Companies today are grappling with the Internet of Things (IoT), a large network of physical devices that extends beyond the typical computer networks, encompassing devices, industrial equipment, sensors, and extended products. For some manufacturers everything they build could feed into IoT, from cars to buildings or even consumer products. While vendors like Cisco Systems, Dell, IBM, Hewlett-Packard, Intel, and Microsoft are all part of the IoT land grab, it is important for end customers not to focus their strategy on the products, technology, or pieces. Instead of focusing on the how of IoT, customers need to be focused on the what of IoT—namely the data. All of the strategy and shiny objects in the world won’t help if the data isn’t accurate, secure, and actionable. The data should always drive the strategy; the implementation tail should not be wagging the data dog. This strategy needs to start at the business level based on identifiable business needs and then filter down to the products.
Accurately Measure The Right Things
A very metrics-driven former employer of mine had the motto “if you can’t measure it, you can’t manage it”. While this was a very true statement, if you can’t measure it accurately then there is probably no point because bad data leads to bad decisions, especially when one doesn’t realize the data is wrong. A great example was the Air France flight 447 crash where pilots apparently pulled back in the middle of a stall because their sensor data told them that they were going too fast. Ultimately this bad data may have caused them to make the situation worse. Just having everything instrumented and feeding data back does not guarantee accuracy. Data is only as accurate as the system and sensor, combined with some sort of actual business knowledge that can help separate the signal from the noise. As an experiment on data accuracy, I used the calorie data of 3 different sensor / systems to track the calories on 5 different bike rides. Using a Fitbit (with heart rate monitor), a Garmin (with heart rate monitor), and Strava (which takes the Garmin’s data feed), I ended up with 3 different sets of data for each ride. Even the two that shared data couldn’t agree on the same outcome:
(Source: John Fruehe)
When only one set of data is available, it’s always assumed to be accurate; but even a single set of data seen through two different lenses can deliver different results. While the general trend of my data holds, it is not correlating perfectly. Which is right? Probably none. For my purposes it might be noise, but realistically, for your business, if you can’t trust the accuracy of the data, will it enable you to make the right decisions? Building a strategy on less-than-accurate data yields questionable results.
Assuming that you’ve done a good job of managing your sensors and input to guarantee the accuracy of the data, the next step is ensuring that your data is securely collected and the integrity is maintained as it moves throughout the system. In IoT, businesses may be starting with a sensor that could be in the facility or out in the field, remotely operating. Data is collected, brought through a gateway, and then moved through a series of servers and applications before it finally comes to rest in the middle of the datacenter, where it can be analyzed and acted upon. Each step along the way requires security and consistency to ensure that the data integrity is maintained. Within the datacenter there is much focus on security and maintaining systems, but often at the sensor / collector level or the initial gateway (which could be in the cloud) data may be more vulnerable. Vehicles may be sending loads of telemetry and service data back to manufacturers, but as we have seen, the vehicles themselves—the host of the sensors—might not be that secure. From a data standpoint this insecurity could lead to erroneous or altered streams that could compromise the validity of the data coming over the gateway. One cannot truly ensure the accuracy of the data unless it can be secured through the whole process. This is often a challenge as security embedded at the sensor level is often the most difficult to implement and maintain; especially, as in the case of vehicles, if the manufacturer loses physical custody of the product to the end user.
The Data Must Be Actionable
Data, data everywhere, and not a drop to drink. All of the data in the world is useless if it can’t be acted upon by the business. This is where the business units (not the IT department) need to be driving the IoT strategy. There is zero value in implementing IoT if there is not an analytics engine behind it to make sense of all of this data being captured. Then, beyond that, the company needs to be able to take action on any outcomes from the IoT data. It does no good to capture large amounts of data, structure it, process it, and then just file it away. IoT data should be used as a real-time feed to help hone decisions and accelerate action, driving agility in the company. There may be instances where collected data can actually create a liability for the company. The National Security Agency received much criticism for collecting mounds of data that still sits unanalyzed because they either do not have the resources or do not know what they may/may not be looking for. And recently, Ashley Madison revealed that a hacker potentially had access to not only embarrassing profiles of their customers, but also profiles that the company had charged individuals to delete but had never actually deleted the profiles. Data being collected and stored without an actionable purpose becomes a liability for the company, not only from the cost of having to store it but also from the perspectives of both legal liability and opportunity cost. In a world of tight IT budgets, opportunity cost can be a bigger driver than many think as every dollar spent to store unused data is a dollar that is not available for other projects with an actual ROI.
Where This Leaves Us
As businesses begin to sit down with the purveyors of IoT solutions, be it companies like Dell or Hewlett-Packard on the systems side, Cisco Systems or Intel on the networks and gateways, or companies like Tableau Software Inc. or SAP SE for the analytics, it is important to think through the real aspects of what you are trying to accomplish and how you plan to use the data. With multiple handoff points you’ll need to ensure that you are securing and maintaining consistency of that data for a clean chain-of-custody on the information. Finally, don’t just view IoT as the means to an end for near-term decision making. Keep in mind that the data being generated may need to live for long periods of time, and your corporate data handling and data retention policies may need to be aligned to the new reality of this type of data, because it is a whole new world in IoT and the data old rules may not cleanly apply.