<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Narendran Sivakumar on Medium]]></title>
        <description><![CDATA[Stories by Narendran Sivakumar on Medium]]></description>
        <link>https://medium.com/@narendransivakumar1?source=rss-216b7c0837c2------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*6xtba240OAg9Iwr8tFafCA@2x.jpeg</url>
            <title>Stories by Narendran Sivakumar on Medium</title>
            <link>https://medium.com/@narendransivakumar1?source=rss-216b7c0837c2------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 07:20:39 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@narendransivakumar1/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[The Future <> and its not just 5G!!]]></title>
            <link>https://medium.com/@narendransivakumar1/the-future-and-its-not-just-5g-aa80f29657c?source=rss-216b7c0837c2------2</link>
            <guid isPermaLink="false">https://medium.com/p/aa80f29657c</guid>
            <category><![CDATA[innovation]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[nlp]]></category>
            <category><![CDATA[5g]]></category>
            <category><![CDATA[strategy]]></category>
            <dc:creator><![CDATA[Narendran Sivakumar]]></dc:creator>
            <pubDate>Tue, 15 Sep 2020 10:31:31 GMT</pubDate>
            <atom:updated>2020-10-23T16:07:50.269Z</atom:updated>
            <content:encoded><![CDATA[<p>Let me be a maverick and start off with a prolog. I am adopting a non-linear narrative for this post, for a reason.</p><p><strong>Prologue</strong>: Please do not miss the postscript. Postscript is, in fact, the main post. Read first few paragraphs and skip over if you are not a 5G nerd.</p><blockquote><strong>Advanced Analytics:</strong></blockquote><blockquote>Offers an introduction to the wide and robust consideration of the vertical industry’s stakeholders. The “From Software To Network” rationale is based on the ETSI NFV standardisation programme introduction: SMURF. The rationale is still evolving and it is recommended that the NSD be included in the final NSD for the main components of the AADU. It has a clear categorisation by the following groups:</blockquote><blockquote><strong>General Purpose</strong>: Provides a general overview of the functionality in an deployed IT infrastructure.</blockquote><blockquote>Transport Network Operators (TNO): These are people responsible forproviding the data collection and analysis, management and orchestration, and the terminating of all links connecting the two networks.</blockquote><blockquote>Core Network Operators (CN): These are people attached to a specific network and provide the data collection and analysis, management and orchestration, and the terminating of all links connecting the two networks.</blockquote><blockquote>Data Centre Operators (DCOs): These are people who are responsible for the provisioning of the data centres. They may be called the “data analytics” people or “data analytics maintenance”.</blockquote><blockquote>Data analytics practitioners (DAPs): These are the people who do data analytics for the AADU. They are responsible for collecting analytics data from the AADU and sorting it into one or more collections.</blockquote><blockquote>Data analytics students (DAPs): These are the people who are studying the AADU at an early stage in its development. They are mostly computer scientists or data analytics professionals and are considered to be in charge of the development of the AADU.</blockquote><blockquote>Data centre administrators (DAFs): following the AADU programme specification, adheres to the overall flow of the programme. The AADU starts with a programme specification and a network slice as a starting point. After that, it is up to the AADU to setup the network slice.</blockquote><blockquote><strong>Data analytics functions (DAs)</strong> can be introduced into RAN that can detect and mitigate any effects of network slicing.</blockquote><blockquote>A network slice awareness and de-icing function (DAF) is created by adding RAN-like functions to the existing ones. It is recommended to include in the NFV architecture any software functions that can be used by the Network Functions Virtualisation (NFV) Orchestrator (NFVO) to deliver the virtualised network resources as well as to provide the virtualised network infrastructure segments.</blockquote><blockquote>A network slice execution environment can be defined from two perspectives:</blockquote><blockquote>A static model of the network that is deployed at the edge, allowing the NFV Orchestrator to deploy the virtual networks at the physical devices (by way of an NFV-QoS framework).</blockquote><blockquote>A network slice deployment environment can be defined in terms of the QoS profiles of the network elements. The QoS profiles include the Packet Error Rate (PER), the Allocation and Retention Priority (ARP), the Allocation and Retention Level (AL): ALT: low, Medium: high, and General Purpose (GPP): drone IP: protocol, channel, and interface pointers of the virtual network elements.</blockquote><blockquote>A network slice deployment context can be defined in terms of the QoS profiles of the network elements. The QoS profiles can be used to fine-tune the NFV architecture, improve the reliability of the network service, or both.</blockquote><blockquote><strong>Virtual Linkser (VL)</strong>: A network slice instance can be configured to use one or more Virtual Links (VLCs) and can be used with different BSAN configurations. The configuration desired depends on the complexity, the availability of the network service, and the requirements of the VLC instance. The network slice instance can also be used as a VNF host for different VNF applications. An example application that can be easily deployed and managed on the VLC is the internet of things (IoT) program.</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*CH32e8Xw1ZJ37U1n" /><figcaption>Photo by <a href="https://unsplash.com/@lazycreekimages?utm_source=medium&amp;utm_medium=referral">Michael Dziedzic</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><h3>Postscript: What if</h3><p>What if — what you had read above is not written by me. What if — what you had read above is not even written by a human.</p><p>The text has been infact written by an AI algorithm. The AI algorithm that wrote that text is a GPT 2 model that was opensourced by OpenAI. I downloaded the simple version of the model that comes pre-trained with only “1.5 billion params”. The sucessor to GPT2, GPT 3 has infact been trained on 1.75 trillion params. GPT 3 is much more advanced. I’ll come back to that.</p><p>According to <a href="https://openai.com/blog/better-language-models/">OpenAI</a>, “GPT-2 is a large <a href="https://arxiv.org/abs/1706.03762">transformer</a>-based language model with 1.5 billion parameters, trained on a dataset<a href="https://openai.com/blog/better-language-models/#fn1">[1]</a>of 8 million web pages. GPT-2 is trained with a simple objective: predict the next word, given all of the previous words within some text. The diversity of the dataset causes this simple goal to contain naturally occurring demonstrations of many tasks across diverse domains. GPT-2 is a direct scale-up of GPT, with more than 10X the parameters and trained on more than 10X the amount of data”.</p><p><strong>A ground shift in NLP has happened</strong></p><p>The traditional NLP approach of training models on large body of labelled data set has not yielded results. In fact, research indicates that pre-trained models based on large labelled data sets do not generalise well for scenarios that are not part of training. It is not so surprising as, we — the humans, do not learn language and develop skills like that. We develop language skills with surprisingly fewer number of examples. That is because we learn language by identifying patterns and the sequence in which they appear — called as <strong>Language Meta Learning </strong>in NLP speak.</p><p>More importantly, the models try to learn how context shapes the patterns and the sequences — called as <strong>in-context learning</strong>.</p><p><strong>What if </strong>— we replicate this for NLP whereby we build models that does not focus on the exactness of the training data set, but focusses on the patterns and sequences, and the context under which they get invoked. GPT family of models — GPT 3 being the latest — achieves this using transformer architectures (basically a complex and very deep neural net that takes the language corpus — entire Wikipedia, Reddit and many more — as input and develops an intermediary machine language that focusses on patterns and sequences). The output of these models is generated based on most likely predicted occurrence of patterns and sequences, fine-tuned over a period of training on vast data sets.</p><h4>In other words, this approach is equivalent to someone going from point A to point B by training to recognise and navigate based on roadside signs as opposed to someone who has to memorise the entire route. The former is likely to be accurate and the former is likely to adapt to unforeseen circumstances compared to the latter where anything out of ordinary can trip the plan.</h4><p>The transformers are, in effect, a language that stores the learnings on how to navigate using the roadside signs. This ground breaking technique essentially means that in real world, fewer number of training is required for a model to adapt and respond.</p><p>The ease at which a language model adapts to newer and untrained information is very important and it determines how close a language model is to adhere to turing compatiability test. A turing compatibility test is an observation and validation measure to asses the degrees of indistinguishability between a human response and a response generated by a language model. Basically, in other words, it means that a layperson cannot discern the source of the response as a human or a machine.</p><p>Model’s learning adaptability is defined in terms:</p><ul><li>Zero-shot: how quickly a model can learn and adapt to a new situation — totally blind and no learning. Equivalent to someone appearing for an advanced calculus exam with no knowledge of calculus or for that matter math.</li><li>One-shot: how quickly a model can learn and adapt to a new situation based on learning from one — yes, one — example. Equivalent to someone appearing for an advanced calculus exam by solving just one problem.</li><li>Few-shot: how quickly a model can learn from many examples. Usually, from 10 to 100 or little more. More like how we all will study for calculus.</li></ul><p>GPT3 takes this to a different level. To quote from the <a href="https://arxiv.org/pdf/2005.14165.pdf">seminal GPT3 paper from OpenAI</a> :</p><blockquote>“we test this hypothesis by training a 175 billion parameter autoregressive language model, which we call GPT-3, and measuring its in-context learning abilities. Specifically, we evaluate GPT-3 on over two dozen NLP datasets, as well as several novel tasks designed to test rapid adaptation to tasks unlikely to be directly contained in the training set. For each task, we evaluate GPT-3 under 3 conditions: (a) “few-shot learning”, or in-context learning where we allow as many demonstrations as will fit into the model’s context window (typically 10 to 100), (b) “one-shot learning”, where we allow only one demonstration, and © “zero-shot” learning, where no demonstrations are allowed and only an instruction in natural language is given to the model...”</blockquote><p>NOTE: Demonstration is the NLP speak for “example of an occurrence” that the model can learn from.</p><p><strong>In our language model which became a 5G nerd</strong>, I quickly used a repurposed GPT -2 model developed by <a href="https://minimaxir.com/2019/09/howto-gpt2/">Max Wolf</a> (as GPT 3 is not open sourced) and used a simple implementation of GPT -2 which comes pre-trained on 124Million params.</p><p>To make it a 5G nerd, I trained the model on <a href="https://5g-ppp.eu/wp-content/uploads/2019/07/5G-PPP-5G-Architecture-White-Paper_v3.0_PublicConsultation.pdf">one document</a> — just one document — which is a public discussion document on 5G released by EU and 5G PPP.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*BSbT8F5XYWqtda5zOUA0uQ.png" /><figcaption>The bureaucratease on 5G that helped the language model to become a 5G nerd. Thanks EU!!</figcaption></figure><p>Look at the sample and random learnt model extracts before even I coded the model to write something about Analytics in 5G:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_YwxEid_pSKBwBURJkLQbg.png" /><figcaption>Not bad eh! Sample extracts just after 200 training runs.</figcaption></figure><p>Here are some of the training outputs from the model — not bad eh!</p><blockquote>“constituent MAC addresses (constituent.h: small number of IP addresses that a small network operator can establish around their own networks). The envisioned architecture does not include any multicast and broadcast capabilities. It can be expected that the 5G system will adopt this feature or that another multicast or broadcast capability would be beneficial to the system, such as, load balancing. The 5G System: — The 5G system is being designed to meet the requirements of a highly mobile and fully connected society. A truly global society therefore requires a diverse set of services and services are recommended to support these services. Service manufacturers, corporates, research and development organizations, telecommunication and cloud computing and networking services producers, research and development organizations and industry organizations should work together to bring about the structured applications common to all of the relevant domains in the proposed 5G system. — The 5G system is being designed to meet the requirements of a highly mobile and fully connected society, requiring both optimized services for various functionalities, and optimizations for the optimization of the user experience. The 5G system architecture in Rel., for instance, presents a number of solutions to different accessibility and service management requirements in order to create a truly global system, in which both the most used service and its components can be isolated and managed independently, avoiding long lists of services.