Oct 10, 2018
| 3 min read

Podcast #30

Exploring the Long Tail of IoT – by Param Singh

Our conversation with Param Singh covered his recent blog posts “Is IoT a Viable Market” and “The Long Tail of IoT”. The evolution of IoT has seen a lot of focus on bespoke solutions, and our conversation explores the distinctions between the types of use cases that are replicable, and those that fall along the “Long Tail”. The importance of business value is a recurring theme in our conversation, as companies that succeed tend to follow the most practical use cases for technologies. One interesting startup worthy of note is Brighton.ai, which is developing speech recognition technologies that are vertically oriented.

What has shaped Param's view of the Internet of Things:

Made in Japan: Akio Morita and Sony – by Akio Morita  

Prediction Machines: The Simple Economics of Artificial Intelligence by Ajay Agarwal and Avi Goldfarb  


We'll notify you weekly about new podcast episodes, upcoming guests, and news. You can subscribe to the podcast and if you'd like to be considered to appear on the podcast contact us.


View Transcript

Good day everyone this is Ed Maguire, Insights Partner at Momenta Partners, with another episode of our Edge Podcast series, and with us today we have Param Singh who is Head of Cisco IoT Platform and Partnerships. Param is a technologist, a writer, thinker, keynote speaker, and a thought leader. We’ve been going back and forth in a couple of interesting conversations; he’s got I think some insightful takes on the state of the market, and I’ve been looking forward to diving into our conversation today. Param, it’s great to have you. 

Hi Ed, it’s always good to talk to Momenta folks, I remember way back to 2013 when we were still wondering what this IoT thing was, so it’s always good to chat. 


Let’s first start with a bit of your background, and what has shaped your perspective on this crazy market that we call IoT. 

As a little bit of background, I’ve spent equal amounts of time in large companies like Apple, Oracle, spent about eight years each, and then three different start-ups, so I’m always looking for the next major wave to ride. I got onto the IoT side in 2013 when I founded a company called IoTracks, and it came from a foundation of doing embedded technologies that Oracle, based in the job I embedded. So, for me it seemed like a logical extension of the work that was happening on devices, but in a more standardized way. So, it was very exciting, and I think it’s been a long journey, these five years have seen a lot of change already, but in some ways the market remains very nascent. 


No doubt. Even as we describe the market, you’ve been publishing a series of some provocative blog posts, and one of them you’ve entitled it, ‘Is IoT a Viable Market?’ and I think that’s a question that a lot of people have, whether it’s a buzzword or a market, but I would love to get a sense of your take on the question asked there, and could you share a bit of the insights in that blog post? 

It came out of a dinner conversation where there were lots of folks poking fun at IoT, and ‘Is IoT dead?’ At the end of the conversation it really was, is IoT a concept, or an architecture, rather than a market? I took the position that, that’s the premise of the question, is IoT a market? The answer seems to be that it is not a market per say, it is an architecture that allows you to feed data into markets, and then the question becomes, ‘What are those markets?’ and we can talk a little bit about that too. 


What’s been the experience you’ve had that have shaped your view of the market? Can you point to some key events or people that have had an impact on your views? 

I think if you look at IoT, literally every company that I can think of has people whose titles seem to be associated with IoT, whether it’s in the enterprise or young start-ups. What has been most informative has been the number of pilots and projects that are done notably in the start-ups and enterprises, but what’s common amongst them is they are two extremes, one extreme is the existing enterprise segments, like enterprise asset management, supply chain, or transportation. On the other hand, we have connected widgets, and more and more all the ones that are connected widgets have a hard time proving business value, and hence all the POC, the pilots that aren’t reckonable, they don’t go anywhere.  

