Monday, March 23, 2015

Internet of Things and Smart Systems


Last month I had the pleasure to lead two roundtables about connected objects during an event jointly organized by Cité des Sciences and the National Academy of Technologies entitled “Connected Objects : a third digital revolution”. The first roundtable gave me the opportunity to discuss about the systemic vision that is necessary to understand and design connected objects, together with a very interesting list of companies:  Terraillon, Medissimo, Connected Cycle, Withings, Kolibree. This post proposes a summary of some of these ideas, with a focus on understanding what make connected objects smart and useful. This is not a summary of the complete discussion; it is filtered by my bias towards system analysis, as is expressed in a previous post from this blog. In a way, this is a follow-up to the talk that I gave at the 2013 Amsterdam conference on Smart Home, which I made available on Slideshare. It is also influenced by discussions at the IRT SystemX, the French technology research institute on “Systems of Systems”.

I will follow a simple outline:
  • First I will take a look at “connected objects” seen as objects with the additional benefits of being connected. My opening statement during the roundtable was that we should not care about this from a technology standpoint (what is added to the object) but focus on the customer experience. This may sound like a boring and self-evident statement, but my experience shows that (1) there is too much excitement about the technology (2) not enough effort with experience design, and that we are bound to see more disappointment with connected objects.
  • The second section will look at Internet of things (IoT) as a component of the Web Squared vision. With this second level of system analysis, connected objects interact with the complete Web in both directions, they bring additional senses (through sensors) to the Web and become contact / interaction points (bringing the Web “closer” to the user, what Joel de Rosnay calls the “clickable environment”). Obviously, the burning engine of this system is the value that we can bring though Big Data, a point that was abundantly made clear during our roundtable discussion and which is the topic of many books. However, a similar observation may be made that data follows usage and that focusing on customer experience first is the best way to avoid technology mirages.
  • The last section of this post will focus on “smart objects”, as defined by Olivier Ezratty – who was also a participant to the roundtable – as “objects that talk to one another”. There is a consensus that “we are not there yet” in the sense that experiences that are built on many connected objects interacting with each  other are rare, tend to involve objects from the same manufacturer, and to follow a closed set of possible interactions. “Smart objects” correspond to a more advanced “connected object ecosystem”, which is necessary to unlock more value to the end user.

1.   Connected Objects Seen as Objects

At a first glance, or at a first systemic level of analysis, a connected object gets two additional benefits from its connection:
  • Access to a deported interface, which is most often the smartphone. The tremendous capabilities of the smartphone (quality of the screen, touch interface, motion sensors, etc.), together with its ubiquity, make it a wonderful user interface for most scenarios.
  • Access to large storage and computing capabilities. This is true for computing, in the sense that one may do more on a smartphone or in the cloud that on the embedded chipset from the connected object. This is even more important for storage, since external storage gives both better capacity and durability. Connected objects have “memory”, as exemplified by the Withing connected scale.
We will look at the deeper benefits later on, but I would maintain that the large majority of connected devices that are available today are mostly selling the benefit of being controlled by the smartphone with some form of extended memory. That is, the additional benefit of true Web, Big Data, Smart System integration are either not there, not developed enough to be useful, or they are readily available in other ways.

My experience, especially in the realm of smart Homes where I have been quite active with my previous job, is that remote interaction with the smartphone does not carry enough weight to last long. One needs to add systemic value to the connected experiences, based on scenarios, context, and personalization to keep the interest going. Otherwise, once the novelty wears off, you are left with the underlying object: if it is a great, useful object in itself, you keep using it, otherwise you simply discard it. This is precisely the content of what I said during the Amsterdam Smart Home talk, with a focus on “life moments”, which are the daily events that may trigger scenarios.

