6 Considerations for a Successful Enterprise IoT Strategy

From having adequate manpower to planning for unpredictability in a mass-deployment environment, organizations need to cover their bases to use IoT applications successfully.

The current generation of IoT use cases revolves around optimization of existing processes, such as collecting smart meter readings. IoT applications may struggle to offer real value, since potential per-deployment savings may be small. For example, in the U.S., the 10-year cost of reading a water meter might be $120. This means the 10-year cost of an IoT meter solution would need to be substantially less than that or offer dramatic benefits to be justifiable. As a consequence, many IoT business plans are wildly optimistic.

IoT data’s value can be very asymmetric—it’s worth different things to different people. For example, utility meter readings at 15-minute intervals are a source of entertainment to home owners, are useful for load planning for the power company and can also reveal when the home is empty to criminals. IoT data’s value is also related to how easy it is to join with other data streams and whether this can be done in real time or not.

Storing data is easy, but accessing it in a coherent fashion afterward is a challenge. Consider whether the data is being collected for data science or operational purposes. Data scientists want untouched raw data streams, but such data is unusable for monetization purposes. Enterprises that want to monetize IoT will need quality, enriched and value-added data. The same considerations that applied to legacy extract, transform and load (ETL) systems apply to operational IoT systems. Just because they use Hadoop instead of a legacy relational database doesn’t mean that 40 years of experience and knowledge related to data processing suddenly becomes irrelevant.

IoT applications differ in that it is generally not possible to “turn off” the IoT. As a consequence, a different operational mindset will be required because customers will be unhappy if their IoT devices stop working for any reason. The machine-generated nature of traffic also will be a challenge—such data can be very bursty and devices will tend to keep sending messages until they get an answer. Service-level agreements (SLAs) will become important. The de facto SLA for older applications is often the human short-term memory limit of around seven seconds. If IoT devices behave like telephone switching and charging equipment, we can expect a de facto SLA of between 50 and 150 milliseconds before the devices start to misbehave.

Stored historical data may be a liability instead of an asset for organizations. Don’t let enthusiasm for IoT technologies hide the reality that they are going to be custodians of the most intimate details of people’s lives. From a device perspective, the industry tends to pretend that security is someone else’s problem. Trying to reconcile the rapidly evolving state of security with the difficulty of upgrading deployed devices, while anticipating consumer expectations for minimal configuration requirements, is close to impossible.

Developers need to consider the implications of mass deployment in real-world environments. IoT devices will have to cope with a wide variety of environments, and especially with erratic connectivity. Developers probably won’t be able to control or mandate the environment in which the device is used. If they are successful and end up with millions of deployments, they will find that everything that can happen will happen, and some things that can’t happen will happen as well.