On the other hand, I can see that existing segments, let’s just say, with enterprise asset management, I happen to like it a lot, by adding contextual information about an asset, and let’s say the asset is in manufacturing, you’re providing net new value to that asset. So, enterprise asset management which again is a well-defined market, there’s a magic quadrant around it. So, within that market there is a segment of growth which is context-based awareness in CBA, and that context can come from sense of data, and it might be just the location, it might be is it on or off, visibility into the data, and just basics that feed value-added information into an existing market, and the end-number of other examples of market segments that are benefitting from the data-feed from IoT. 


In many respects when you talk about enterprise asset management, or asset monitoring and product monitoring, these are established market segments; what is the difference when you start adding new data? Are we just expanding on existing concepts, or do you see really new ideas emerging out of this drive to connect assets in great numbers? 

It certainly is adding the contextual information, but I think once you have that visibility, and you can correlate it to other elements, other content feeds, then those two together really give you better visibility, predictions, ability for optimization, which is not possible by itself today. So, I think implicit in that is the fact that IoT data by itself doesn’t give you all that information, the traditional attributes of an asset don’t give you that information. It is in context of both of those feeds that it becomes valuable, and you can see this happening in all kinds, in transportation, in supply chain, and a variety of other market segments. 


I know there have been a number of proof of concepts in pilots that have been designed to establish value, and I think when we look at this market I’d love to get your take on why what we call the industrial IoT market has been a lot slower to take off, then some of the initial expectations, if we go back about four or five years ago when there were some quite ambitious growth expectations in the market. What’s your take on the pace of adoption and the alignment, or misalignment with expectations that had been set back, when a lot of the real corporate interest in IoT was taking off? 

If you look at the promise of IoT, it seemed to be that you attach some sensor to a device, you connect it and the outcome will be realized. I think we’re all realizing that it’s harder for several reasons, and I outline them in one of my blogposts on LinkedIn, which was ‘The Long Tail of IoT’, and the majority of these pilots etc. are on the long tail, and the reason tends to be, it has required a greater degree of customization by used case. The sensors vary, the prequels vary, the connecting into existing systems is not as simple as one might have expected, again in enterprise software we should have known that, but I think hype and enthusiasm would lead us to believe it would happen right away. 

Also, the cost structures involved in the sensor, many times it might be okay for a pilot, but they’re definitely not going to be okay at a scale where the business benefit doesn’t justify the very-very high cost, and the costs throughout the layers of IoT, it could be the sensor itself which are new in many cases, or it is the fact that they need to be maybe packaged and ruggedized, because they’re not going to work in the industrial environment; taking a Raspberry Pi and trying to put it on the shop floor might work for a pilot, but you’re going to need hardened sensors and gateways for those environments. The connectivity is more complex than one might initially look at, and then the degree of customization required to integrate these into the enterprise applications, I think also has been underestimated. 

Other business factors are probably not adjusting to the business model, trying to sell using existing perpetual licensing mechanisms, not understanding who has the budget, whether its OT or IT, and engaging perhaps doing a pilot with one side and not involving both, because when you get to connectivity it’s probably IT, but when you actually look at the business outcomes it might be OT. I think some of the other factors are that the devil really does lie in the details of the business outcome; so, one really has to go through that, look at the technology, look at the scalability, look at the cost structure, or the total cost of ownership, and the time needed to realize value. 

So, again, these items are all listed in the blog post. But the key is, that as you walk into a project, being able to have a checklist that says, ‘Will this regularly scale?’ and if you go to the checklist you might save yourself and the company a lot of angst and pain by either architecting it better, where you’re looking at both business factors as well as technology factors, or looking at the cost of ownership up-front, rather than having to run into this after the pilot has been deployed. 


Do you think there’s been a lack of methodologies in terms of being able to calculate ROI? Certainly, there’s been a fair amount of discussion that early proof of concepts was more focused on just connecting devices and seeing what could come out of the data. What have been some of your observations about the challenges of being able to quantify ROI to be able to scale? And I think the point you made earlier was that early proof of concepts, or early projects, may have underestimated the amount of integration and effort needed to bring a project to fruition. 