</blockquote><p>In the above extract, look at the way the language model, in braces explains what constituent MAC addresses mean. In the document there is a fat glossary which explains the terms and MAC addresses are <strong>NOT</strong> defined there. I did a plagiarism check to discover the source and it comes from Wikipedia!!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-Phrs14lseR0bypKyGvA9w.png" /><figcaption><a href="https://plagiarismdetector.net/">Plagiarism checker </a>has validated that rest of the text is original</figcaption></figure><p>In fact, I did push the entire text written by the model for a plagiarism check just to be sure. Look what I found out:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*r2uP0gmlu4_0zF-od0zOsA.png" /><figcaption>97% original and 3% plagiarised is as original as it gets!!</figcaption></figure><p><strong>Conclusion: Ground has shifted decisively in AI</strong></p><p>The future is here right now, powered by AI. AI will take over most of the knowledge synthesis, summarisation and complex conversational tasks as we found out when I coded the model to begin a paragraph on 5G begining with Advanced Analytics:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*TwqWsre2IV9wP3vTNhXQkA.png" /><figcaption>GPT2 model that is a 5G nerd coded to generate a 1023 word writeup on Advanced Analytics</figcaption></figure><p>The question that beckons us right now is how we can use AI to reimagine our work, our enterprises, our lives and our world. The future is already here folks, and it is waiting for you …</p><p>Credit for the Tensorflow based wrapper of GPT 2 Model goes to Max Woolf and his blog <a href="https://minimaxir.com/2019/09/howto-gpt2/">MiniMaxir</a>.GPT 2 is developed by <a href="https://openai.com/">OpenAI</a>. GPT 3 is the successor to GPT2.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=aa80f29657c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Rise of Multiverse and what it means for CMOs]]></title>
            <link>https://medium.com/@narendransivakumar1/the-rise-of-multiverse-and-what-it-means-for-cmos-cc5ce2d410ef?source=rss-216b7c0837c2------2</link>
            <guid isPermaLink="false">https://medium.com/p/cc5ce2d410ef</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[business-strategy]]></category>
            <category><![CDATA[strategy]]></category>
            <category><![CDATA[cmo]]></category>
            <category><![CDATA[customer-experience]]></category>
            <dc:creator><![CDATA[Narendran Sivakumar]]></dc:creator>
            <pubDate>Mon, 20 Jul 2020 23:27:39 GMT</pubDate>
            <atom:updated>2020-07-22T19:00:04.240Z</atom:updated>
            <content:encoded><![CDATA[<p>First things first. I hope all of you are safe and well.</p><p>And as anticipated COVID has proved and disproved every model that has been conceived, which is actually not surprising if you follow what goes behind these models.What is surprising though is that there is so much in the universe that is unknown and unfathomable even now.</p><p>Now to the subject of universe or the rather the multiverse. Most of you will know multiverse if you are Marvel Avenger aficionados or followers of scientists like Stephen Hawking or Neil deGrasse Tyson. Well, I am not going to delve deeper into the world of Marvel and Avengers Endgame. I am actually going to talk about a cultural phenomenon of Multiverse that is inexorably growing in our deeply polarising world.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*AQzT84QsjDe-7tOR.png" /><figcaption>Where am I? — Source: Wikimedia Commons, Author: Silver Spoon</figcaption></figure><h4>Polarisation is the norm. Cancel is not just an innocuous verb. It is a culture now.</h4><p>Without delving deeper into any of these and without taking sides, Bari Weiss one of the proclaimed centrists quit New York Times as its opinion editor. She did so with an exquisitely carved <a href="https://www.bariweiss.com/resignation-letter">resignation letter</a> as one would expect from Bari Weiss, an indictment of the strange circumstances that we find ourselves in:</p><blockquote>…that truth isn’t a process of collective discovery, but an orthodoxy already known to an enlightened few whose job is to inform everyone else… Twitter is not on the masthead of The New York Times. But Twitter has become its ultimate editor.</blockquote><p>You would imagine that this is a devastating critique of the newspaper of record. And this will impact its growth. May even spurn existing customers. Surprisingly, no!</p><p>NYT increased its base by 1 million subscribers in just a quarter — from <a href="https://www.nytimes.com/2020/02/06/business/new-york-times-earning.html">February </a>to <a href="https://www.nytimes.com/2020/05/06/business/media/new-york-times-earnings-subscriptions-coronavirus.html">May </a>2020. This can only indicate that people are increasingly seeking what conforms to their beliefs (and outrage!) and to quote Ms Weiss:</p><blockquote>…. (Why) edit something challenging to our readers, or write something bold only to go through the numbing process of making it ideologically kosher, when we can assure ourselves of job security (and clicks) by publishing our 4000th op-ed arguing that Donald Trump is a unique danger to the country and the world?</blockquote><p>Without going into the correctness of it, one can’t fault NYT as a business. It is indeed driven by the market forces and is rewarded by its customer segments on the direction it has taken.<strong> The market it seems loves echo chambers.</strong></p><h4>Brands &amp; Marketers — you have a problem now!</h4><p>Amidst this whirlpool of righteousness, cause driven activism, liberalism, fundamentalist liberalism, nationalism and many more ‘nesses’ and ‘isms’, Brands have to chart a path that is delicate and sensitive in a way that it appeals to all but is endearing to none. Or they are not seen as espousing someone else’s cause until and unless that cause is ethical and moral like <a href="https://www.bbc.com/worklife/article/20200612-black-lives-matter-do-companies-really-support-the-cause">Black Lives M</a>atter.</p><p>The key conundrum for brand and marketing strategists is that there is no clearly defined boundary between a brand appeal and a brand cause or endearment. If a brand is built around a cause or is associated with a cause like for example Apple and its absolutist position on <a href="https://www.nytimes.com/2016/02/18/technology/apples-stance-highlights-a-more-confrontational-tech-industry.html">privacy</a>, then the overlap between appeal and endearment is welcome and in fact, that overlap is a competitive advantage. Given that scenario, <strong>the marketing managers need to create campaigns that, nudges customers and the segments they belong to, from the brand appeal onto brand endearment.</strong> Think of Apple and its campaigns — the focus is on creating brand endearment. Think of Nike.</p><p>On the other hand, if you are a brand that is easily replaceable with low switching costs for consumers like Cola brands, or a mass market brand or a brand that dances to the performance metrics like Monthly Active Users, Daily Active Users, Gross Additions etc, then the overlap can be a problem. In fact, it can be a deterrent.<strong> Think of a digital native with hockey-stick growth in terms of users and that it treats all of its customers and segments as one universe. It will struggle to reconcile the opposing poles of customer opinion, just like in the case of NYT, expressed in terms of market forces. It is inevitable that who you (the brand) are and what you represent will be attractive to significant chunks of customer segments and, at the same time, repulsive to other significant chunks of segments in the same universe. M</strong>oreover, as the world gets swept by viral events — no pun intended — and daily outrages, who you are and what you represent changes daily in the eyes of your segments. Think Twitter and Facebook. Widely disparaged. Widely used. Widely embraced.</p><blockquote>Worse, the customer segments are not even a constant. The segments change and evolve driven by events.</blockquote><p>What one collection of customer segments likes, the other detests. There can be common ground one day for one cause or event, not the other day. You cannot appeal for the lowest common denominator as well because it will only commoditise the brand proposition and strip away all its uniqueness — the LCM of uniqueness is called mundane.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0I7jdK0YpkTbUMwzk9Am8A.png" /><figcaption>How events continuously influences customers, shapes segments, and creates Universes</figcaption></figure><p>So, what is a strategy where the Go-to-Universe of a brand is a nebulous quicksand, getting pulled and pushed, form shifting and form shaping, driven by vicissitudes of events beyond the control of the brand?</p><p>The answer is to not to treat that there is one common universe where all customers and brands exist. The answer is to continue to slice and dice the Go-to-Universe into multiple parts — Multiple universes.</p><p>We are moving into the era of Multiverse.</p><p><strong>Multiverse is here to stay!</strong></p><p>There is a world of a difference between a traditional and a multiverse world for brand managers and marketeers.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*2IC6Pb-1tesRn3D3j_W6ww.png" /><figcaption>Difference between Traditional and Multiverse Strategies</figcaption></figure><p><strong>Personalisation will become ineffective! — </strong>In a multiverse world, Personalisation, as we know based on traditional segmentation methods, will be outdated and ineffective. How can you personalise based on a past understanding of a customer behaviour, segment and universe when all are prone to change — daily, monthly or anytime. Enterprises need multiple levels of personalisation strategies now.</p><blockquote>Personalisation Strategies for the Universe, for the segments within the universe, for the customer within the segment. For the segment that moved into this universe from another and for the customer within the segment who moved along. For the customer from another segment in another universe who moved into this universe and this segment…</blockquote><p><strong>Targeting based Customer Segments will become inconsequential. Live Digital Twinning of customers will become the norm! — </strong>The multiverse world would demand marketing strategists to create virtual replica of customers and their evolving behaviour and response actions and one that updates itself for every change and along with it, updates the segment and the universes. Marketeers and brand strategists will engage with these virtual twins or digital customer twins on an individual basis and create moments of differentiated experience and value.</p><p><strong>Branding will move towards creating bespoke Brand experiences on demand for an individual — </strong>The omnibus consumer brand that stands for all and attracts all will be ineffective. In fact, in a multiverse world, if a brand stands for all, in effect, it will stand for none. With pervasiveness of digital mediums in the form of phones, wearable and other digital devices and an understanding of ‘who, what, where and how’ from digital twins, brands will have the ability to create specific brand experiences, in the moment and on demand, for an individual.</p><p><strong>Enterprise Strategy will be freed from staid powerpoint charts and will be wargamed frequently </strong>— The once a year endeavour of creating enterprise strategy will no longer be sustainable. To lead value creation in the multiverse world, Enterprise strategy itself will be extremely agile, wargamed repeatedly with all stakeholders and effectiveness of response actions simulated in the virtual world leveraging deep learning AI models, cutting edge actor — critic architectures and AI driven simulations.</p><p><strong>Artificial Intelligence is the oxygen of Enterprises’ Multiverse Strategy — </strong>AI will be pervasive and its capabilities will power the enterprise response — from branding to marketing, from proposition development to 1:1 engagement, from orchestrating business processes to helping employees, and from self-correcting AI capabilities to responsible AI operating models. The AI fabric wrests on an enterprise meta model that brings together agent based architectures, transparent policy and audit architectures, goal based simulations and reinforcement learning architectures that continues to ensure that the Enterprise is alive and responsive to evolving multiverses.</p><p>In summary, the world is changing at a faster clip, from a monolith to one where people increasingly inhabit social, cultural and market bubbles called multiverse. Products and propositions will be focused for an universe amongst multiverse. The more common a product gets across universes, more likely it is to become commoditised. It is imperative for branding and marketing strategists to recognise that customers will exist within an universe and can move across universes depending on the events that are beyond brand control. Straitjacketing or cookie-cutter process centric approaches will not yield desired outcomes. It is important to evolve strategies that treat each individual customer in the moment and create the experience. The specificity of that experience will drive the enterprise outcomes. The entire strategy will be driven by innumerable signals and triggers from within and external to enterprise, all captured as data. Enterprises, brands and marketeers will have to employ scenario planning, wargaming and other simulation techniques to refine and fine tune their response strategies. And in a multiverse, your strategy is only as good as its relevance to the current moment. As soon as the moment changes, segments dissolve, float away and new ones emerge, and the new segments coalesce into newer universes or merge into another universe. Multiverses are like volcanoes, always churning, getting moulded and shaped into new ones, by events within and outside its control.</p><p>In the end, if you do not like any of this, the best recourse is to a product proposition that is universally appealing and endearing. Black swans can happen too!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=cc5ce2d410ef" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What lies behind the epidemic models used for Corona Virus tracking and public policy response]]></title>
