Nest – The Internet of Things?


I always hated to program heating system thermostats. Ridiculous key combination, repetitive tasks and at the end a build-in predictive algorithm that thinks it knows better and always guesses wrong. Adding insult to injury some of the learning thermostat purposely show the wrong temperature – which is considered a feature.

With winter coming another dreadful task to change programming of all thermostats was necessary. Rather than going through another set of strange key combination I bought the Nest betting on its iPhone apps as remote control.

The Nest is build by a home automation company co-founded in 2010 by former Apple engineers Tony Fadell and Matt Rogers. The device oozes Apple-like focus on design and its packaging shows similar attention. I have to exclude the funky shaped screwdriver, which seems to be more “for show” than anything else.

Installation was reasonably quick with the only hiccup was due to our system having two fan speeds and the Nest only supporting one. We like to run the fan constantly at low speed to circulate air – too bad. I had to go online and search as well as get the old thermostat’s manual to resolve the fan connectivity issue.

Once the nest was installed, powered-up and connected to WIFI it started with the now common consumer experience of “Updating”. I watched the “blue bar” with a sinking feeling of doom – luckily the device finished with still been responsive and I could connect to it from the iPhone app – programming was a breeze with a convenient temperature control.

Nest does what it is supposed to do – it is a good learning thermostat. The motion sensor is a cool feature allowing the device to automatically toggle “away” or “home” mode based on the absence or presence of motion.

The apps on iPhone, IPad and computer all work fine and the Energy History is useful. I can see the room temperature on my computer in the notification area and I have access to the Nest app from everywhere – cool.

nest history

The Nest cloud service however is underwhelming especially considering the price of the device.

  • Where are the real-time statistics about inside /outside temperature, humidity and motion?
  • Where are the attached New York weather service with forecast and predictions?
  • Where are statistic reports and comparisons based on the “collective intelligence” of thermostats of other New York fellows?
  • Where is the efficiency report for our house?

The Nest is a beautiful hardware device that can be remote controlled by apps with innovative features. As an example what the Internet of Things could look like or a showcase of what cool services devices connected to the cloud could offer it is a disappointment.

Email – The Largest Social Network?

It is interesting to think of e-mail as a giant social network, which would put Facebook squarely into the number two spot in the market. The immediate questions that arise are: what is different and why has e-mail not adopted more social features.

Both services have identity, profiles and allow users to connect and share information. However social networks are more focused on sharing while e-mail is still the preferred channel for private communication.

The role of a private communication channel might be a reason that e-mail providers haven’t adopted social features more aggressively. This in turn could be the reason that the market of online e-mail services, such as gmail, yahoo mail, Hotmail, has not produced a player like Facebook, which exploits network effects to dominate.

Does social media analysis apply to email too and how similar do the networks behave? One way to find out is to run a test. The playground for e-mail and text analysis is the Enron email dataset.

Enron Corporation was an American energy, commodities, and services company based in Houston, Texas. Before its bankruptcy on December 2, 2001, Enron employed approximately 20,000 staff and was one of the world’s major electricity, natural gas, communications, and pulp and paper company, with claimed revenues of nearly $101 billion during 2000.

The Enron scandal, revealed in October 2001, ultimately leads to Enron’s bankruptcy. Kenneth Lay, Jeffery Skilling and others went on trial for their part in the Enron scandal. Lay was convicted of all six counts of securities and wire fraud for which he had been tried. Skilling was convicted of 19 of 28 counts of securities fraud and wire fraud and acquitted on the remaining nine, including charges of insider trading.

To be able to run social media analysis tools on the Enron data set, it needs to be convert into a social network like structure of profiles, connections and messages. Connections are formed between actors if they had an email exchange and all messages authored by an actor are combined into the profile.

The first step is to analyze the network by scoring the influence of actors, i.e. how information has spread through the company. The following screenshot shows actors with Enron email ranked by influence:
Analysis 1

The list of the 10 top highest scored actors includes Lay and Skilling:

Rank Email Description
1 Jeff Skilling former CEO of Enron
8 Kenneth Lay CEO and chairman of Enron

The second step is to cluster actors according to topics of interest based on e-mails they have authored. Topics are extracted by the system automatically and topic-related clusters group people that used words, phrases and concepts in a similar way.