Yes, totally. In fact, all my integrator friends when they look at IoT, of course they have the mechanisms of calculating it, it’s a well-honed art from integrators, but I think just the cost structure of deploying the pilot in many cases blows the ROI out of the water, or the level of customization and the surprises along the way. So, I think those have been the challenges, and talking to your friends in the industry who have been doing ROI studies, and looking at the variables that go into it, and taking those into account early-on in the process might be a good pragmatic step. 


Are there any technology hurdles or enablers that you think may be under-appreciated, at least in the market? I know you’ve got a great appreciation for the complexity of technologies across the stack, but where may there have been disconnects in either viewing how easy technologies would be, or how much leverage there would be, or how difficult it would be to deploy certain parts of the stack? 

I’ll just take one to kick it off, we all knew whether its 50 billion devices, or 10 billion, or 100 billion devices, that there’s a vast diversity in the type of equipment, and the type of sensors that would be needed, across many markets. But the complexity in the protocols, the traditional legacy protocols, I learned protocols that I didn’t know existed, even from the enterprise side or industrial side. So, that protocol layer, the connectivity layer, and then the integration to the enterprise, completely underestimated, majority of projects get bogged-down in that connectors to different sensors. So, the protocols, the connectors, the network layer, and making them all work seamlessly, and I think even more importantly you may get a sensor, or a type of machine, connected all the way through and get data. But then if you wanted to go back and make any changes, you needed a new tag of data from a sensor from the machine that you hadn’t accounted for before, you must go and redo that process. I don’t think there’s a well-architected process to deal with the variation, or deal with going and making iterative changes, it’s been extremely time consuming. 


Well that sounds again very much like the role that an integration platform would play, but as you’re probably closely acquainted with, as many are, there are many-many platforms. I’d love to get your sense of the state of the market of platforms, and I know there are hundreds by some counts, but why haven’t we seen one that’s emerged which could be close to a best practice, or a de facto standard in the market for exactly the types of situations that you’re looking for? I would preface that by saying you had companies like Informatica, Talend, Dells Boomie, SnapLogic and other traditional data integration and enterprise application integration platforms, but IoT seems to be very fragmented and we’ve seen this cambrian explosion of solutions. 

I think if we break it up into two types of platforms, ones that are focused on the cloud, the application enablement platforms, and the ones on the edge, I think as the large cloud vendors start building out their platforms and make them available at scale, I think we’ll see consolidation and alignment around those a lot sooner; you’ve also seen some exits of application enablement platforms. I think on the edge, just getting back to the variants of the types of machine and sensors, I think you see a lot of industrial vendors that have equipment they’re selling, early management systems, this, that and the other. Because of the variants they all want to optimize for their set of sensors, and then extend from there, so, I think we continue to see more variation on the edge, and we might start seeing some consolidation around the big cloud vendors who are all focusing their energies on adding IoT, IoT hubs, or IoT pass platforms on top of their current infrastructure. 


You had eluded as well to maybe ‘walled-garden’ is extreme, but the highly proprietary mindset of traditional industrial companies, and the challenges of connecting OT or operational technology with information technology. Could you share some of your perspective on bridging that transition, what are some of the most challenging aspects that you’ve seen in bridging the traditional industrial tech and information tech markets, and cultures? 

Traditionally I think we all know that OT and IT only interacted minimally when they had to. More and more as you have these security breaches, the nuclear virus etc., it has become clear to the organizations that they do need to start working together on the networking site, or if an OT department starts installing the network, how do you ensure users that are bringing their own devices on the IT side aren’t connecting it to the OT network? So, I think just because they’re being forced into working together, it doesn’t mean that they may or may not come kicking and screaming, and it also varies by organization, there is no one-size fits all. So, I think the integration is starting slowly, the dialogue starts, and I think the security considerations will drive a lot closer-working relationship between IT and OT, but it will take time and it will vary. I think the integrators also will play a critical part where there might be a default integrator that can be trusted on both sides. You’d have to pick one that has had an OT practice for a considerable amount of time, so that they can come in, take the set of best practices from both sides, and get everyone to the table. 