            <link>https://medium.com/@narendransivakumar1/what-lies-behind-the-epidemic-models-used-for-corona-virus-tracking-and-public-policy-response-1e6bf974bc47?source=rss-216b7c0837c2------2</link>
            <guid isPermaLink="false">https://medium.com/p/1e6bf974bc47</guid>
            <category><![CDATA[policy]]></category>
            <category><![CDATA[public-health]]></category>
            <category><![CDATA[business-strategy]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[coronavirus]]></category>
            <dc:creator><![CDATA[Narendran Sivakumar]]></dc:creator>
            <pubDate>Sun, 12 Apr 2020 20:17:54 GMT</pubDate>
            <atom:updated>2020-04-13T18:51:49.703Z</atom:updated>
            <content:encoded><![CDATA[<p>When I looked ahead for 2020, I imagined 2020 to be a year of so many things, 2020 as the year of pandemic is definitely not one among them. Let me begin this post with a hope that the worst is behind us and we will be spared from anything worse.</p><p>The CoVID 19 pandemic is a pandemic like never before. Certainly we have had infectious diseases before — Bubonic Plague, Spanish flu, Ebola, SARS, MERS etc. CoVID 19 stands out because of its sheer infectiousness. I am not an epidemiologist or an Infectious Diseases expert, so I will refrain from commenting on CoVID 19 itself. What I am interested though, is that, for the first time, data and modelling is being leveraged by policy makers to orchestrate the response strategy. Many of you would have already added few more jargons to our arsenal — Flatten the curve, Community Transmission, Social Distancing, Super Spreader etc.</p><p>A lot of Government’s strategy is actually informed by science (data and modelling). UK was the first to parrot that their strategy is driven by detailed modelling. India did the same. So did many other countries. This post attempts to lift the hood covering these impenetrable models that power the projections and policy making. We have also seen a lot of doomsday statistics and model driven projections which are too horrid to be even considered. However, in the day and age of “virality” of social media, these doomsday projections have caused severe panic amongst general population.</p><p>This post also looks at how these models can be applied beyond the world of epidemiology and in business.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*RyZbkYkMh8dVxQ4Z" /><figcaption>Photo by <a href="https://unsplash.com/@giacomocarra?utm_source=medium&amp;utm_medium=referral">Giacomo Carra</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><h3><strong>Stochastic Models</strong></h3><p>At the outset, Stochastic models are models that are used to capture a trend or an outcome (a probability distribution) that is occurring over a period of time. Stochastic models are frequently used in Epidemiology to project the trend of growth and spread of infections. It is these models that are being used to project the spread of Corona virus.</p><p>This post is not going to explain these models, but will seek to highlight the assumptions that underpin these models. So next time, when you see a doomsday statistic that causes panic, there are several layers of assumptions that are behind these projections. It is henceforth important to treat models and the numbers they project as advisory. They do not project reality.</p><p><strong>SIR Model</strong></p><p>The most basic model used in Epidemic forecasting is the SIR model where the S means Susceptible to the infection, I means currently Infected, and R means Recovered from the infection. What this model tries to obtain for a general population is, given the introduction of the infectious disease (called as Index Patient), what is the probability that it will grow from an isolated, yet infectious disease to an epidemic.</p><p>So the equation goes like this: S → I → R</p><p>What we are looking at is a magic number called R0, called Reproductive Rate. R0 basically looks at the number of susceptible (S) individuals that an infected patient spreads the disease. R0 &gt;1 indicates the growth epidemic and R0 &lt; 1 signifies the declining infectious trend. For Corona Virus, R0 is anywhere between 2 to 3, signifying extreme infectiousness.</p><h4><strong>Assumptions behind the model:</strong></h4><p>Whilst this model has been used to study the spread of epidemics like Malaria, and evaluate effectiveness of response actions (India used a variant of this model to evaluate the effectiveness of its polio vaccination drives), it is important to understand the assumptions that underpin this.</p><blockquote>The model assumes that probability of people who are susceptible, probability of people who are likely to be infected and probability of people who recovered are known from past experience</blockquote><p>More importantly, the model also assumes:</p><ul><li>Immune response of individuals are constant and uniform</li><li>Rate of progression where the Index patient infects a Susceptible patient is actually a random number (plucked from Poisson distribution)</li><li>Infectivity (how virulent the virus is) and Infectiousness (how many susceptible people are affected) are both assumed</li><li>The proportion of infections that occur before symptoms develop is good predictor of how easy an infection is to control</li><li>The population mixing is assumed to be random (i.e. the index patient and so many other infectious patient will randomly come into contact with the susceptible patient). Population mix assumption is vital as that determines a group of susceptible individuals over period of time, the rate at which they will get infected, the rate at which recover, the rate at which they spread (R0) before quarantined and the rate at which they may get reinfected, if there is a possibility.</li></ul><blockquote>However, mixing patterns, as we have seen on the ground for CoVID 19 (Wetmarket from Wuhan, Patient 31 of South Korea) are seldom random. There is a significant behavioral and geographical element in determining population mixes. In some cases, in Netherlands for example when measels outbreak broke out, Community mixes were found to be quite important as due to certain community beliefs, a section of population refused to get vaccinated resulting in higher R0 detected only when community mix was assumed as opposed to random mix (R0 will be lower in a random mix).</blockquote><ul><li>Saturation or, in popular parlance, Herd Immunity, is still being debated. Saturation is a probability that the host antibodies kill the virus and thereby develop immunity to the virus. The likely saturation rate in a group of susceptible individuals is still being debated and they are either ignored (mostly the case for CoVID 19 projections) or assumed.</li></ul><h3><strong>SEIR — A variant of SIR</strong></h3><p>This model is a variant of SIR and it introduces a new progressive state (Exposed) in the equation. SEIR stands for:</p><p>S(Susceptible) → (E)Exposed → (I) Infected → (R) Recovered</p><p>In layman terms, the model assumes that people once infected, develop lifelong immunity against the disease and hence the rate probability of transition from exposed to infected will be lower in a random group mixing consisting of people who have developed life long immunity.</p><blockquote>Life long immunity occurs when the body generates antibodies to successfully ward off the infection. The immune systems of ours remembers what antibodies worked against a particular pathogen (virus, bacteria..) and then when a similar virus attacks, generates the same antibodies. Apparently, this immunity detail about antibodies is stored in memory cells. In case anyone is wondering to whither vaccination; the vaccines are essentially live virus (but significantly less virulent) injected in the body to gently nudge the body to remember its immune response and generate antibodies to kill the virus. That’s the reason why most of us take flu shots just in time before the flu season, so that the body is prepped up with enough antibodies waiting to pounce on anything remotely familiar.</blockquote><p>In probability terms, R0 tends to be quite low if the proportion of Exposed Individuals within a mixing is higher than the proportion of people without immunity. This is sometimes called, now infamously used by all politicians, as <strong>Herd Immunity</strong>. Typically, in a random sample population mixing, it takes 60%-70% of people with immunity to completely kill the virus transmission. This I am sure will ring a bell with many of the readers who closely follow the strategy of Sweden, UK and perhaps even US.</p><p>You can even play with your SEIR projections <a href="http://www.public.asu.edu/~hnesse/classes/seir.html">here</a>.</p><p>Let look at the assumptions here, in addition to SIR ones:</p><ul><li>To begin with, the spread from index to susceptible population, and the proportion of exposed, infected and recovered people are assumed at the beginning of the model run.</li><li>The population mixing is also assumed. There are multiple mixing strategies — Heterogeneous population, Homogeneous population or Dynamic (Random) Population. This is a crucial assumption as the S — E — I — R parameters are dependent on the population mixes.</li><li>The probability of people exposed to susceptible individuals getting infected is also assumed.</li><li>The probability that, among the exposed people, the proportion of people without immunity is assumed. And so is the probability of infection to these exposed people without immunity.</li><li>Lastly, the probability of people who recover from being infected is assumed and this assumption follows the similar assumption made for SIR model.</li><li><strong>One of the most important assumption is that the population once assumed, stays fairly static and the population characteristics do not change over the period of time.</strong></li></ul><h4><strong>How are these random parameters estimated then?</strong></h4><p>It is entirely left to the modellers on how to arrive at the initial estimation of these parameters. Given that most models begin with an initial set of assumptions, epidemiologists and professional modellers tend to use a technique called <strong>Maximum Likelyhood Estimation — MLE</strong>, for picking the initial sets of parameters so that they are not totally random.</p><p>MLE is a technique of picking the right assumptions (parameters in the world of modelling) that will generate projections closest to the actual observations.</p><blockquote>In other words, lets assume that there is an architect. He really likes his neighbours house. He wants to build something similar for himself. First, he wants to build a 3D model to ensure that he gets the planning right to build the house. The architect first creates a floor plan with measurements and feeds the plan to his modelling software to generate the 3D model. <strong>In the MLE terms, the plan with measurements are set of numbers that generates a house and that is closest to the architect’s neighbours house. The Maximum Likelihood Estimation is that set of measurement parameters in a plan that closely resembles the observation (neighbours house).</strong></blockquote><p>Now that we understand how MLE is used to arrive at assumptions, it will be interesting to understand how this works in the case of estimating parameters of Corona Virus epidemic modelling:</p><ul><li>Since Corona virus is new and has no prior occurrence (unlike the reality of neighbours house), the modellers will use the distribution of past epidemics, preferably from the same family (like SARS or MERS or even HIV), as the initial distribution.</li><li>Based on available little data on the infection dynamics, coming from the countries and WHO, they will further adjust whether SARS is the right distribution or anything else. Once more data comes through, then they will get actual distribution to work with to pick the assumption parameters.</li></ul><blockquote>It is worthwhile to note that modellers and epidemiologists always lag the disease spread when a new epidemic occurs. And as the epidemic spread dynamics changes per population type (disease dynamics varies from country to country), any general or even adapted model is likely to underestimate or overestimate the actual occurrence.</blockquote><p>So the early projections from these models always needs to be taken with a pinch of salt, especially for new epidemics. These projections are a good tool for policy makers to plan public health response strategies. These are not meant for general public consumption as they are likely to be induce widespread panic, which has unfortunately been the case with what we have seen with couple of models projecting doomsday projections for US, UK, India etc.</p><h3><strong>Are there alternatives to MLE?</strong></h3><p>Naturally, the world of modelling and statistics is not standing still. Instead of MLE of the Frequentist approach, Markov Chain based and Approximate Bayesian Computation of Bayesian world are increasingly used to estimate the parameters.