The business model of such connected objects is, most often, to sell the object with a premium that corresponds to the improved service. The challenge is twofolds: (a) to deliver enough recurring value (b) to reduce the ownership cost/burden of the “connectedness”. The keyword “recurring” is important: it is easy to sell a connected object with the combination of aesthetic design and technology excitement, but if there is not recurring value, the business model is not sustainable. The ownership costs grows over time, with issues such as battery replacement / charging, network setup and update (the first wifi/Bluetooth pairing is easier because of the excitement of the recent purchase, the later ones, once the box and user guide card is gone, is trickier). Genevieve Bell, the anthropologist from Intel, explains this very well: each connected object is begging for attention, from its notifications to its battery charge, creating a true “cost” of ownership.

The Nurun design agency has a nice way of formulating the value challenge:  the value of the connected object, defined as immediate (out of the box), aggregate (over time) and emerging (such as social benefits) must outweigh the replacement cost. This is the only way to generate a sustainable business. As of today, and if the systemic value that we will discuss in the next sections is not there, there are few objects that will pass this test, and their success comes more from their intrinsic value (objects that prove useful and pleasant to use irrespectively of their connection) than their “connected” status.
To win this difficult equation, the technology battle is the excellence with service delivery (especially the app on the smartphone). Companies need to develop state-of-the-art skills with respect to digital software. This makes perfect sense since this is a “service business model”: delighting the customer with superior services associated with the object. This is where the “Web Giants’ lessons”, together with Lean Startup principles are relevant.

Giving memory and control to objects require sending data to the service provider, which prompts the issue of respecting data privacy. Our second roundtable was dedicated, among other things, to the issue of data privacy. It would require a different blog post to deal with this issue, but I would like to point out that, as of now, the absence of real customer value from her/his collected data is a bigger issue than the respect of data privacy, which is a great way to introduce the next section. Waze is a great example to remind us that applications which generate true user value gets access (consent) to lots of personal data.

2.       Internet of Things and Web Squared

If we pursue our thinking, and look at a more global system analysis, the connected object may benefit from two additional features:

  • It may send information to the cloud, using it sensors. The object is a mobile / wearable / personalized data capture device, which feeds “global services”, hosted on the Web. This corresponds to the “IoT as the senses of the Web” metaphor.
  • It may receive information from the cloud, and act upon them. The animated object (from a simple display – where the connected device acts as a more ergonomic and more convenient device than, say, the smartphone – to a complete animated behavior change – for instance a small robot whose interaction embodies a web service).

These two sides are the two parts of the “web square” vision: the connected object becomes part of something bigger. The object is not enough; it’s the digital experience that matters. The value does not come solely from what is in the object (sensors & actuators included), it comes in a large part from what is in the cloud. If you are not familiar with the “Web Squared” concept, the move “Eagle Eye” is a great and entertaining illustration.

Because of its very large scope, it is difficult to categorize the different business models that such a system approach enables. Without completeness, we can list:
  • Personalization and contextualization of a digital experience: the connected object is used to enrich a positive experience that is sold as a service. This leverages the first direction: object to cloud. Ambient digital experiences rely on connected objects.
  • Reciprocately, connected objects may better the digital experience though improved usability. In his book “Enchanted objects: design, human desire and the internet of things”, David Rose writes about glaneability, gesturability and wearability, which are three compelling reasons to use connected objects as a way to interact with the Web.
  • The complete B2B scope of connected objects, which is huge and outside the scope of this blog post, fits nicely into this object-to-cloud system view. Although the B2B model has its own logic, there are clear dependencies and cross-opportunities. For instance, metering devices, which are introduced with their separate business logic, may become part of customer digital experiences.
  • IoT/Cloud integration is a way to improve efficiency of business processes (I will not qualify them as digital, most business processes nowadays are digital in some form). This efficiency improvement may translate into cost reduction that may be passed to the consumer. Improving efficiency may come from cost reduction, (cf. metering example), risk avoidance or better characterization (insurance),  or performance improvement (cf. automated logistics). The “pure case” (when the value created by the efficiency gain is larger than the connected device) is uninteresting (objects become connected and the cost of their “status upgrade” is not seen by the customer) but common (more and more objects will be built with “connected object capabilities” – think of your printer’s ink cartridges).  The hybrid case (where the created efficiency value is not enough, but part of the business case for the connected object) is the most interesting one.