In addition to that, we are starting to see that there are job roles being defined in IT which are targeted at interfacing with the OT organization. I don’t have the list of job titles, maybe that’s not a marked article, but as you start seeing folks being defined in the IT organizations whose job it is to go work with OT, they become the bridge, and if someone is trying to sell into an organization identifying those kind of people, will become increasingly important because then they can be the catalyst, they can be your guides through the organization. 


It seems that what you’re saying is, you do need that domain expertise and that ability to bridge these multiple domains to successfully push through some of the cultural differences. I want to circle back to another one of your blog posts which you alluded to earlier, and that’s ‘The Long Tail of IoT’, and get your thoughts on how this idea of a long tail market, where you have these highly specialized, potentially very vertical used cases and solution areas, impacts the development of the market; and how you would contrast this long tail of IoT to, for instance, the enterprise software market? 

It’s something very current in my mind, and just to set context for everyone, the long tail of IoT, or the concept of long tail was from Chris Anderson’s book, ‘The Longtail’, and he’d applied it to websites, but I find the distribution on a standard curve that Chris highlighted, where on the website the majority of people just go to a handful of sites, and then everything else very specialized is in the longtail. In IoT, I think we’re seeing the same thing, the traditional enterprise companies and industrial systems, those are the systems that are first getting the value of IoT systems, of IoT technologies of this concept, the market has proven, and then the recent ones just don’t have it.  

So, for me enterprise production monitoring, if you look at Oracle for instance, Oracle has taken an approach of going vertical market-centric, so they’re integrating sensor technology into the Oracle production monitoring, and Oracle asset monitoring first, versus pushing a horizontal platform. That to me makes a lot of sense, because they know that business, they are taking sensor IoT data, but they’re making it a part of their standardized applications. I believe that’s a much more pragmatic way of pushing IoT in the short-term. 

Another point, back to the platform side, is that if I were a software vendor, or a start-up trying to build an IoT platform, I might look at the industrial guys who have their individual platforms, and say, ‘You need to swap out your platform with my new innovative thing’. I think that’s a hard battle to win, because say in buildings, building management systems, those systems have been communicating to their set of heaters, coolers, chillers, what have you, for a long period of time, and trying to come in and displace that, versus finding a way to integrate with them, I think is a more pragmatic way, and all of these examples set in the center of the standard curve, rather than on the edge, or the long tail. 


I think an insightful analysis in your ‘Long Tail’ pieces, is that there are a lot of applications that potentially are not replicable, and the challenge is that particularly if you’re a vendor, you don’t want to go too far down the path of non-markets as you describe it. Could you provide a bit more color on avoiding non-markets, and how a company that’s investing either from the adoption perspective as an adopter, how to avoid going too far down the path of just building a bespoke one-off solution, and how participants or vendors, and technology providers in the market can avoid going down that rabbit-hole of non-replicable solutions? 

I go back to the dinner conversation that I was having with a couple of colleagues, there was a guy from LA, and from the Valley etc., so the joke became that everyone in LA has a script, a movie script, the same is true for Bollywood, and everyone in the Valley has a start-up idea. I think in IoT everyone has an idea of how they could add a sensor to something and believe that would turn into a market.  

I think the answer is, you probably won’t find self-referencing customers, the definition of the market for most of the things that are on the long tail, or just an idea to connect a particular sensor. I have a checklist in that blog article that you can go through, and one of them is, ‘Are the customers self-referencing?’ So, if you’re a large company and you’re looking at IoT, looking at the use cases, or offers, sales offers that you have in market, a news article as an example again, where you have asset monitoring, or you have production monitoring, and you may want to see inside, so you may want some predictive models built on top of it. They’re adding sensor data, proven market, you understand what the sensor data can add value to, and then a large company you have to show returns, faster.  

