The Blind Men and the Elephant parable, from Indian origins, reminds me of how I see the IoT private industry fail in their frustrating attempts to integrate a profitable network. The parable goes like this: it is a story of several blind men who touch various parts of an elephant to know or learn what it is. To one blind man, he identifies the leg as a pillar. The tusk feels like a spear to another. And so forth. This East Asian parable is used primarily to provide insight into the value of relativism of information, the need for communication, and the emphasis to reach the totality of truth. In other words, failure to collaborate and integrate information produces an imperfect world. Without integrating the whole IoT network, one cannot reach the final objective: increasing actionable value from Big Data.
In the case of IoT, sensors, and Big Data, I have observed that the HW industry cannot vouchsafe the quality of information from its data sources, in most cases, sensors. For example, in one case, I noticed that one specific sensor supplier became the critical path for data in measuring human breath compounds. But the quality of that data had not been confirmed by scientific or medical standards. With no medical expert involved with the sensor design, the sensor manufacturer provided whatever data with no bearing on the diagnostic objectives. Since I have seen various studies regarding the human breath, I observed that there were a wide variety of volatile organic compounds and other molecules, totaling over 2,000 according to one study. Those studies relied on the large, expensive GC/MS machines. How would a minute and simple sensor calibrated to measure one set of molecules, in this case, be able to provide an accurate reading that many compounds identified by the standard equipment defies logic. Note that these observations were made about the SV company, Theranos, where the HW did not consistently deliver accurate results.
Now, the sensor manufacturer has one objective – to sell sensors, and that is the manufacturer’s agenda. But it was up to the product designer and integrators to design the right HW/SW tools and actionable dashboard needed to attract and retain consumers. The IoT product has to accumulate accurate data and provide actionable intelligence for the consumer. Now, based on one IoT component provider relying on a single source sensor supplier, it will deliver inaccurate results – false positives or other defective data.
Is accurate data in IoT important? Recent studies indicated that anywhere from 44% to 50% of the purchasers abandoned their wrist health monitors within 6 months because the consumers concluded that the devices were inaccurate. I was not surprised, since I have an acquaintance who, in testing his multi-faceted sports wrist sensor, discovered that many variables impacted the quality of the data. Hairy wrists, dark epidermal, and even sweat produced many false readings. In other words, what is the point of accumulating big data that is grossly inaccurate?
Another factor not taken into the critical path of an IoT network is the integration of telecommunications and data storage technologies – indispensable data management and transmission tools. Data simply does not rise into a cloud and then magically turns into actionable data. The data must be directed to hardware within data centers — specific servers containing the database applications and storage capacity in order to store and process the data. That means leasing racks at data centers and these racks have to be close to data sources and access points. For example, Netflix sets up its CDNs close to the demand for its content, not in one central location at Los Gatos, its corporate headquarters. Data has to be posited in a network efficiently. Look at the Facebook huge headquarters – 90% of the employees are managing the hardware so that the Facebook software performs flawlessly. Without hardware, there is no software.
Then comes the software development itself. How will the software be designed to provide actionable data for consumers or clients? How does it analyze the data efficiently and appropriately? One big data analyst related to me an interview question of how to analyze data for a San Jose police force regarding criminal activities in the district. He recommended to use “heat” maps to show the criminal activity. I pointed out that the example failed to provide actionable data. The focus is to provide actionable data: show probability of felony and misdemeanor crimes and link those results to place police closer to high priority crimes. Here I observed that the data analyst did not understand the appropriate algorithms for the data. It must be frustrating to get to the final step in setting up an IoT network and not be able to gain any actionable value from the data.
Without integrating the three major IoT components appropriately – the source of the data, the sensors, telecom/data design, and the appropriate big data algorithms – what do you have? Blind men touching elephants. Each team places priority on one part of the IoT data stream and forgets to link and integrate the other components into one fluid system for actionable information for consumers.
Does the IoT deliver the right actionable data to the right audience? Big data for the medical industry is valueless if not providing the actionable data for the healthcare provider, as an example from before. Yet as of now, I notice so many missteps in IoT integration. What will your actionable data do? Therefore, these diverse teams need to be coordinated and integrated in their efforts to create an actionable IoT data stream.