieve this end, and we are delighted that OfCom chose to cite our report, as it shows they are open to innovation and aware of the limitations of traditional testing methods. That being said, we retain a number of reservations about the methodology used in the OfCom study.” The response goes on to list specific limitations (available as an annex to this document). In summary, taking account of data that is truly representative of phone users’ experience can facilitate in obtaining more useful and accurate data for consumers.
Quality of service The Annual Plan explains in section 2.14, “Even in areas where fixed and mobile networks offer widespread coverage, consumers and businesses have expressed concerns about the quality of service that they receive from communications providers.” At OpenSignal we agree that it is important not only to measure and report on coverage, but also on the quality of the connection that the phone user experiences. The crowdsourcing approach that OpenSignal adopts allows for both of these measures. The variable that we use to measure coverage is our ‘time on’ metric, which is the proportion of time a phone user on a given network and technology (e.g. 3G, LTE etc) has access to that network. Quality of service and performance data collected includes variables such as upload and download speed, latency, data success rates and signal strength. International collaboration As part of the work area “contribute to and implement public policy defined by Parliament” is the activity “Engage with the legislative programme of the new European Commission.” The Annual Plan explains that Ofcom
9 http://stakeholders.ofcom.org.uk/market-data-research/other/telecoms-research/broadbandspeeds/mobile-bb-nov14
“actively participate[s] within European regulatory networks including the Body of European Regulators for Electronic Communications (BEREC), the European Platform of Regulatory Authorities (EPRA), the European Regulators Group for Audiovisual Media Services (ERGA) and the European Regulators Group for Postal Services (ERGP). We are also active within the Radio Spectrum Policy Group (RSPG), which advises the European Commission on spectrum policy-‐ making.” OpenSignal understands Ofcom collaborates with a number of international bodies also striving for the same goals of improving connectivity for consumers. There are challenges with adopting consistent methodologies to report on mobile network coverage across borders. An additional benefit of crowdsourcing data from users’ phones it that it is agnostic of nation states, allowing for consistent frameworks for reporting on and comparing coverage between countries. OpenSignal has engaged with many multilateral organizations in this space through presentations, private meeting and strategy consultations, including BEREC, the CEPT and the European Commission. OpenSignal presented on crowdsource methodology at the International Telecommunications Union (ITU) Quality of Service Development Group in Dubai in November 2014. summary In Summary, OpenSignal supports much of the Ofcom Annual Plan’s priorities. However, it is important to consider the methodology by which these actions are carried out, and OpenSignal believes that crowdsourcing user experience data is a powerful tool, and for the most part, necessary, for obtaining and publishing accurate data and achieving the aims laid out in the consultation document. For more information, contact Ellie Ereira at
[email protected] or +447929854557.
ANNEX
The OfCom Mobile Broadband Report: a response on November 14, 2014 on the OpenSignal blog. Web version available at http://opensignal.com/blog/2014/11/14/the-‐ofcom-‐mobile-‐broadband-‐report-‐a-‐response/ Posted
Yesterday OfCom released a report [pdf]10 on the current state of UK mobile networks – a report similar to ours on the subject from last week11, which found that 4G speeds had halved over the past year in the UK. OfCom focussed their report on network performance, rather than coverage, and cited our report to show that independent analysis is being carried out on the comparative coverage provided by MNO’s. Importantly, OfCom carried out their study by comparing performance in sites around the UK, both indoor and outdoor, where all four networks were present – and their findings were different to ours, with higher reported speeds on both 3G and 4G. In the OfCom report EE ranked as the fastest 4G LTE network while Vodafone were the fastest 4G network based on OpenSignal user data. We feel very strongly that these kind of reports are exactly what OfCom should be doing, and that any and all independent data made available to consumers is a positive step leading to a more efficient market. We have made clear in the past that we want OfCom to do more independent testing, and make use of different datasets (such as OpenSignal) – so that better consumer information on network performance will force the operators to compete more in terms of the actual cellular service they provide. We are supporters of all independent information that helps to achieve this end, and we are delighted that OfCom chose to cite our report, as it shows they are open to innovation and aware of the limitations of traditional testing methods. That being said, we retain a number of reservations about the methodology used in the OfCom study. 