The following tables show the top three topics Lay and Skilling dealt with in their email and the top terms that constitute the two main topics:

Email Topics Topic_2, Topic_9, Topic_1 Topic_6, Topic_4, Topic_9
Topic Terms, Concepts and Phrases
Topic_2 enron, hou, agreement, corp, legal, credit, master, north, development, america, original, isda, 713, fax, transaction, ena, review, shackleton, counterparty, 2001, comments, questions, financial, draft, power, …
Topic_6 business, company, management, group, services, risk, development, trading, capital, time, president, communications, market, technology, firm, financial, companies, board, year, enron, ceo, internet, position, …

The next analysis step is to use topics to pull-up clusters of similar actors and to filter and analyze communication flow:

Analysis 2

The analysis tools used above have been designed to identify networked customers that will respond to marketing activities by brands on social networks. E-mail is still one of the most important channels for brands to connect with customers (permission based) and above analysis has shown how similar social and e-mail networks are regarding dissemination of information.

Will we see them merging moving forward?

Next Generation Of Mobile Apps

Mobile is changing the behavior of users to an app model – apps are installed and used to accomplish a specific task. This leads to fragmentation that favors services that take a new or optimized approach to address the needs of mobile users.

cloud service

We have seen such a phase of fragmentation before with Web portals such as AOL, MSN and Yahoo. The typical portal user behavior was to adopt other portal services once they had signed-up for three or more services offered by the portal such as e-mail, IM or photo sharing. This did breed a culture of just-good-enough services offered by portals.

Google with its search engine disrupted the centralized content and service model of portals by making it possible to find content that is distributed across the web. Portals therefore lost their stickiness, hence users became much more inclined to adopt best-in-class standalone services to create, consume or share content.

Facebook from its experience and market strategy feels like a content portal to me with the difference that it aggregates user profiles rather than information. Facebook Home can be seen as an attempt to counter the thread of fragmentation caused by mobile apps.

Facebook services replaced by best-of-breed apps could seriously undermine the stickiness of the social network. Apps don’t allow strategies such as rendering web pages inside an app. Offering integrated web browser capabilities avoids app context switching and keeps users inside the Facebook experience. Currently the only way to counter fragmentation by apps is defensive, i.e. to not allow other services to access and migrate users’ connections.

The current generation of mobile apps has taken existing web bases solutions and optimized them for the mobile form factor as well as touches and gesture control. A good example is the transition from web to mobile search – the basic user interaction model hasn’t changed – you post a query and the system responds with a result set that allows you to navigate to the requested information sources.

Apps are starting to disintermediates search by offering results that are task or context aware avoiding the typical query, result, refine loop. Google Now with its just-in-time delivery model of answers exemplifies the change of user behavior and expectation – take my context do some magic in the cloud and give me the right information at the right time.

The next generation of mobile apps will perform tasks for users in the cloud. Again, taking search as an example, mobile will change the model from positing a query to automatically running a context-dependent, standing query and pushing the results to the user.

The current opportunity in mobile – until the next wave of consolidation occurs – is to create best-in-class services that take a user’s context and data and predict what the user needs to accomplish a task even before the need arises.

What’s Next For Online Shopping?

Shopping is fun. People go shopping with friends, for occasion, to discover something new, for retail therapy or just because. Discovery shopping, where consumers don’t know what they are going to buy until they encounter it, represents 70% of offline shopping.

Retailers have optimized for discovery shopping for centuries with exhilarating in-store experience, exclusive shopping events, entertainment and personal connections. Offline the focus is centered on Brand message, product showcase, and customer loyalty.

Online shopping has been falling behind. Most e-commerce stores are based on a vending machine model. You either instantly find what you’re looking for or leave. The search-based interaction model with its focus on functionality and usability is optimized for transactions and task based shopping at the expense of product discovery and impulse purchases.

The use of computers to buy goods has transformed shopping into a rational and deliberate process relying more on the brain than the heart. Arranging products according to categories triggers task shopping – customers find what they need or leave. Telling a beautiful product story motivates browsing – shoppers will hang around.

Serendipitous shopping has as goal to offer consumers a wider range of choices versus today’s shopping sites with their sole focus on price and features. Most online ecommerce sites use a search-based interaction model to help visitors find products of interest.

Future sites will not; they will offer a window-shopping experience, where a shopper spends time looking and interacting with each visual and each visual is designed with the emphasis on telling a product story and to instill the urge for an impulse purchase.

Pinterest is leading the way with an experience that allows social users to keep up with trends and to shop for ideas. The Pin action or button extracts from sites or uploads images and places them onto Pinboards that can be shared, favored and followed.

Pinterest however hasn’t come up with a strategy that places the services in the category of e-commerce marketplaces. The service lacks the basic functionality of a shopping site such as a catalog of product, a shopping cart, price comparison, or a buy button.