</p><p>Woah, I know I may have thrown the book of applied statistics at you! Let me quickly declutter this:</p><blockquote><strong>Frequency vs Bayesian</strong>:</blockquote><blockquote>A Frequentist approach is basically, to use the same architect and his neighbours house example, how confident that the architect is about the probability that the measurements given in the plan will result in the house similar to neighbours house. The confidence or lack of it of the architect stems from the past experience of the outputs of his plan (not related to this house, but generally speaking) and the similarity of the output (3D projection) to the target output (actual house measurements).</blockquote><blockquote>A Bayesian approach focus is on the outcome, not on the output. To again use the architect and house example, a Bayesian approach means the following: should the architect go ahead and commission the contractors to build the house based on his measurements?</blockquote><h4>Markov Chain Approach</h4><p>Unlike the MLE, where the focus is on finding the estimators that generate the assumption parameters that closely reflect the reality, the Markov Chain approach is different. The focus is on generating assumption parameters based on prior beliefs and generate a random sample with an outcome of closely matching the random sample with the reality — called as posterior belief. Then continue to adjust the assumption parameters in light of the new belief (posterior belief) and generate samples. Continue this process till the outcome of generating samples (posterior belief) closely matches the reality (evidence). Markov Chain, essentially a state transition diagram between S → E → I → R, is an algorithm used to generate samples by performing random transitions from S to E (proportion of Susceptible people come in contact with Exposed), from E to I (proportion of Exposed People not having immunity and hence getting Infected, cue conditional probability), from I to R (from Infected to Recovered). This output is then compared with the evidence, if they do not match, new probabilities and assumption parameters governing S → E →I → R gets generated. Monte Carlo simulation is used to generate the random parameters, plucking from any of the known distributions like Poisson, Weibull etc. The process is run till the model matches the observations or acceptably close to observation, then the simulation is run ahead of time, as a forecast, by letting transitions of S → E → I → R run governed by those assumption parameters that resulted in the closest match between model output and available evidence .</p><h4>Assumptions governing Markov Chain approach:</h4><ul><li>As with SEIR model, the population mixing is initially assumed. However, as the parameters are continuously learnt, the model keeps up with the even dynamic populations.</li><li>Prior knowledge is always assumed. For example, prior knowledge of spread of Corona virus in China informing US modelling.</li><li>It is always assumed that there exists significant body of evidence in the current model application context (figures of Corona virus spread as recorded in US for projecting US nos) is essential. Otherwise, Markov Chain approach will continue to use China as evidence and project for US. The model therefore needs continuous updates.</li><li>And another assumption is that location of the individual is a constant and the time effect post contact is not assumed (basically, the exposed person gets infected immediately, not with a delay). <strong>This is quite important given that now we all know how hard Social Distancing is and we have learnt of new terms like Super Spreader, and how spatial (space and movement related) dynamics are critical to project the spread of virus.</strong></li></ul><blockquote>More importantly, Markov Chain based projection is not advisable for scenarios where evidence is thin, in the geography of application. I.e. US evidence is thin, so projections will have to wait till more data is generated. Projecting Texas spread based on NY is also tenuous as the geographical context is different.</blockquote><h4>Are there approaches that realistically assumes individuals as not static, but as dynamic in location and time?</h4><p>In cases where evidence data is hard to get by and it becomes computationally expensive to model transition of individuals across states in dimensions of both space and time, <strong>Approximate Bayesian Computation (ABC) </strong>approach is used.</p><p>Instead of the four states of SEIR model, the ABC approach uses 8 states {S,E,I,R,S*,E*,I*,R*}, where S,E,I,R are the count of individuals within each compartment and the (*) indicates the transition of individuals within a compartment given a time and location. The transition probability of an individual in one compartment transitioning in location and time within and outside is derived from an exponential distribution. The simulation is run as long as the counts projected in each compartments are acceptably closer to the reality.</p><p>Some of the assumptions of this model are:</p><ul><li>The intensity of epidemic per location is assumed and given the computational demands, the location is generally a large area (like NYC for example)</li><li>The intentity of epidemic per location also changes over time. The rationale is that over the period of time, the infection may be peaking or may have peaked or abated.</li><li>The time spent by individuals as they transition is derived from similar epidemics (from China for example) or assumptions made on early data (for Dallas projection, from NYC)</li><li>The transition probabilities from stage (Exposed) to another (Infected) is also assumed</li></ul><blockquote>One advantage of ABC over Markov Chain based approach is that ABC do not rely on prior knowledge. Or simply put, past is not assumed to influence the future. The current data alone is sufficient enough to predict the future.</blockquote><h3><strong>Success stories of epidemic models in Policy and in Business</strong></h3><ul><li>It is not all doom and gloom. These models have been used extensively by public health professions to manage and even measure the effectiveness of policies like universal vaccination polices, polio eradication drive etc.</li><li>These models are also used by Epidemiologists to define health response protocols and government policies around transportation, border control especially during outbreaks such as SARS, Ebola, MERS, ZIKA etc</li><li>Since these models represent the spread of virus amongst the human beings, the models in certain ways also simulate the interaction dynamics between human beings. Some of these models, especially the Reproductive Number, R0, is being used by businesses and marketeers especially, to test the “virality” of their messages and campaigns amongst the target segments.</li><li>The spatial properties of the model can be inversely learnt from the outcomes, by applying backward propagation techniques to discover influential parameters in terms of location and time. Linked to individual interaction, they can be used to discover newer market segments and behavioural patterns</li><li>Increasingly similar models are being looked at to study the spread of viruses in the cyber world. As the cyber-viruses operate in similar dynamics, organisations are looking at studying the transition dynamics to refine and formulate effective cyber security response strategies focused on containing the spread, flatten the curve and then eliminate the virus.</li></ul><p>In the end, models are just models, cloaked by numerous assumptions. They are neither meant nor intended for doomsday statistics triggering panic amongst general populace.They are at best a guide and a tool in the hands of policy and decision makers to shape policy, define strategy, and test the effectiveness of policy and strategic interventions.</p><p>Here’s to a pandemic free 2020!</p><p><strong>References</strong></p><ul><li>Hans Nesse — Global Health — SEIR Model <a href="http://www.public.asu.edu/~hnesse/classes/seir.html?Beta=0.9&amp;Gamma=0.2&amp;Sigma=0.5&amp;Mu=0&amp;Nu=0&amp;initialS=10&amp;initialE=1&amp;initialI=0&amp;initialR=0&amp;iters=15">http://www.public.asu.edu/~hnesse/classes/seir.html?Beta=0.9&amp;Gamma=0.2&amp;Sigma=0.5&amp;Mu=0&amp;Nu=0&amp;initialS=10&amp;initialE=1&amp;initialI=0&amp;initialR=0&amp;iters=15</a></li><li>van den Driessche P. Reproduction numbers of infectious disease models. <em>Infect Dis Model</em>. 2017;2(3):288–303. Published 2017 Jun 29. doi:10.1016/j.idm.2017.06.002</li><li>Yaesoubi R, Cohen T. Generalized Markov Models of Infectious Disease Spread: A Novel Framework for Developing Dynamic Health Policies. <em>Eur J Oper Res</em>. 2011;215(3):679–687. doi:10.1016/j.ejor.2011.07.016</li><li>Bayes Theorem <a href="https://towardsdatascience.com/ever-wondered-how-epidemiologists-simulate-covid-19-deaths-32e3aca7531d">https://towardsdatascience.com/ever-wondered-how-epidemiologists-simulate-covid-19-deaths-32e3aca7531d</a></li><li>Ghassan Hamra, Richard MacLehose, David Richardson, Markov Chain Monte Carlo: an introduction for epidemiologists, <em>International Journal of Epidemiology</em>, Volume 42, Issue 2, April 2013, Pages 627–634, <a href="https://doi.org/10.1093/ije/dyt043">https://doi.org/10.1093/ije/dyt043</a></li><li>Hamra G, MacLehose R, Richardson D. Markov chain Monte Carlo: an introduction for epidemiologists. <em>Int J Epidemiol</em>. 2013;42(2):627–634. doi:10.1093/ije/dyt043</li><li>Brown GD, Porter AT, Oleson JJ, Hinman JA. Approximate Bayesian computation for spatial SEIR(S) epidemic models. <em>Spat Spatiotemporal Epidemiol</em>. 2018;24:27–37. doi:10.1016/j.sste.2017.11.001</li><li>Stochastic compartmental modelling with Markov Chain Monte Carlo <a href="http://sherrytowers.com/2016/01/02/stochastic-compartmental-modelling-with-markov-chain-monte-carlo/">http://sherrytowers.com/2016/01/02/stochastic-compartmental-modelling-with-markov-chain-monte-carlo/</a></li><li>A Gentle Introduction to Maximum Likelihood Estimation <a href="https://towardsdatascience.com/a-gentle-introduction-to-maximum-likelihood-estimation-9fbff27ea12f">https://towardsdatascience.com/a-gentle-introduction-to-maximum-likelihood-estimation-9fbff27ea12f</a></li><li>Cole SR, Chu H, Greenland S. Maximum likelihood, profile likelihood, and penalized likelihood: a primer. <em>Am J Epidemiol</em>. 2014;179(2):252–260. doi:10.1093/aje/kwt245</li><li>Sunnåker M, Busetto AG, Numminen E, Corander J, Foll M, et al. (2013) Approximate Bayesian Computation. PLOS Computational Biology 9(1): e1002803.<a href="https://doi.org/10.1371/journal.pcbi.1002803">https://doi.org/10.1371/journal.pcbi.1002803</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1e6bf974bc47" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[From a Telco to a Techco: Review of the year gone by and what’s in store for 2020]]></title>
            <link>https://medium.com/@narendransivakumar1/from-telco-to-a-techco-review-of-the-year-gone-by-and-whats-in-store-for-2020-fc8deb08cea5?source=rss-216b7c0837c2------2</link>
            <guid isPermaLink="false">https://medium.com/p/fc8deb08cea5</guid>
            <category><![CDATA[internet-of-things]]></category>
            <category><![CDATA[strategy]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[telco]]></category>
            <category><![CDATA[digital-transformation]]></category>
            <dc:creator><![CDATA[Narendran Sivakumar]]></dc:creator>
            <pubDate>Mon, 03 Feb 2020 22:42:39 GMT</pubDate>
            <atom:updated>2020-02-14T07:55:52.254Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>Strategic divergences are now real</strong></p><p>Gone are the days when the telco industry used to think alike – at least broader principles. 2019 saw a stark strategic divergence with US, Indian and to some degree Australian telcos beginning the transformation towards a digital services organisation. Connectivity is just a product line. The focus is to encourage customer to subscribe to a stack of digital services on top of connectivity. However in the UK and Europe market, the priority was and continues to be on cost and optimisation. Convergence was and continues to be still a buzz word. The rest the world, barring Africa, has moved beyond convergence.</p><p><strong>The squeeze on capital</strong></p><p>The cost of capital has risen for telcos, more so in the UK and Europe region, triggering a spate of asset unwinding – cue Vodafone and TDC a la NetCo and ServCo. Balance sheets are stressed throughout and as Liberty Global and AT&amp;T found out, investors are now becoming activistic.</p><p><strong>Bold Vision was and will continue to be rewarded</strong></p><p>The only telcos that starred last year were the ones with bold vision and willing to bet big. No vacillations. Not caught up in their own connectivity subscriptions trap. And more importantly, are willing to show to their customers and the world that telcos can go beyond connectivity into agents of digital experience and digital transformation. They are Comcast, Verizon, Elisa, Airtel, KPN and Reliance Jio.</p><p><strong>The age of 5G</strong></p><p>5G finally crawled out of the network organisations of telcos into collective consciousness of the commentariat – the politicians, the spies, the economists, the journalists, the technologists, the regulators, and at last the telcos. Whilst politicians were busy battling the Huawei threat, the telcos were talking up the benefits of 5G – the speed, the reliability, the low latency – and the wondrous impact all of them have on all manner of industries and the consumers. If the past is any guide, and if the current price plans are any indicative, telcos seam to deal with 5G the way they dealt with 4G albeit the much touted addition to convergence. That is a sign of telcos commoditising themselves – an untenable strategy given their capital strategy and cost structures.