So, there, just by applying looking at your existing portfolio, looking at the existing offers, and then looking for adjacency and then extending them first, and then putting the other ideas that folks come up with, in incubation, is a prudent way. You can have your criteria of how you rank them, and I’ve offered some. If you’re a start-up, again you want a couple of reputable sales, so avoiding the ones that are cool or maybe close to your own personal heart, stepping back and being a little objective, and looking at the ones that are reputable, where the cost-structure is viable, the returns aren’t over multiple years, and we can get a return over several quarters. This much is also true for start-ups. So, I think applying some traditional business thinking to this would help delineate non-markets from the ones that you can extend in a meaningful way in the short-term. 


Yes, you alluded to the bright shiny object’s syndrome, which I guess is a syndrome that appears in technology and financial markets commonly, but there are some newer technologies that are quite relevant and do have some meaningful impact. I’d love to get your thoughts on the potential leveraging and downstream implications of the increasingly powerful artificial intelligence machine learning technologies. How do you see these technologies impacting the evolution of the market? 

I think AI is a fascinating topic, just like blockchain and IoT, AI is now being touted by everybody. I was at this event recently at the Hass School of Business, it’s the innovation forum run by the Garwood Centre. We had this professor, Avi Goldfarb, and Avi has written a book, and what is really interesting is, the book is about AI, but it’s called ‘Prediction Machines’, he has some YouTube videos as well which I thought were very well presented, where the hypothesis is that trying AI at this time isn’t about building an iRobot, the movie, or building robots, artificial intelligence in that sense, but you’re really building these mini-prediction machines, and you’re applying prediction models to precision theory. So, it’s very pragmatic in that sense, and a strongly recommend the book, ‘Prediction Machines’, by Avi Goldfarb, and watch him on YouTube.  

Machine learning AI, we have found that it is a lot more pragmatic to look at things like adding machine learning algorithms to a known product, so you can add vision machine learning to a video stream, we have a video product, and one of the use cases that was fascinating was, using machine learning to detect errors at scale on a fabric manufacturing plant. They were using these emerging technologies but you’re applying it again to a very pragmatic problem, which can be solved. Another learning from that is, you really need to identify folks that can do the machine learning, or train the model, and get people who have been doing that in that space for a while, so you’re not trying train people from the ground-up.  


Do you have any thoughts on blockchain technologies, and the potential utility or relevance of distributed ledger technologies in connected industry and IoT? 

I’m sure there are used cases for traceability, and I hear my friends talk about them in different context, and diamonds supply chain, so its again tracing it through, or even co-chain. But again, I would caution that distributed ledger technologies unless they get to scale, applying it to an area like IoT which itself is emerging, and I believe there are constraints in the number of transactions you can process. So, either in IoT, if you’re doing some tracing, you need to just work on the events; so, you reduce the number of things that need to be part of that tracing, or the technology, the blockchain technology where it meets the scale. I believe there are lots of people looking at problems like that. But I would still say it’s still nascent, especially when applied to IoT. 


Yes, we certainly are in the early stages. I think a lot of people have drawn parallels between the Internet in the early-1990’s and the state of blockchain technologies today, but there is a lot of interesting innovation happening. 

When you look at projects that have been successful or either companies that are very successful in adopting Connected Industry strategies, are there any lessons that you can draw from a success story, and can you point to any notable examples of what we’ll call successful IoT projects and implementation, that really stand out to you? 

I’m sure there are plenty in manufacturing. I found a recent announcement between PTC and Rockwell particularly interesting, because you have a company that does have an IoT stack, they also have Kepware that does device connectors. On the other hand you have an OT leader in Rockwell, and there’s clearly a correlation or value, so I think maybe strategic partnerships like that are the ones where you can start bridging a lot of the things we have been discussing, whether its connecting, whether it’s taking one market segment and combining your strengths and marketability, I think that’s the kind of thing I find intriguing in the industry at this time. I believe there are several other opportunities like that. In fact, my next article is going to be, ‘Best of Breed’, which is, how do you take a particular industry, identify the layers, the key integration points, the points of risk, and how do you identify the best of breed vendors within that? But then the integration of partnership can be very-very deep, they can’t be at a cursory level because the devil does lie in the details. 