I am a big fan of what Facebook has achieved with the Facebook’s Like button and the Open Graph protocol and believe that like button as an one click expression of affection for a brand or product is a great signal to be exploited by a shopping engine.

What will online shopping look like in future?

Shopping Next

For me it is a combination of Pinterest, Facebook’s Like ecosystem and shopping catalog with same day delivery:

  • A highly visual experience that encourages browsing and discovery through product stories focused on trends and shopping ideas.
  • A social button similar to the “like” that connects consumers with brands and their products and turns web pages into product objects that can be aggregated.
  • A shopping engine that aggregates products and generates product stories based on the “like” signal besides offering the typical shopping comparison and transaction features.

Social Recommendations: Homophily versus Heterophily

Homophily—the principle that “birds of a feather flock together”— have been studied extensively in online network formation.

networkConnectivity breeds similarity and targeting the social network neighbors is an efficient strategy of audience identification for advertisers or for matching up people that have similar interests.

Services that use Facebook Open Graph to push out activities of their users herald this as great way to discover new and interesting objects such as movies, restaurants, and products. Netflix recent integration with Facebook is a good example.

The Facebook’s Graph Search introduction showed a similar use case: restaurant recommendations by friends. Besides the question of density of such “likes”, I question the value of recommendation services based on social activities of ones family, closest friends or colleges.

Assuming that these are the people one communicates the most with and therefore one is most similar with recommendation most likely lack diversity. Like in midlevel settlings in the countryside where people could not travel further than ~5 miles life would become very boring very quickly.

Heterophily – “the love of the different” – describes human tendency to organize in diverse groups. Network formation in this case organizes homophilious networks through weak ties into a larger heterophilious network.

Heterophilious network allow the spreading of new and unique content creating serendipity. This correlates with finding experts or taste masters in the offline world; recommendations by subject matter experts.

Algorithms that use social activity to recommend unique or tail content have to take this into account. They would have to aggregate my activities (e.g. likes) and that of my network and find similar clusters of users in the open graph. These clusters could be used to derive new and interesting recommendations analogous to collaborative filtering algorithms.

Until such algorithms come into existence I will stick with strangers on Yelp or Zagat and the friendly concierge.

The Long Tail – R.I.P.

The commercialization of the web and the proliferation of the mobile experience threaten to put the long tale of content to rest. Popular and paid content will outrank the obscure and unique.


To understand one has to look at the three ways that content is discovered:

Web Search – Algorithmic

Algorithmic is the Google way. The algorithms of search engines crawl the web, analyze and match user queries with publisher content, and serve ranked results.

Sophisticated machine learning algorithms analyze queries suggesting or re-writing them with the goal to guide users or providing instant answers. The unintended side effect is that users are funneled down the beaten path of popular (i.e. head) queries.

The commercial implication of search traffic has fostered link spam to a level, which threatens the search ecosystem. The last few Google search algorithm updates have been focused on penalizing sites that are deemed to have too many [monetized] pages and rewarding site freshness and authority for tail queries. Spam removal is good – the side effect however is that there is less incentive for publishers to produce and maintain content for tail queries.

Monetization of search is becoming more difficult because of increased competition for commerce-related queries and the disruptive effect of mobile in removing search as middleman. Google just introduced paid-inclusion for its shopping vertical and paid-placement with its Product Listing Ads, two strategies that are known to favor the head and starve off tail content.

Social Media – Crowdsourced

Social hasn’t reached the winner-takes-all state as web search yet. Using social activity – filtered by a user’s curated network of friends – to identify relevant content is however the Facebook way.

User submits, shares, comments, votes on content and Facebook through its algorithms determines what is popular for a given user and what is not. Popularity, defined by the number of content-related activities generated by ones network, does not favor tail content, i.e. the majority of people will not be heard.

Facebook recently overhauled its News Feed experience by changing the visual representation mainly by providing users more filters and by introducing larger thumbnails. Filters – if adopted by users – have the ability to create a larger surface for tail content. Larger images on the other side most likely favor head content.

The commercialization of the social stream has not reached the level of web search yet. The shift from desktop to mobile however has reached an infliction point and it can be only a matter of time until monetization pressure will change this too.

Mobile Local – Contextual

Mobile disintermediates web search. Users expect results relevant to their context and location. They are more likely to accept default results as answers rather than to search, analyze and refine. Such now-and-here decision-making naturally favors content in the head rather than tail.