</p><p><strong>B2B, rather than the B2C, will be the proverbial magic wand</strong></p><p>Telcos have not yet managed to unshackle themselves from the connectivity trap. The full potential of 5G and complementary technologies like fibre will only be realised by creating propositions that will transform industries and the societies at large – just like how electricity and steam engine transformed the world. Consumers will be an added beneficiary. The biggest beneficiaries will be industries. This requires telcos to create compelling value propositions for the industries, propositions that are industry specific, deeply integrated to networks, truly disruptive and unlocking the full potential of 5G.</p><p><strong>Regulators 2.0</strong></p><p>Given that some of these B2B propositions will have to be deeply integrated to telco core networks, telcos have a head start in creating new value streams. Failing which, they risk inviting the regulators to open up access to core network for the over the top players or tech companies to integrate their products or propositions, similar to how legacy telcos were forced to open up last mile access to all players through local loop unbundling, only that this new regulator led putsch will be far more painful.</p><p><strong>Advanced AI + Automation = Cognitive Telco</strong></p><p>Telcos will move few steps to embrace an enterprise wide AI couple with automation. AI and automation will converge given the combined benefits of both. AI will grow out of data scientists into an engineering discipline whereby sophisticated models called as strong AI will be performing at the levels of a human being and will increasingly invoke automated actions without any human intervention. 2020 will show whether telcos embrace it or avoid making a bold bet on becoming a cognitive telco powered by AI.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fc8deb08cea5" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[5G and Net Neutrality – Policy Recommendations for Telcos and Regulators]]></title>
            <link>https://medium.com/@narendransivakumar1/5g-and-net-neutrality-8b05148c6802?source=rss-216b7c0837c2------2</link>
            <guid isPermaLink="false">https://medium.com/p/8b05148c6802</guid>
            <category><![CDATA[customer-experience]]></category>
            <category><![CDATA[net-neutrality]]></category>
            <category><![CDATA[regulatory-compliance]]></category>
            <category><![CDATA[5g]]></category>
            <category><![CDATA[business-strategy]]></category>
            <dc:creator><![CDATA[Narendran Sivakumar]]></dc:creator>
            <pubDate>Wed, 27 Nov 2019 15:50:26 GMT</pubDate>
            <atom:updated>2019-11-28T10:32:45.222Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>5G and Net Neutrality – Policy Recommendations for Telcos and Regulators</strong></p><p>As 5G investments soar, the telcom industry is stuck in a bit of a quandary. I have touched upon this in the past posts about business models, monetisation strategies and structural limitations. Let’s be optimistic and assume that telcos will somehow manage to tide over all of this.</p><blockquote>Will they then raise a glass of champagne and declare victory, or is there something that will still bother them – something stuck within that is difficult to spit it out and rankles inside, petulant and irascible?</blockquote><p>I am talking about how uneasy the shining armour of 5G – network slicing driven by user services – sits with the net neutrality regulations. To clarify, this is not a big problem in US as it is in EU. FCC’s liberal interpretation of net neutrality has given a leg up to telcos in the Big Apple to differentiate services on price and quality. However for Europe, it is a different story. And it is safe to assume that global regulators appear, more or less, inclined to embrace the European stance than the American one. India may go either way, and it will be political decision there.</p><p><strong>A brief about Net Neutrality</strong></p><p>Net neutrality, in plain English, is about ensuring that internet remains open for all and that the internet service providers do not get to control and differentiate services based on who delivers the service and how valuable the service is to both the consumer and over the top service provider. Think of internet as a motorway and let’s assume there is an enterprise (operator) that lays and governs the motorway. Net neutrality means the enterprise can only charge or even forgo tolls, but cannot charge tolls based on the colour of car, the number of passengers, the profile of passengers etc.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/720/1*Qrl73W5FDHSNpdKl5h2wGw@2x.jpeg" /></figure><p>Meter gauge vs Highspeed Maglev, Image Credit: Scott Barbour/Getty Images</p><h4><strong>Net neutrality in European context</strong></h4><p>BEREC, abbreviated from Body of European Regulators for Electronic Communications, is a supranational regulator that sets, amongst other things, telecom regulations within Europe.</p><p>BEREC, in April 2016, came up with a landmark regulation – Telecom Single Market regulation, imaginatively dubbed as ‘Save the Internet’ regulation. The regulation, unsurprisingly set up saving the internet. It declared internet truly open and no one can act as a gatekeeper. More importantly, it had far reaching implications for telcos as they can differentiate services by quality or price only unless the same is available in uncontrolled form for the rest. Basically, you can differentiate but you have make what you differentiate available to all, yes but no!</p><p>Quoting, verbatim, the crux of the regulation from the BERRC <a href="https://ec.europa.eu/newsroom/dae/document.cfm?doc_id=58910">website</a>, stripped of legal gobbledygook:</p><ul><li>Article 3(1) of the Regulation sets out the principle that end-users of internet access services have the right to access and distribute the information, content, applications and services of their choice.</li><li>Article 3(2) provides that agreements between providers of internet access services and end-users must not limit the rights of end-users set out in Article 3(1).</li><li>Article3(3) bans blocking, throttling and discrimination between content, applications and services, subject to certain limited exceptions. Traffic management, to optimise the quality of the services transmitted, is possible as long as it is reasonable. The third sub-paragraph lays down restrictive exceptions from the obligation not to engage in traffic management measures (and not to block, slow down, alter, restrict, interfere with, degrade or discriminate between specific content, applications or services) in order to either comply with legislation, or to preserve the security of the networks or to prevent exceptional/temporary congestion management.</li><li>Article 3(4) provides that any traffic management measures have to comply with the principles of necessity and proportionality as regards the processing of personal data and to the relevant Union framework in this area.</li><li>Article 3(5) makes clear that services (other than internet access services) optimised for specific content, applications or services (sometimes called ‘specialised services’) can be offered under certain conditions, including that they are not detrimental to the quality of general internet access services.</li></ul><p>The fundamental guardrails of net neutrality are:</p><ul><li>Open for all</li><li>Non discriminatory for end users arising from any B2B or internal arrangement</li><li>Non discrimination of network traffic</li><li>Reasonable, Necessary and Proportional, should management of network traffic arise due to extenuating circumstances like congestion</li><li>Fairness whereby traffic optimised for specific application service is not at the cost of other internet traffic</li></ul><p><strong>5G and issues with Net Neutrality</strong></p><p><strong><em>5G New Radio, Massive Mimo and Beamforming</em></strong></p><p>5G introduces a new antenna technology whereby the antenna is no longer a fixed monolithic physical unit, but an 8x8 array of individual movable antennas. It truly comes to play in congestion management or providing extra throughout boost to a local area. Consider a busy street, during peak hours flooded with people, and putting immense pressure on networks. In the pre 5G world, the network experience provided by multiple towers in the vicinity of busy street will immediately get into congestion control mode by degrading throughput for all users in the vicinity to accommodate for demand spike. It is proportional, necessary and reasonable. What 5G does is enable a feature called beamforming where multiple individual antennas reorient themselves away from the planned orientation towards the busy street, the source of new traffic demand, and create a narrow virtual radio channel to meet the increasing demand. It is great from customers in the busy street, they get to enjoy high speed great service. However, what is given to them is actually taken away from someone who is now underserved because all that attention has shifted to the busy street. This calls to question the fairness principle of net neutrality. Who is to decide what is fair – serving the folks on the busy street or serving folks in an area where the demand is low. The traffic is optimised at the cost of someone’s general internet access.</p><p><strong><em>5G Network Slicing</em></strong></p><p>5G architecture has three predefined traffic types using which the the 5G system can differentiate and steer the traffic. Telcos, on top of that, can define their own slides. To quote, verbatim, from 5GS release specification <a href="https://www.3gpp.org/release-15">v15</a>:</p><blockquote>The characteristics of each slice are defined in terms of QoS, bit rate, latency, etc. For a given slice, these characteristics are either predefined in the 3GPP Standard or are operator-defined. There are three types of predefined slices: type 1 is dedicated to the support of eMBB, type 2 is for URLLC and type 3 is for MIoT support. These predefinitions allow inter-PLMN operation with reduced coordination effort between operators. As for the operator-defined slices, they enable more differentiation among network slices.</blockquote><p>There is no regulatory clarity on this. One could argue that network slice is a form of optimised network for services but on the other hand, it is hard to prove that network slices do not discriminate general network traffic. They indeed do. That’s how network slicing works where the user plane function within 5G system discriminates user traffic based on network slice identifiers and each slice of traffic from thereon is assigned predefined QoS Class Identifiers (QCI). The QCIs drive predetermined traffic management parameters in terms of throughout, latency, jitter, traffic prioritisation etc. This does not sit well with broader thrust of net neutrality regulations (3.3 and 3.5 of BEREC Regulations)</p><p><strong><em>Public policy and regulations need to keep up with time</em></strong></p><p>The need for regulations was well grounded. It all started off with some telcos in EU restricting OTT VoIP calls. In the past, Telcos did shoot themselves roundly in the foot. It is preposterous to think now that telcos thought about monetising Whatsapp / FaceTime / Skype calls. It is no surprise that they find themselves in a commodity business conundrum. Having said that, what 5G enables, in the parlance of railways is not a standard meter gauge railway, but a high speed maglev railway. To expect telcos to treat high speed maglev railway in the manner similar to meter gauge is akin to killing the industry by ruinous policy. The capex that is required to build 5G networks is too large for telcos to charge the services like a commodity. At the very least, some leeway must be provided to charge high speed services differentially. However, from a regulatory point of view, it begs a question as to when a differential treatment of service becomes discriminatory, anti-customer and predatory. Telcos, themselves, are struggling to answer that with contours of business models that will underpin differential treatment and monetisation, beyond the three broad classification. I do not think regulators will mind that, but will insist on proportionality amongst the three classifications, which is not a great outcome for telcos. For telcos to eke out a positive outcome, it is incumbent upon them to define business models and differential treatment of services to granular detail. And more importantly, run multiple scenarios of these at play in simulated environment that is reflective of real world, not some unrealistic scenarios. Even more important, telcos should commission a broad econometric study on societal, customer and business impact. There is too much at stake to be left to regulators and layers alone.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8b05148c6802" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[5G : Finding Telco’s True North]]></title>
            <link>https://medium.com/@narendransivakumar1/5g-finding-telcos-true-north-42d18392e955?source=rss-216b7c0837c2------2</link>
            <guid isPermaLink="false">https://medium.com/p/42d18392e955</guid>
            <category><![CDATA[innovation]]></category>
            <category><![CDATA[5g]]></category>
            <category><![CDATA[5g-wireless]]></category>
            <category><![CDATA[ar]]></category>
            <category><![CDATA[business-strategy]]></category>
            <dc:creator><![CDATA[Narendran Sivakumar]]></dc:creator>
            <pubDate>Mon, 28 Oct 2019 17:34:01 GMT</pubDate>
            <atom:updated>2019-10-28T22:44:21.253Z</atom:updated>