That’s very insightful. I think the interesting evolution in the market has come around these partnerships. Just recently we saw an insurance company, Munich Re, buying an IoT edge platform solutions company Relayr, and you’re seeing this in many respects, that partnership or that acquisition really validates the fact that there is truly much greater value from combining the companies, your business value, than there would be even just through partnership. So, it’s kind of nice to see that. 

How do you see the market evolving over the next decade? Are there some catalysts that you’d be looking at, or any forces that you’re keeping your eye on? 

Going back to the Relayer thing, a friend of mine sent me an excerpt of that agreement, and in that excerpt, there was this item, I think there was an observation that instead of focusing on five or six markets, a company should focus on one or two. So, if you go to any of the IoT vendor sites that are looking at different industries, they tend to cover and hedge their bets across five or six, because who knows what might be attractive to a customer. But I think this observation was, focus on one or two, and go deep and succeed in those, and then you can extend it, you’d probably be a lot more credible in the markets that you have selected to focus on. So, I thought that was an important observation for the next few years, is don’t go as broad as possible, and I’m certainly taking that to heart; take a particular market. I think that will pave the way for what happens in the subsequent years, is, industry-specific solutions that take the industry platforms, bringing out data, make it contextual, and then deliver value to customers at the right price-point from talk of cost of ownership, and do it in a replicable way. I can’t see five or ten years out. 


Well, it’s hard enough to predict the weather in the next week! I want to get a sense of any interesting or promising start-ups or technologies you’ve seen recently? Are there any interesting companies you’re keeping your eyes on? 

There’s one, especially because of all the AI hype, it’s a company called Brighton.ai and they’re doing solving for some very vertical used cases in speech. The idea would be that in particular industries you need speech for very particular purposes, that instead of general-purpose platforms like you have from the top platform vendors, they will customize and deliver you solutions; so, for industrial if you have particular environments that need speech optimized, it’s a lot easier to optimize it for that area. I think Brighton again is consistent with the discussion we’ve been having of narrowing the focus, delivering against it, and then expanding from there. 


That’s really interesting. I had a conversation earlier on one of our blog posts with Mike Flannagan over at SAP, and he’s talking about his view that ultimately so much compute will be essentially driven by speech, speech recognition, and that we will have a much more ubiquitous fabric of services and technology embedded around us. I think that’s a super-interesting area, it ties in well with the other themes I think you brought up, which is the need to go vertical, and not try to boil the ocean as it were in connective industry. 

I have one final question that I always like to ask, which is a recommendation for a good book or resource for our listeners. It can be anything that you would gift to a friend. 

I used to always quote a book from long ago, it was called ‘Made in Japan’, by the President of Sony at that time, Akio Morita. He talks about his identifying, accidentally I might add, the need for a Walkman was to have portability of music, shrinking it in a time when everyone was carrying around boomboxes. It’s a fascinating book in any case because I think the iPod, this, that and the other continued that theme of music portability. I think this can be applied to many-many industries where you step back and look at the underlying transformation and being the first to get to it.  

I really-really like Avi Goldfarb’s book, ‘The Prediction Machine’, because it is a level of clarity on how you can use AI today. They break it down into a very simple systemic way of using these prediction machines and filling in gaps between data as a way that you can start doing it today, rather than waiting for whatever future of AI thoughts are envisioning. 


Those are great recommendations, I’m definitely going to make sure I get them on my Kindle. This been a great conversation Param. 

Again, this has been Ed Maguire, Insights partner at Momenta Partners, and we’ve been speaking with Param Singh who is Head of Cisco’s IoT platforms and partnerships. We will have links to the books in the show notes, and certainly if there are any questions or comments, we always welcome them. So, Param, thank you so much again for taking the time to speak with us today. 

Wonderful, always good to talk to you and to Momenta, I appreciate it.