If we get back to our topic of “consumer connected objects”, the common umbrella for most of the digital experience that may be enriched in a systemic way through connected object is the “assistance” experience, from immediate to long-term assistance and coaching, from Siri or Google Now to healthcare services. Assistance requires context (personal data), knowledge (big data), personalization (machine learning) and problem solving (automated reasoning). The consensus during our roundtable discussion was that there needs to be a fair amount of each of these capabilities to provide a cloud assistance or coaching service that can bring sustainable recurrent value to the user. Most wearable devices, intended to be a piece of a wellbeing / health care digital experience, are still very far from meeting this goal, as can be seen by the mediocre smartphone applications and the fast decline in usage rate once the initial excitement wears off. The buzz about “quantified self” gives a good illustration of this point. Three years ago, I conjectured that quantified self would fit perfectly a combination of narcissists, geeks and people who enjoyed “system thinking” (if seeing a curve with the evolution of your bio-data tickles you, you’re in). This is a fairly large intersection – I fit just in there J – but it is still a niche. For most people, the “quantified self” experience left to oneself – which is exactly what I enjoy about my Withing connected scale – is too abstract and must be replaced by a true coaching experience.

This leads to two major challenges:
  • To generate knowledge from connected objects, one needs smart algorithms and data, and a little bit of time. To get data, one needs both usage and consent from the user, which both requires to deliver actual value. There is a chicken-and-egg conundrum: one needs data to deliver value, and value to attract the data. In most cases, the solution is to follow a two-step process, where the value generated by Big Data analysis comes in a second step, with a different value model to bootstrap usage.
  • The combination of information and technology required to produce a viable assistance or coaching service is accessible to large & technology-focused companies such as Apple or Google, but difficult to reach for most companies. This means that a partnership is required between different players, one of which is the connected object manufacturer. The object is not the center of the business value, what matters is reaching the critical mass of content (what to say), contextualization (when to say it), personalization (how to say it) and knowledge (why to say it). It is possible to add human interaction into the system loop, as a way to add knowledge or reasoning, which brings the digital experience closer to traditional coaching, but this usually comes at a cost. Still, there is likely a sweet spot for hybrid assistance or coaching services that mix the benefits of smart technology and human care.

The technology challenges associated with this systemic level are different from the previous one from Section 1. The first key domain is API (Application Programming Interfaces) architecture, used both for data collection and ecosystem partnerships. The importance of API comes from the second observation that partnerships will be required. The second domain is obviously data mining, as in Big Data, which must be either a strong point of the aspiring connected object contender, or one of its preferred partners. As stated earlier, most of the connected devices that have been introduced so far, especially in the field of connected health, fall short on delivering value with their data mining abilities.

There is a related challenge to this system vision, which is the social acceptability of “smart” cloud-based coaching services. The question of the “social acceptability” of artificial intelligence is also worth a separate discussion which I will address in my next post. During a presentation at the NATF, Dominique Cardon explained to us that the word “algorithm”, which had a very positive image a few decades ago as a symbol of technological progress, is now seen with more suspicion by the average citizen. “Smart” systems are under scrutiny, whereas social systems, where recommendation comes from the community, have a clear trust advantage.

3. Smart Objects Ecosystems

The ultimate goal of smart objects is not to improve, enrich or extend existing experiences, it is to create brand new ones. This will happen once smart objects start to act as a smart system, which supposes that they interact with one another. To build truly exciting and new experiences supposes an adaptive and intelligent behavior, obtained from distributed control and autonomous smart communication between objects. Today, the “state of the art” for mashing-up connected objects and web services seems to be IFTTT, which is deterministic, rule-based and centralized.