            <content:encoded><![CDATA[<h3>5G Strategy : Finding Telco’s True North</h3><p>In the Telco “change &amp; transformation” series, I have covered quite a bit about the key drivers for Telco transformations and why the industry is in need for new inspirations and motifs. Change <a href="https://medium.com/@narendransivakumar1/change-series-1-netco-and-servco-old-wine-in-a-new-bottle-or-fortifying-against-hedge-fund-led-244542cea17c">series 1</a> dwelt into emergent business models for telco whereas <a href="https://medium.com/@narendransivakumar1/change-series-2-market-disruption-and-the-need-for-telcos-to-establish-new-symbols-and-new-aef7d6f61f7a">series 2</a> focused on a structural and behavioral impediments that telcos have to find a way to either surmount or avoid.</p><h4><strong>A brief note on 5G</strong></h4><p>I do not want to use this post to explain away 5G. That is a project of its own. I do intend to cover 5G in detail as and when the release 16 is available for consumption. To provide a very short primer, 5G is all about massive speed (capacity), massive connectivity (devices) and ultra low latency (high reliability). 5G can be summed up by this oft used graphic:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*7JUN81i4gfuiIaEc.jpg" /><figcaption>What is 5G</figcaption></figure><h4>5G investments</h4><p>According to a GSMA report, telcos will invest upwards of $1 Trillion dollars on 5G and allied services, yep — that $$ value is not a typo! The per annum capex translates to rougly 17% of total operator revenue. It is indeed a big deal. A very big deal with lots at stake. So far, most of the investments are towards a phased up-gradation of networks, from 4G (LTE-A) to 5G NR (New Radio — Non Standalone) and eventually to 5G NR (New Radio — Standalone). Without getting too much into the nitty-gritty of spectral frequencies and capacity, except to note that higher the frequency, higher is the capacity to carry data (voice &amp; data). Most telcos are currently moving from 4G to 5G NR NSA as a first step of upgrade, which essentially means using new radio equipment to latch on to higher spectral frequencies than they are currently latching on to in the 4G world whilst still using the control logic (control plane in telco speak) of 4G world. Essentially, visualise a human being’s brain (control plane) with all attendant fallibleness powered by a jet engine (5G spectrum) instead of a lumbering diesel locomotive (4G spectrum). What this means for the consumers is an added boost when it comes to speed, by an order of 10x to 20x, all things being equal. Most of the advanced 5G features like timeslicing, programmable traffic flows, and what makes up the control plane functions (brain) will be part of the 5G NR Standalone rollout, which is atleast 3–5 years away. So when you hear hype of 5G, what you see (what is 5G- triangle picture) is not what you get, atleast for now!</p><h4>5G Products and Monetisation Strategy</h4><p><strong>Be that as it may,</strong> it is the right time for the industry to think deep and formulate a holistic and a sustainable 5G strategy from a products and monetisation viewpoint. As change series 1 and 2 indicate, if telcos where to sell 5G the way they have sold 4G, then they are on the fast route towards commoditisation. What telcos need is a calibrated strategy that points towards True North: Confluence of New Demand, New Services, New Business Models.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/733/1*WsO_Ik6_g646eZbut_QfEQ.png" /><figcaption>5G Success Critera</figcaption></figure><h4>AR, VR, ESports, online multiplayer games: 5G Usecases and Demand</h4><p>The “trillion dollar” question that is confronting the industry is where will the new demand come from. More than one CEOs have been quoted in the industry reports wondering the same. With 5G — and as it was with 4G, one gets a familiar sense that industry has embarked on a ‘leap before you think’ strategy, in part (or largely, depending on where you sit) due to regulatory and competitive pressures. Now having leaped, the industry is thinking about creating new sources of demand that will consume the capacity that 5G unleashes. Most operator strategies are primarily focused on trialing 5G. Whichever way the demand story pans out, at the very least, I fully expect operators to use 5G to offset network capacity constraints in the near short term, as a relief measure. As for the usecases, the industry is split. There are vague assertions but I am yet to see a convincing usecase that can be conceived and led solely by operators. It is all about partnerships. When it comes to AR and VR, the consumer technology that powers them will shape the industry. Right now, the industry (Apple, Google, Facebook etc) is working on building an easily wearable, light weight, and mainstream VR glasses/headset with a wide Field of Vision (jargon for providing a wider view of the real/projected world). A 2019 marketwatch report in fact is forecasting possible adoption in the medical, retail, manufacturing and other process based industries. So the demand is there, but primarily oriented towards B2B segments rather than the consumer end of the spectrum.</p><p>That leaves online gaming — esports, multiplayer of that sorts. If there is one industry that is growing at a fast clip — that is gaming, especially the online multiplayer games like Fortnite, PUBG, Call of Duty etc. Netflix views Fortnite as its chief competition.</p><p>Gaming, the online multiplayer variety, is actually the one that is generating voracious demand. I was told that a particular release of Fortnite caused so much congestion that it briefly threatened to bring down one of the operators network. It is to no wonder, the biggies are entering the cloud gaming market. Google with Stadia, Apple with Arcade etc are few casing examples. Games are predominantly confined to Consoles and PCs (yes!) and are only now expanding to the mobile ecosystems (iOS and Android). The experience so far though good, to quote a <a href="https://www.forbes.com/sites/erikkain/2018/10/12/heres-every-android-device-compatible-with-fortnite-battle-royale/">Forbes contributor</a>, is not great:</p><blockquote>Epic has done a fine job with the mobile version of this (Fortnite) game, but there’s just no denying that a controller or mouse-and-keyboard setup is better than touch controls when it comes to shooting and building</blockquote><p>Perhaps, the industry with investments from Google, Apple and the rest, will start to push a mobile centric experience. Gaming, especially when it is primed for mobile experience, continues to be a good bet for new sources of consumer demand. Till such point, serious gaming although attractive, will always be a console and PC centric experience, hence more hungry for bandwidth of the wrong kind (Fiber, not 5G).</p><p>So to conclude, the demand is likely to be B2B for AR/VR and B2C for Online Gaming, at-least in the short to mid term. In the long run, to quote John Maynard Keynes, everyone is dead!</p><h4>5G Services: B2B is the new frontier of growth</h4><p>So we discovered that the much vaunted AR and VR is still in its infancy from a consumer view point and there is a strong application of it the in the B2B world. Gaming, as we discovered, is still confined to consoles. There is an emergent category of games, online multi player or mobile centric ones — which are a potential segment for 5G leverage. 5G allows telcos to assure experience for specific services in the wireless world like never before. The focus of that experience so far has been on the consumer. I think, on the contrary, the focus of that experience should be the B2B consumer, not a B2C consumer. Yes, I am advocating a reorientation of focus towards B2B than B2C in the short to mid term. There is lies the opportunities for 5G targeted at B2B. I discussed in <a href="https://medium.com/@narendransivakumar1/change-series-2-market-disruption-and-the-need-for-telcos-to-establish-new-symbols-and-new-aef7d6f61f7a?source=---------2------------------">this post</a> as to why B2C growth requires new symbol and new metaphor from telcos beyond just unlimited plans.</p><p>Traditionally, in the B2B world, capex funded transformations are the traditional “get of jail” card employed to modernize and fuel growth. In a gloomy macroeconomic environment, the growth in capex cycles are tepid at best and, worse, is predicted to remain tepid for a next two to three years as per <a href="https://www.spratings.com/documents/20184/1481001/Global+Corporate+Capex+Study+2019+Curbed+Enthusiasm/086812c1-e10f-e2e7-2be0-89c7e34d8bc7">S&amp;P global capex survey</a>. Capex tends to fuel productivity and eke of efficiencies. In absence of that, B2B enterprises, across sectors, will resort to eking out efficiencies within the current regimen. That will mean, leveraging technology — sensors, devices, IOT platforms, edge analytics, edge services — to drive productivity and efficiency improvements. It is here we see a huge 5G potential.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/781/1*Ewcm0E1o3JxgxuylkpsR5Q.png" /><figcaption><a href="https://www.spratings.com/documents/20184/1481001/Global+Corporate+Capex+Study+2019+Curbed+Enthusiasm/086812c1-e10f-e2e7-2be0-89c7e34d8bc7">Source: S&amp;P Global Market Intelligence, S&amp;P Global Ratings. Universe is Global Capex 2000</a></figcaption></figure><p>Process based manufacturing industries, utilities and energy industries present a significant potential for 5G Services. <strong><em>Some of the 5G services in the manufacturing area include realtime machine analytics and remote control, remote operations leveraging connected IOT devices and sensors, AR/VR led training for better employee productivity, AR/VR led assisted operations — all fulfilled by dedicated 5G slices powered by a private 5G network for large establishments or a dedicated Multi-access edge network integrated to the enterprise connectivity backbone.</em></strong></p><p>In the B2B segment, 5G elevates Telco positioning from connectivity services provider to a business value provider who can have a material impact on firms Capex/Opex and returns. The future for Telcos in the B2B segment looks very promising. The key is to build the right product proposition, conceive right service experience, and facilitate the right business outcomes for their enterprise customers — all wrapped up in a surgical Go-to-Market strategy.</p><p>I will conclude this post by stating that a successful 5G exploitation requires Telcos orchestrate new demand fulfilled by new services and facilitated by new business models. In the next post, I intend to elaborate upon the business models and the art of possible for Telcos.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=42d18392e955" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Change Series #2: Market Disruption and the need for Telcos to establish new symbols and new…]]></title>
            <link>https://medium.com/@narendransivakumar1/change-series-2-market-disruption-and-the-need-for-telcos-to-establish-new-symbols-and-new-aef7d6f61f7a?source=rss-216b7c0837c2------2</link>
            <guid isPermaLink="false">https://medium.com/p/aef7d6f61f7a</guid>
            <category><![CDATA[5g]]></category>
            <category><![CDATA[market-disruption]]></category>
            <category><![CDATA[business-strategy]]></category>
            <category><![CDATA[marketing-strategies]]></category>
            <category><![CDATA[telecommunication]]></category>
            <dc:creator><![CDATA[Narendran Sivakumar]]></dc:creator>
            <pubDate>Sat, 12 Oct 2019 17:04:32 GMT</pubDate>
            <atom:updated>2019-10-13T14:36:18.794Z</atom:updated>
            <content:encoded><![CDATA[<h3>Change Series #2: Market Disruption and the need for Telcos to establish new symbols and new metaphors</h3><p>That Telcos will have to change is beyond obvious. There is a broad industry consensus on the need for change. Across the horizon, the contours of change that is emerging requires careful consideration. This post is part of a larger series which aims to present a perspective on the need and vehicles adopted for change and transformation. <a href="https://medium.com/@narendransivakumar1/change-series-1-netco-and-servco-old-wine-in-a-new-bottle-or-fortifying-against-hedge-fund-led-244542cea17c">Series 1</a> took a provocative view that the ultimate decision makers — the shareholders — are running out of patience and if things are left as it is, then there is a good chance of the industry being split across NetCo and ServCo lines. I present reasons as to why — to cut a long story short- it is about returns, returns and returns.</p><h3>5G is the new buzz</h3><p>In this post, I direct the gaze towards the contours of change we are seeing in the market. I’ll be facetious and facile if I were to opine that Telcos are doing nothing against the shift towards the default Utility business model. The reality is far from that. With all that talk (and some hype too) of 5G, there is already a buzz in the market.</p><h3>Telcos are the only one investing</h3><p>In fact, in a gloomy macro-economic environment punctuated with recessionary fears in US, UK and EU, trade rifts between major trading blocs, geopolitical flareups in the middle east, central banks are turning dovish in order to boost consumption and consequently boost the industry. Most industrial sectors are already feeling the brunt of macroeconomic slowdown and are in a belt-tightening spree, slashing capex and optimizing opex. There is one industry — Telecom — which is singing a different tune or swimming in the opposite direction — pick your own metaphor. All thanks to (or forced under regulatory overhang) 5G and Fiber— take your pick again.