1) The data is up to eight months old: With the current pace of network upgrades, using data up to eight months old is not likely to reflect the current state of network performance. OfCom’s data collection ran from March to June, while our OpenSignal report covers the three months leading to October 1. This is especially a concern when it comes to 4G measurements, as our UK report showed that average 4G LTE speeds across all networks had fallen 15% from March to October, meaning that OfCom may be overstating the speeds currently experienced by real consumers. 2) The type of device used: OfCom tested using a Galaxy Note III, a mobile phone which is available across all networks. They justify this by saying they want to test the network without it being affected by the user’s device, as different devices experience different speeds. While this is
10 http://www.ofcom.org.uk/static/research/mbb.pdf 11 http://opensignal.com/reports/2014/10/uk-networks-report/
a legitimate way to judge comparative network performance, we do not feel that it is the best way to show the real user experience of the network. Testing using a consumer device in this way sits at an uncomfortable halfway point between using specialised testing equipment and data crowdsourced from real user devices, and is unable to fully capture the full spread of user experience. Many users do not have high-‐end devices, so testing using a device such as the Galaxy Note III is likely to overstate speeds. Our network averages represent the fact that users have many different devices, and our data is therefore naturally weighted by device market share – as every user counts the same – giving a more accurate picture of the diversity of the UK’s mobile device usage. We feel our methodology complements the OfCom data, as it would be too time consuming to road-‐test using a full spread of devices and OfCom do not claim to be directly measuring the user experience, despite using a consumer device for their testing. 3) The location of tests: OfCom test both indoors and outdoors (a positive development from the days when network testing included almost no indoor tests) but they test 50% on each, which is not necessarily representative of typical use. Above all, by testing in areas of good connectivity (where all four networks are present) they are potentially skewing their results to the faster end. 4) Testing on EE ‘double speed’: OfCom ran their tests using the EE 4G ‘double speed’ sim (without fully explaining why, as they stated they intended to test both the ‘single’ and ‘double’ speed tariffs, but ended up only testing the faster one). Many EE 4G users are not on the faster speed tariff and so reported speeds for EE are only representative of the experience available to users who are paying for the faster data, and therefore possibly not representative of those who are paying for 4G LTE on EE on the ‘single speed’ network. 5) What OfCom are actually measuring: OfCom are not measuring the typical speeds users actually experience (and this is by design), and therefore overstate the speeds consumers are likely to get. This explains why we report the mobile networks to be slower on both 3G and 4G than OfCom, as our testing is directly measuring performance as experienced by users rather than modelling it based on controlled tests. 6) Lack of Coverage data: OfCom rely on coverage data self-‐reported by the operators themselves, looking at the proportion of premises covered in the UK. This metric is an attempt to combine raw geographic coverage with is impact on users, but we feel that, in isolation, it is some way divorced from the actual availability of networks for consumers. OfCom cite our ‘time on’ metric, which looks at the proportion of time users have a connection, as an alternative metric for coverage and we feel it is entirely complementary to the more traditional geographic metrics used by OfCom, as it helps put the ‘premises’ figure into perspective. OfCom’s testing methodology is not able to gather accurate data on coverage, and this means that their report on network performance cannot be entirely complete, as it only records data from where all networks are present. While differences in performance are right to be noted, and can be significant, what is more significant is the actual availability of the network itself.
The rise of independent reporting is vital for the on-‐going success of a competitive mobile market in the UK, and we feel that OfCom’s report is important for bringing questions of network performance to the forefront of consumers’ minds. We do, however, believe that additional useful and up-‐to-‐date information can be made available to consumers through additional techniques, 12
such
as
crowdsourcing,
and
would
13
encourage
the
OfCom
[pdf] and OpenSignal reports to be read side-‐by-‐side to paint a more complete picture of the current state of network provision in the United Kingdom.
12 http://www.ofcom.org.uk/static/research/mbb.pdf 13 http://opensignal.com/reports/2014/10/uk-networks-report/