Mobile apps are a convenient way of consuming content on a device, if they are readily available in the experience and have establishes a habit.  This is a steep hill to clime for tail apps and an equivalent to navigational search queries – to easily locate and access content sources – hasn’t been established.

Location, images, videos and audio on mobile on the other side can open up a complete new way of identifying tail content by utilizing technologies such as bar code reading or facial/object recognition.

Mobile hasn’t produced a winner-takes-all service yet. Web search – by not indexing and surfacing user comments in blogs – has spawned social bookmarking sites and ultimately social networking – allowing user’s to share their [tail] content.

Mobile and social media favors the popular today. History indicates that we will see new and innovative services that allow people to express themselves on mobile.

Evolution of Online Marketplaces in the Mobile Ecosystem

Based on comScore data I looked at my favorite candidates for the winner-takes-all online marketplace game:

web mobile properties

The interesting fact is that both Google and Facebook are reaching 50% of their audience on mobile now and Amazon and Apple are close behind. I added Wal-Mart into the mix because with its distribution network it could still have a shot at becoming a major online marketplaces player.

The second interesting comScore statistics looks at app versus mobile browser time spent:

mobile behavior

The mobile ecosystem shifts consumes away from web search on the desktop into a more fragmented world of vertical apps. Online marketplaces are going mobile and marketplace apps will dominate.

Consumers use only a couple of apps consistently and it is extremely hard to get into that list. One or more players out of the short list shown above will dominate shopping search using online marketplaces on mobile.

The options for startups and retailers are limited: either join one of the marketplaces as tenant, join the ecosystem (APIs) to add value on top, or innovate outside of generic shopping search.

Google versus Amazon: Same-day delivery to lure customers

The buzz around same-day-delivery has increased lately, fuelled by the entrance of Amazon, eBay and Wal-Mart into this market segment. Now Google reportedly is throwing down its gauntlet too!

Amazon push in this space shouldn’t come as a surprise. Convenience and instant gratification of immediately delivery will pull consumers into its online ecosystem away from brick-and-mortar stores.

The years Amazon has spent to streamline its business model to be run with razor-thin margins will enable it to absorb the cost for same-day or free delivery in order to gain market share.

Google, jumping into the delivery business by acquiring BufferBox to compete with Amazon Locker and reportedly building Google Shopping Express, needs context to be understood.

Google recently has revamped the experience (e.g. support of “nearby stores”) and the business model (to paid-inclusion) of its shopping vertical. It also has acquired Channel Intelligence a commerce enabler that offers feed generation, optimization, reporting and consulting.

Obviously Google is in the process of assembling a new marketplace engine to better competing with Amazon and the like:

google stack

Google uses its Product Search (aka Froogle) shopping comparison engine ecosystem to start the transformation. By changing the catalog to paid-inclusion it gains a direct merchant relationship removing free riders.

The fact that Google in that process doesn’t mind loosing Amazon as customer – a significant hit in product selection – should become understandable now.

Merchant benefit from Google’s offering, if it helps them to compete with Amazon, especially if Google initially subsidizes shipping, which is reasonable to assume.

The new Product Listing Ads that have been introduced by Google are attractive to merchants since they provide a greater control over sales – the higher a product is bid up in results the more traffic it sees.

A strategy as outlined above hits all the checkboxes for success: its starts with an existing ecosystem, it rewards early movers, and it has the chance to pull-in customers and merchants a like through network effects. It would create an attractive offering of local products and merchants.

What needs to be seen is, if it can reach economy of scale quickly enough to fend off Amazon.

The benefits of Google revitalizing its commerce marketplace should be obvious. In addition Google will gain deeper insight into the purchase process and history of customers.

Behavioral data acquired in such a way will help to improve ad targeting in its network and allows to better compete with Amazon’s and Facebook’s entrance into the market of ad exchanges.

Google Glass(es) – The Anatomy of Apps

Having information projected by a heads-up-display (HUD) into ones vision is an intriguing idea. Google’s consumer-targeted device Glass deserves credit in pushing the boundary and hopefully the product will ship.

There are many things that come to mind if one thinks about the implication of day-to-day use of such a technology, among them privacy and dorky looks. However considering smartphones and their uniqueness and social acceptance almost everywhere success for Glass is more a matter of functionality and lasting coolness.

The apps Google has shown so far haven’t convinced me that I need Glass. The sport-focused scenarios make sense since they showcase the hands-free benefit of Glass. However in most of the other scenarios I am happy to use my phone to get time, take a photo, or record a video.