This long-term goal of what would be described as a smart object ecosystem is close to the general topic of this blog, in the spirit of Kevin Kelly organic/grown distributed systems. I have already mentioned the ideal of “calm computing” in a previous post. The three key principles of calm ubiquitous computing are the invisibility of the “smart dimension” (computing) – which does not mean invisible objects – the ability to stop down the assistance and the implicit machine learning – the smart object ecosystem must learn from the user, not the other way around. I often quote Adhoco, a Swiss smart home solution provider, as an example of existing smart objects service that learns from its users and operates under calm computing principles.
Another direct application of biomimicry is how to design a fail-safe smart system. Not only the control should not be centralized but distributed across the system elements, following complex systems design principles, in addition smart functions and devices must operate as assistance to lower level functions, and not the opposite. Following a simple principle that says that the more sophisticated the device, the more likely it will experience failure, control must be designed so that lower level automation may function independently from the more advanced “smart” or “adaptive” component.

A traditional failure of advanced smart home systems of the previous decade was the introduction of complex computerized systems to control low-level but vital automatic functions (such as light control or blind opening). The inevitable computer problems would then translate into a dysfunctional house, which is anything but smart. The appropriate modern design is both to avoid single-points-of-failure, and to make sure that high-level control logics are introduced as fail-safe “adds-on” to lower level of system logic.

The business model of smart object ecosystem is to provide with a new experience, or with a new level of satisfaction linked to an existing digital experience. The conviction that I expressed during this 2013 Amsterdam Smart Home interview is that it requires an “operator”, that is a company that takes responsibility for installing and maintaining the smart system. The alternative approach is to believe in interoperable standards and let the customer be the integrator of his own system, from her home to her health digital environment. It follows from the low level of maturity of “smart object interaction”, which we all agreed on during the roundtable, that letting the user be the integrator is a risky business scenario. This DIY (Do It Yourself) approach has two drawbacks, which I have seen firsthand with smart home experiments: first the smart object setup is a barrier to most consumers, but a few passionate geeks; second, the simplicity required by DIY installation prevents from delivering a true “smart system” behavior, which keeps the experience within the range of what is achieved with Section 1 and Section 2 approaches. This means that the value associated with the “target ideal experience” (of a smart home, for instance) is simply not there, which prevents raising a sustainable fee for the new service. The business challenge is to setup a brand with a clear promise, a distribution and installation network and the service maintenance infrastructure. I see the move of US telcos, such as AT&T or Verizon, towards smart home operators, as a significant signal. 

The technology challenges associated with this “third level of system integration” for connected objects are precisely linked to system integration. The skills and the software challenges are also different from what we saw in Section 1 and 2. Open system design, fault-tolerant architecture and machine learning are three key components of what is necessary to build a smart object ecosystem. Although digital and Big Data skills are bound to play an important role (level 3 encompasses levels 1 & 2), this is foremost a “system of system” integration game. Usability, availability, reliability, maintainability and adaptability are the key design challenges for smart object ecosystems.


On the one hand, I am expecting a lot of disappointment with respect to connected devices in the years to come because I believe that many of the barriers identified in this post still have to be removed. A very crude summary would be to say that connected objects are not enough, and that we are still far from the promise of smart “assistance” through connected objects, because there is a critical mass of data, learning and service innovation that requires more time and energy that what has been spent so far. On the other hand, I am a big believer in connected object because I think that the best is yet to come. This is just the beginning, not only will technology improve in all directions (miniaturization, better performance, new interaction channels, much better speech, image and pattern recognition, etc.) but time will make the conundrum identified in Section 2 become a virtuous cycle. The more time passes, the more we accumulate data which enriches the value that may be delivered through connected objects.

I believe that the distinction made here between three business models for mass market connected objects is useful, because it emphasizes different challenges and different expectations from the user. These three dimensions are not exclusive, many connected object strategies are bound to be a combination, but it matters to focus on the true benefits brought to the consumers. Here is a last attempt to summarize these three models:
  • The “service” business model uses objects to deliver improved services, with a key contribution of digital software excellence to the connected object experience.
  • The “efficiency” model uses objects to deliver better efficiency through big data, with a technology focus on API architecture and data mining.
  • The “new business” model provides a new experience that is made possible by a system of connected objects, where the technical challenge is the combination of usability, resilience and machine learning, from a “system of systems” point of view.

Technorati Profile