</p><h3>Convergence and unlimited plans as product strategy</h3><p>The wheels of change are already set into motion anticipating 5G. The symbols and metaphor of change are data, data and more data. It is about how do I (as in Telcos) can get Consumers to develop a voracious appetite for data eventually they will end up consuming the tremendous capacity that will be unleashed onto the market because of 5G and Fiber. Telcos do realize that to generate returns, they have to get customers used to consuming data — data in vast vast quantities. Perhaps, drawing inspiration from Noble prize wining ‘Nudge theory’ on behavioral economics, Telcos have started to launch a slew of “unlimited” plans. And “Convergence” is the new game in town for Telcos who all have duo/triple/quadpay propositions.</p><blockquote>Although the intentions are right, the symbols and metaphors are the same old-old. What is needed are the new symbols and metaphors. Data, Data and more Data is not an apt metaphor of growth and disruption. Convergence as a symbol of change, although it has a ring to it, is perhaps a good starting point but not the end.</blockquote><p>I will try to unpack as to why it is so. Let me, for simplicity sake, conflate data unlimited as a both symbol and a metaphor.Lets take the UK market — partly I live there and more importantly, because the market is under continuous, active and penetrating regulatory oversight which produces excellent studies. <a href="http://telecoms.com/500197/bt-starts-making-noise-about-convergence-with-halo-launch/">BT </a>has launched a new convergence service — called <a href="http://telecoms.com/500197/bt-starts-making-noise-about-convergence-with-halo-launch/">Halo</a>.The symbol is unlimited data, and the metaphor is the consumption of voracious amount of data regardless of underlying access medium — wireless or fixed. There are so many operators which have launched or launching something similar globally. Just search using the keywords “unlimited plans”, and the story will unravel itself.</p><p>The problem with data unlimited as a symbol and metaphor is that the market is structurally inhibited to embrace it in the current form. The reason is because the market is structurally positioned to monetize customers data consumption — higher the data, higher the price. Structures are created, on the mobile services end, to enforce this by creating bundles with data caps and automatic demotions to ultra low speed tiers. The fiber product continues to depend on advancements in technologies across TV (4K, UHD, 8K etc), home devices etc and OTT products (Netflix, Gaming etc) to create demand for data.</p><h3>Why behaviours need to be changed for market disruption to take root</h3><p>For simplicity sake, let us focus primarily on Mobile as that is where greater investment thrust lies. UK market, because of Telco created structures, is a capacity surplus market. A market becomes capacity surplus when its network utilization does not exceed more than 50% to 60%.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/575/1*0dPX92CO8oXy0ohg7kiKjA.png" /><figcaption>Enders report on capacity utilization</figcaption></figure><p>In layman terms, it is a market where people consume less data and they are attuned to that behavior. Some consumers prefer more speed with caps on data usage where as some consumers prefer more data with a cap on speed.</p><blockquote>This is exemplified by a CCS Insights Survey of UK and US consumers, 2019: <strong>where more number of UK customers preferred Unrestricted speed with a limit on mobile data compared to Unlimited mobile data with a restriction on speed</strong></blockquote><p>The Ofcom data is quite stark about this structural limitation.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xmb7R3nChomc5kgB8q1Uvw.png" /><figcaption>Subscriber distribution vs Data Consumption (&amp;speed): Not much of a difference, source: Ofcom data</figcaption></figure><p>The Subscriber distribution vs Speed chart is remarkable, in that it shows a bulging middle. The tepid growth on the top (&gt;25GB) despite a net 4–5% decline in prices is indicative of a structural limitation.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5ariGyTf40yK_lf_o3pN-w.png" /><figcaption>Per capita Mobile Data Consumption Growth is linear in UK, source: Ofcom data</figcaption></figure><p>The mobile data growth so far is linear despite 75% of entire subscriber base, by and large, being on higher speed and data cap &gt;2GB. Still, the per capita consumption is hardly encouraging and in fact tends to the lower deciles. It all portends a structural problem in the market. The structural limitation will prevent a breakthrough growth in data consumption and it is inadvertently caused by telcos. Telcos because of the fact that they have monetized every unit of data consumption have hitherto created a behavioral disposition amidst subscribers that is averse to high data consumption on mobile. No amount of unlimited plans will shift that as people have crafted lives around this — for example, most listen to downloaded music now, most watch downloaded movies, most play online games at home during evenings, very few watch live sport on their personal devices (in UK, that is centered around lively pubs, wouldn&#39;t be different anywhere). I do not consider burst in traffic during events like festivals etc and this is a technology issue, not a behavioral one.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xu5o4tI-PFBsRzCVEFMc6Q.png" /><figcaption>Redefining the new Normal: Target Subscriber vs Distribution to realize the true benefits ‘unlimited data’</figcaption></figure><h3>Reliance Jio – an example of market disruption created by behavioural nudges</h3><p>Telcos hence need a new symbol and metaphor that will propel consumers to radically shift their behaviors. India, which has now the highest data consumption per capita (8GB) leapfrogged the world due to what is now commonly called as Jio-effect. Jio is a greenfield mobile operator who offered unlimited data, free voice and messages and all packaged in 4 simple plans. That alone was not enough, according to Jio executives, to change a notoriously thrifty Indian consumer behavior where even the premium tiers of subscribers were not used to consume data more than 1GB per month. Jio went ahead to bundle services that will trigger the voracious need for data. Simple, Jio offered content (OTT entertainment and gaming) for free. And they marketed not data plans which customers are actually paying for, but they only marketed the free bundled services exclusively. <strong>In other words, they marketed the demand (services) whereas they sold the supply (data). A new symbol and a new metaphor. </strong>The rest, as they say, is history. The symbol and the metaphor for Jio was not unlimited data with free voice and messages at a price that is one-fourth to one-third of the median plan cost.</p><p>That is not an impossibility, history is replete with examples where consumers have changed their behavior and Jio is, but, the latest example. What I’ve stated so far broadly holds true for most of OECD countries — barring Nordics. Nordics have certain uniqueness which I will not delve into in this post.</p><h3>New Choice Architectures for customers to maximise 5G potential</h3><p>What is therefore a new symbol and a new metaphor — which nudges the customer behavior. The answer lies not in more unlimited plans. <strong>The answer lies in the world of behavioral economics. Can telcos build new choice architectures that will nudge customers to shed their reticence and become voracious consumer of data.</strong> As the Jio story foretells, the success of business strategy lies in creating the demand — a new symbol and metaphor; supply is bound to happen. This requires Telcos to strike partnerships outside natural adjacency — Video is obvious, Live Sport, Live Social Gaming etc all requiring heavy data consumption.</p><p>How different will this be from bundling — the answer lies with how deeply the services are integrated (as business models become increasingly digital subscription led, cue <a href="https://medium.com/@narendransivakumar1/what-can-telecom-operator-learn-from-apple-a165c787cbef">Apple shift</a>), differential service experience (cue network slicing and 4K mobile displays) and the sheer physical reality of packet round trip delay time (benefits of locating content caches closer to the demand for Live Gaming, Live Sport overlaid with AR, immersive content etc).</p><p><strong>Be that as it may,</strong> one thing is clear, Telcos have to radically simplify product structures and shift customers to unlimited plans, as default. <a href="https://medium.com/@narendransivakumar1/change-series-1-netco-and-servco-old-wine-in-a-new-bottle-or-fortifying-against-hedge-fund-led-244542cea17c">Telcos do not have the luxury of chipping away</a> at structural limitations, they need to demolish them.</p><blockquote>Telcos have to redefine the New Normal. Unlimited — as the new basic plan — is perhaps a good way to begin.</blockquote><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=aef7d6f61f7a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Change Series: #1 NetCo and ServCo, Old wine in a new bottle or fortifying against hedge fund led…]]></title>
            <link>https://medium.com/@narendransivakumar1/change-series-1-netco-and-servco-old-wine-in-a-new-bottle-or-fortifying-against-hedge-fund-led-244542cea17c?source=rss-216b7c0837c2------2</link>
            <guid isPermaLink="false">https://medium.com/p/244542cea17c</guid>
            <category><![CDATA[telecommunication]]></category>
            <category><![CDATA[private-equity]]></category>
            <category><![CDATA[business-models]]></category>
            <category><![CDATA[business-strategy]]></category>
            <category><![CDATA[venture-capital]]></category>
            <dc:creator><![CDATA[Narendran Sivakumar]]></dc:creator>
            <pubDate>Tue, 01 Oct 2019 19:20:35 GMT</pubDate>
            <atom:updated>2019-10-01T21:12:03.321Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>Change Series #1: NetCo and ServCo, Old wine in a new bottle or fortifying against hedge fund led crusade?</strong></p><p>Whilst the last two posts dealt with Telcos possibly becoming a <a href="https://medium.com/@narendransivakumar1/what-can-telecom-operator-learn-from-apple-a165c787cbef?source=---------3------------------"><strong>digital subscription business</strong></a> and why <a href="https://medium.com/@narendransivakumar1/behaviour-and-culture-key-for-telco-transformation-4b00a7588599"><strong>culture and behavioural</strong></a> change are pivotal, this post will try to unpack a business model shift that has become vogue in UK and Europe market. That shift is a deliberate strategy of Telcos to split themselves into NetCo and ServCo to unlock hidden value and deleverage the balance sheets.</p><p>This is not something new, but there is a reason I used the word deliberate. Deliberate, because it is not government or regulator driven as we had seen in the case of New Zealand (erstwhile Telecom) and Australia (with NBN and Telstra). We had seen softer form of separation like the structural separation of OpenReach from BT enforced by the regulator.</p><p>Europe has lately seen a spate of Netco and ServCo split, beginning with TDC. Vodafone is the latest to announce that. What is happening now is different. It is not a regulator led putsch that is forcing the telcos to tear themselves asunder. Something larger is afoot. The business model shift is not driven by customers as well. The shift is largely due to a palpable churn within the investor community and changing profile of investors.</p><p>Telcos have been traditionally a favored portfolio among investors, especially institutional investors. Owing to stable cashflows and predictable margins, and also due to the Capex heavy nature of infrastructure investments, Telcos have been more than generous in parting excess cash back to investors as dividends. All major Telcos that you can think off, has a long dividend paying record. Pension funds and long term return funds were thereby attracted to Telco stocks like bees milling around. Such stable holding structure also provided Telco board and executive management some breathing space to run the business. As long as dividend flows, no hard questions were asked of the management.</p><p>However, that arrangement is changing. Some Telcos, in recent quarters, have either slashed dividends completely or provided very tepid dividends much to the chagrin of the institutional shareholders like pension funds which are reliant on dividends to pay back their holders a monthly income. Some of them inevitably sold off or pared their holdings, attracting a newer form of investors.</p><blockquote>Enter Global Private Equity firms and activist investors — cue AT&amp;T, BT and TDC. According to a bain capital report, on an average, a significant number of PE/Buyout funds meet their expected return target within a span of 24 months and the average return on invested capital for Telecom is around 2x. The Internal Rate of Return of a typical 10-year buyout fund with diverse portfolio is anywhere between 30 to 40x, significantly outperforming other PE/Institutional funds and S&amp;P index.</blockquote><p>Now the picture becomes more clearer. In telcos, you have an industry with a rich record of dividends and richer assets, slipping worryingly from investors eyes. And at the other end of the spectrum, you have PE/Investment/Buyout funds looking at the value hidden within telcos in the form of assets and quickly make a return — the making of a perfect storm. Some Telcos like TDC have already taken the PE path and are now being split into NetCo and ServCo.