It is an interesting exercise to think about what features characterize a useful app on Glass.

Task Accomplishment

At first glance one could assume that Glass as a location and context aware device would be a good fit for a traditional map application. The limited capabilities of the HUD make this questionable. I believe that functionality such as a virtual compass that communicates location information associated with objects is more appropriate.

A more interesting scenario is the lookup of knowledge about the objects one is dealing with as part of a task, i.e. searching for information about people, places, things, etc. Glass would be super useful as a context-aware knowledge lookup service.

Have you ever been at a conference and walked around wondering who all these people are or struggled to remember a face or name. Glass could recognize attendees for you and in doing so enhance the conference experience and lower the barrier to networking (versus pointing your smartphone at the face of an attendee to achieve the same).

Glass could help with its context-aware, hand-free information lookup to identify objects in your environment.

  • If you are an avid mushroom hunter, Glass could be indispensable in identifying if one of your priced discoveries it tasty and edible or not.
  • If you are hunting for your next house, Glass could provide you information about the neighborhood right there based on a real-estate database such as Zillow. No cumbersome map-based research after coming home.
  • If you are at a used car auction or a yard sale hunting for bargains, Glass could offer you value estimates and supporting information based on eBays database of past person-to-person transactions. No more buyer’s remorse.

An app would understand the tasks that need to be accomplishes, identify the things that are relevant in Glass video and audio stream, map the things against a database of information, and display the most relevant information to the user. That would be cool.

Serendipitous Discovery

Many objects in our environment are in plain sight to us but are obscured by a noisy environment and are only found by accident if at all. Serendipitous discovery describes a process in the background that facilitates chance encounters with objects that have location information.

Glass is the ideal device to create serendipitous encounters with objects in our environment based on context such as location, motion, audio, video etc. Scenarios are the discovery of a historic place, of new products while shopping, an interesting person while strolling around etc.

Serendipitous encounters require a new type of personal recommendation that take into account not only a persons history of interests and encounters but also introduces a form of randomness. The challenges to develop such a system should be obvious; too much distraction or irrelevant recommendations will cause frustration and discontinued usage.

Fluid Micro-Coordination

Another interesting app for Glass is to form groups and to facilitate coordination using hyper awareness – the awareness created by real-time communication across different locations.

Such micro coordination would allow a group to form spontaneously and converge using Glass as the application that coordinates time, place and details of and event in a fluid and non-interrupting way.

The Glass app would inferred context from a users device such as location, activity, audio, video, etc. and automatically shares it with others in a group.

The key would be the proper use of the HUD to update each member of the group in a moment-to-moment fashion about current context and location convergence without interrupting ongoing social activities.


It is great to see eBay being on the move again and coming out kicking.

During all the time I have been engaged in building, running and analyzing two-sided marketplace the original online person-to-person listing / biding model of eBay has always been my favorite. Its additional network effect of buyers becoming sellers and vise versa adds to the fun figuring out the ecosystem.

eBay doesn’t take possession of either the seller’s good or the buyer’s payment. Not having inventory or liability is always a good thing in my books compared say to the single-sided marketplace that makes up the retail business of Amazon.

A person-to-person listing intermediary always has the challenge to create trust between participants in its service. eBay’s feedback system is unique in that transactions create a history of trades for each participant. The trading profiles thus aggregated create a form of reputation and allow evaluating the trustworthiness of actors.

eBay’s challenge in competing with more traditional marketplaces has always been its overwhelming product universe caused by the uniqueness of  each of its product listings. Finding a product requires expert effort in form of searching, refining and scanning the many listings – maybe the new Cassini Search Engine will address this problem.

eBay has been a market transparency engine for unique or used goods. Its database of transactions is a tremendous asset that allows putting a price tag on everything. A good example is eBay Motors and its ability to create transparency in a market that is controlled by Kelly’s Blue Book and its valuation of used cars by insider auctions. The price setting power of online person-to-person transaction is a tremendous asset for the web.

Part of the current success of eBay has been its ventures into the non-auction shopping market targeting Amazon’s marketplace.  Hopefully this doesn’t deemphasize the investment into its market transparency engine.

It should not, since it is its unique advantage in the heated battle of two-sided commerce marketplaces that is brewing between Google, Amazon and Facebook:

  • Amazon – “database of consumer transactions” with its rich customer purchase history,
  • Facebook – “database of connections” with its rich user profile and interest data,
  • Google – “database of search purchase intent” with its last click advantage,
  • eBay – “database of  person-to-person transactions” with its pricing setting power.