Some technical recitals
The "network edge" as it were can take on many different forms: it can have a variable population and have a shifting edge, or it can have more of a static population and be geospatially fixed. An example of a variable-population, shifting-edge network is something like Radio Frequency IDentification (RFID) which sticks to shipping pallets loaded with inventory moving between companies. Another example of a variable-population network is obviously the Internet. An example of a fixed-population, geospatially-fixed network is your office local area network.
What are some of the investment problems here?
The complications with a shifting-edge network are many: (i) the edge devices (these are wireless) have to be constantly "visible" to a radio receiver which means receivers have to be everywhere or transmitters have to be very powerful. In either case these requirements increase the cost of deployment to the user. Otherwise you can go to a mesh network where all the edge devices are transceivers (i.e. talk to each other) - the problem with this is that one can't do proximity measurement for geospatial plot (i.e. you will not know exactly where your stuff is). As a rule, the more powerful things are the more power they consume and otherwise have a tendency to be more complex. This in turn makes wireless edge devices more difficult to maintain. This was part of the problem causing the histortically slow adoption rate of RFID. Yet another another issue for investors in RFID was that the world had previously implemented a lower-tech version of the variable-population, shifting-edge network called barcoding (is that what that was?). So since there were a number of pre-RFID inventory and supply chain applications that were geared toward this type of near-real-time-data [barcode], then RFID-based data sourcing was morally more of a product evolution for these barcode-sensitive software applications than a revolution/creation of a new software space based on true realtime. As a result, we have seen a number of RFID middleware application companies coming to life [that link to the pre-RFID apps.] but not really any broad innovation of new business management applications. In the venture world, we have yet to hear the sonic boom of investment returns from RFID-related software at any level.
So what can we learn from this?
The physically distributed energy industry in which we invest has many different needs for operational data and thus a great need for widely distributed networks - some may be variable-population, shifting-edge some may be fixed-point. What we know is that the edge devices had better be low cost to deploy, very low maintenance ("thin" to use an industry expression) else failure rates cause adoption rates to creep at best (see history of RFID).
What we also know is that the data characteristics had better reflect a step change to what the industry is accustomed or otherwise be wrapped around business process that is not currently software-application centric. If not, we are back on the evolution path as investors rather than the revolution front and making very little money as investors in newly-developed software.
One side note and one prediction
The side note - New purpose networks almost invariably start out as device manufacturing entities whose goal it is to get into network services business for the purpose of getting into the software business. Unless you have something special at the networking device/management level...you'd better be planning version 1.0 of your business process software system. I've seen it too many times.
The prediction - Operational intelligence networks today are sold to customers for their internal ownership and management. This is the equivalent of building and maintaining your own phone system. I predict this will change. As the novelty of the new networks diminshes and their size grows they will of course be flipped back out to the technology providers for management - because managing a 1,000-node network is hard to do. This means again that the whole architecture of the network should be geared toward "low-touch" design as opposed to fast-paced version obsolecense revenue models (e.g. desktop computing). This will be a point of distinction between success and failure of many companies.
Some conclusions
In my judgment, there is a significant investment opportunity in the energy industry around real-time, operational intelligence networks. Why?
The raw material underlying the oil industry is (i) flammable/toxic, (ii) fungible, and (iii) is a commodity reflecting material, daily price volatility. This means there are three different motives for knowing exactly how much "product" is where at all times (i.e. environmental/safety, theft, and financial risk). Interestingly to investors, this means a sale to any one of these three corporate interests is possible which creates sales model options - this is good. As a result of the raw material characteristics and where it is found there will be a need for special types of network edge devices that are very rugged, very low maintenance and can operate in volatile environment and still get the data where it needs to go.
These new data types for the oil industry applications will be materially different than those to which the industry is accustomed. They will be instantaneous. As a result there is an application play here. This application play will possibly be a new breed of software required because this new type of data didn't really pre-exist and thus legacy apps. aren't "barcode enabled." As a result, they will have to be created anew and will have to think differently.