</p><p>So, the strategic imperative for Telcos is either be forced to split by investors or do it by themselves in order to generate more return to satisfy the long term holders. They have three business model options along with EV/EBITA multiples as depicted in Business Models Returns Equilibrium: a cost efficient and scale Utility model, a customer growth and engagement centric Digital Subscription Model or a hybrid Digital Utility model. In a utility business model, NetCo is where the value lies, with critical infrastructure assets and guaranteed demand, coupled with low interest rates to fuel investments. The strategy for NetCo will be to squeeze out costs, whip them lean and drive return multiples.</p><figure><img alt="The Business Model returns equilibrium" src="https://cdn-images-1.medium.com/max/1024/1*7rI9SBh7kvt78rTBSASuYw.png" /><figcaption>The Business Model Returns Equilibrium</figcaption></figure><p>The future performance of ServCo will determine the viability of Telco business models other than Utility business model. If ServCo, manages to unshackle from the Communications Service Provider legacy of selling commodity voice and data services, and moves into becoming a digital services provider with an enterprise valuation of 6–10x commanded by digital companies, then Telcos can not only demonstrate the viability of a model outside of Utility, but also hope to fortify against very keen activist / buyout firms seeking better returns.</p><p>So yes, the recent trend of ServCo and NetCo split is not a Old wine in the new bottle — as a benign regulatory push. It is real. Most Telcos want to occupy the middle of the Business Model Returns Equilibrium — a telco version of having the cake and eating it. That requires a concerted strategy to decouple business and evolve Digital Utility business models where one organisation is a lean and efficient utility and another a digital services focused agile entity. Left to activist /PE investors, they would cleave telcos into two business models and have the best of both worlds — Utility and Digital Subscriptions.</p><p>Recent attempts at evolving a NetCo and ServCo based business model is primarily to fortify themselves against investor led activism where the board may quickly loose control of the narrative to activist investors. The devil is in the details — will the NetCo be lean and value generative and will the ServCo have a return multiplier effect. It all depends on the change telcos want to drive. The nature and magnitude of the change, perhaps can wait for another post.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=244542cea17c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Behaviour and Culture – key for Telco transformation]]></title>
            <link>https://medium.com/@narendransivakumar1/behaviour-and-culture-key-for-telco-transformation-4b00a7588599?source=rss-216b7c0837c2------2</link>
            <guid isPermaLink="false">https://medium.com/p/4b00a7588599</guid>
            <category><![CDATA[culture-change]]></category>
            <category><![CDATA[people-management]]></category>
            <category><![CDATA[business-models]]></category>
            <category><![CDATA[business-strategy]]></category>
            <category><![CDATA[telecommunication]]></category>
            <dc:creator><![CDATA[Narendran Sivakumar]]></dc:creator>
            <pubDate>Tue, 24 Sep 2019 18:34:21 GMT</pubDate>
            <atom:updated>2019-09-24T19:19:33.676Z</atom:updated>
            <content:encoded><![CDATA[<p>Last week I wrote a <a href="https://link.medium.com/Iw2iYP8zf0">post</a> that emphasised about how Apple is subtly changing its business models to a subscription based business model. I also harped upon the need for Telcos to reimagine the business models lest risk becoming a commodity player like any other utility industry. Further more I ended the post with a tantalising poser about Telcos willingness to transform, not about its ability to transform as there is no doubt about the latter.</p><p>This post is a continuation of the same thread and I also want to reflect upon the keynote I delivered last week about Cognitive Telco – the lean and intelligent Telco in the Netflix’s mould.</p><p>Before I come back to the topic at hand – that is about culture and behaviour; let me indulge a bit in elaborating Cognitive Telco and the allusion to Netflix. The fundamental premise of Cognitive Telco is about shifting the paradigm of Telco enterprise from subjective, human dependant and manual interventions by design to Value driven, AI led, human augmented and automation by design. That’s a mouthful and will need another post onto itself. In brief, the paradigm is about changing the current state of Telcos where every process is intentionally designed to be dependant on human being’s wisdom, subjectivity, discretion and intervention. Think about a new product launch, business planning, new campaigns, store operations or even network operations. Can we invert the paradigm where AI driven automation is leading the charge and where humans comes to play only to handle exceptions.</p><blockquote>Essentially, from expressively subjective to invisibly objective.</blockquote><p><strong>Be that as it may</strong>, the focus of this post is not about the glorious benefits of AI coupled with Automation, of which there are many. This post is about the behavioural shift that is required to embrace the paradigm of Cognitive Telco and new Business Models.</p><blockquote>I see both of them as two sides of the same coin; as new business models requires the kind of lean, AI driven and massive machine scale fidelity within the enterprise – but I digress again!</blockquote><p>Key determinant for any change is changing the behaviour and culture of Telcos. It is easy for a digital native (cue Spotify) because, well, they are native only to the digital era. Not so much for Telcos which has been in existence for more than 15 decades – yes, you read that right. Out of these 15 decades, this industry has been the sunrise industry attracting the best of talent for 14 and half decades – not bad eh! In a way, <strong>Telcos are the original natives before the current crop of digital natives. </strong>However as much as it is a source of pride, it is a big hindrance.</p><p>The average age of employees within Telcos is roughly anywhere between 40 to 50 years. Amidst them, there exists the collective wisdom of 150 years of the industry – the perceptions, preconceptions, strong feelings about what will work and what won’t, ideas for growth and transformation and ideals. However, sometimes, history shows that there is always a need for an outsider to stir things up.</p><p>Stories of such outsider led disruption abound: the fabled Alexander slicing the gordian knot using a sword; the Ghengiz Khan changing the method of war fighting with cavalry for good; or much closer to modern times, Steve Jobs and Apple changing the digital consumption for good and recently Amazon and Jeff Bezos for retail.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/550/1*O3OG1lsxolvDfzyAXUfM4w@2x.jpeg" /></figure><p>Image courtesy: <a href="http://www.ancientpages.com/2018/04/07/gordian-knot-and-how-alexander-the-great-managed-to-outmaneuver-the-problem/">Ancient Pages</a></p><p>Telcos, so far, have been quite insular. We do not hear or see lot many examples of cross industry talent being brought in to drive change. Let me clarify, this is not just confined to executive levels, but at at all levels of the organisation – from developers, architects, analysts to business leads, commercial and product strategists. Only when multiple perspectives, diverse views and rich experiences come together, any cultural change will take root.</p><blockquote>Inbreeding of talent can otherwise only seek to reinforce status-quo and all of us know too well what happens to an industry that is facing severe market disruption when it chooses the well trodden path – cue Nokia?</blockquote><p>The need for cultural and behavioural transformation cannot be understated as Telcos seek to transform the business models and try to internally transform itself by adopting paradigms such as Cognitive Telco. If not, let me close this post with a quote from Einstein:</p><blockquote>Insanity is doing the same thing over and over again and expecting different results.</blockquote><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4b00a7588599" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What can Telecom operators learn from Apple]]></title>
            <link>https://medium.com/@narendransivakumar1/what-can-telecom-operator-learn-from-apple-a165c787cbef?source=rss-216b7c0837c2------2</link>
            <guid isPermaLink="false">https://medium.com/p/a165c787cbef</guid>
            <category><![CDATA[business-strategy]]></category>
            <category><![CDATA[apple]]></category>
            <category><![CDATA[economics]]></category>
            <category><![CDATA[subscription-economy]]></category>
            <category><![CDATA[telecommunication]]></category>
            <dc:creator><![CDATA[Narendran Sivakumar]]></dc:creator>
            <pubDate>Tue, 17 Sep 2019 21:36:51 GMT</pubDate>
            <atom:updated>2019-09-17T22:23:33.342Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>What can Telecom operators learn from Apple</strong></p><p>Last week, we saw the blockbuster Apple event. Apple launched a trio of iPhones, next evolution of a watch and an entry level iPad. These are exciting devices and have for sure stirred a great deal of consumer interest. Be that as it may, the event witnessed a paradigm shift when it comes to Apple business models and it is that shift that holds great many lessons for Telcos.</p><p>The usual business model for Apple is selling devices – a lot of them and at premium prices. That model is changing. The launch of Apple TV+, Arcade gaming, and Apple Music service signifies that shift. That is, Apple is slowly but inexorably shifting towards a subscription business model. Heck, even Apple Care/ Warranty is now offered in a subscription mode.</p><p>In a way, it makes a lot of sense. The devices have indeed become a commodity, not withstanding Apple’s superlative marking. When enterprise core strengths are rapidly approaching a commodity space, and in Apple’s case, the entire device industry has; it is important for the enterprise to innovate around its core strengths and change the business paradigm. Apple’s core strength is around delivering experiences that delights the customer – create a legion of loyal customer base. It’s not about their exquisitely engineered hardware of the devices, but about how the hardware and the software come together to create harmony for the consumers and that has made Apple what it is now.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/563/1*PjcBDEtkIPursjLX66cH7A@2x.jpeg" /></figure><p>Now Apple is pivoting its business model around its legion of customers and providing them with innovative digital services with a characteristic Applesque elan. Devices therefore become the essential via media to deliver these services. And these services cannot be costed as a one time cost for consumers, hence the recurring subscription model with no end in sight.</p><blockquote>It’s not the devices stupid!, paraphrasing President Clinton’s famous quote.</blockquote><h4>There is another industry whose core strength is getting commoditised rapidly and is looking to expand the value chain. I am referring to Telecom Operators – once the sunrise industry that betrothed the digital revolution, without whom there will be no iPhone, no Netflix, no Facebook and no google.</h4><p>It is high time for Telcos to borrow the leaf out of Apple’s book. One may argue that Telcos are into subscription business selling SIM cards, voice and data. But are these real subscription businesses? Where is the non linearity of growth associated with subscription businesses? Where are the cash flow strengths? None at all. Truly, subscription business models are meant to clawback significant network investments made in the form of huge capex outlays, financed through debt. Moreover, the capex cycles repeat themselves every 5 years – cue the capex outlays for 5G. In other words, Telcos by selling subscriptions are trying to earn revenues in order to pay for the investments.</p><p>In an industry that has saturated, with cellphone penetration exceeding 100% is most mature markets, even incremental revenue growth is at stake. In addition to that, the average revenue per user is only going down. The competition is severe with low pricing power. Regulator is bearing down the neck to even accelerate network investments. This is a classic sign of an industry that is on the throes of getting a commodity trap. The only way out is to reimagine their business model.</p><p>Digital Subscription Businesses offers Telcos the non linearity and pricing power. Because, they are not limited by cell phone penetration and lack of new cellphone subscribers. A consumer can buy limitless number of services from a Telco other than commodity connectivity.</p><blockquote>Just like how for Apple, devices become an essential via media; for Telcos, connectivity becomes the essential one.</blockquote><p>It is therefore possible for Telcos to pivot around their mature customer base (which is neither growing rapidly nor shrinking rapidly) and sell deeply integrated digital services that depend on connectivity to realise true experience. Can Telcos be imaginative and assemble an array of impressive digital services? Even if they do, can they create harmony between Over-The-Top digital services and their connectivity channels just like Apple created harmony between hardware and software? All are eminently possible. History perhaps will judge Telcos harshly when they ceded the entertainment, music and wallets to digital native industries. But nothing prevents them from building partnerships with these innovative organisations and integrate the services deeply to the network experience. Can they do it? Yes certainly they can. Will they do it? Perhaps, this will require an another post to elaborate.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a165c787cbef" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>