[This article was originally published at IoT Agenda.]
Industrial IoT projects are conceived inside enterprises every day under a variety of circumstances. The CEO attends a conference. A competitor releases a connected product. An engineering team sees the future. Operations wants to reduce costs. The marketing department seeks new revenue. Customers ask for help. No matter the origin, an individual or small team receives a mandate to embrace the Internet of Things and create value for the enterprise.
The First Question
What insights, if achieved, would provide significant value to our industry, our operations, and our customers?
For many organizations this will be clear, but if your industry or team is new to the connected product / service provider model, the end goals may start out fuzzy. Devoting significant resources to digital transformation is unwise without a clear target in sight, but some teams may need to proceed further down the path before the answer crystallizes.
The Prototype Phase
For those with a good understanding of the desired insights, the next step is to determine what measurements must be taken to enable their derivation. Following this, you must determine whether or not the required data can be reliably obtained. A business model requiring accurate ongoing health information for Schrödinger’s Cat should not be pursued, regardless of market demand. If available sensors and environmental conditions (or quantum mechanics) prevent your current team or tech stack from obtaining critical data, the project should be paused until such expertise or advances in technology can be found or developed.
Your prototype is now up and running. Sensor(s) on a thing send data to a server or the cloud for presentation on a simple dashboard. Basic commands are sent back down to control the device. Good progress. Now the real planning begins.
Feeding the Machine
You have an incoming flow of data, and a reasonably clear targeted insight. The next step is to process and clean this data into normalized, trustworthy information for analytics tools and machine learning systems. In the real world outside of the warm lab, your production system will encompass myriad devices and shifting conditions. Sensors will fail, firmware will have bugs, message corruption (whether spurious or nefarious) will occur. You cannot feed your data scientists (or their algorithms) junk food all day and expect quality results over time.
The production system must enable the inspection and flagging of all incoming data as trusted (from authenticated source, expected format, etc) or untrusted (suspicious source, out of range values, etc). Only the trusted data is sent to power analytics and machine learning, while untrusted data is retained for review and troubleshooting. In well-designed systems, data flagged as untrusted (not thrown away!) can be can be analyzed to find bugs, detect intruders, and improve learning algorithms. In many cases, it can later be repaired and made useful again for deriving insights.
This is the riskiest phase of the project with most opportunity for failure. To scale a prototype to production does not mean just adding more sensors and devices – to scale an industrial IoT system is to add complex, robust, and secure capabilities that match the stringent requirements of large enterprise systems like Role-based Access Control (RBAC), complete audit history, and chain of custody. Plan accordingly.
Acceleration and Growth
With the architecture plans complete, you can now model the costs to build, operate, and maintain the production system over time. Business model in hand, investments and functionality are scheduled to match opportunities and timelines for cost savings and new revenue streams.
The first milestone is often internal cost savings – reduced unexpected downtime and fewer maintenance truck rolls through predictive maintenance. System infrastructure must meet enterprise requirements and account for the inclusion of additional modules in the future. At this stage though you only need to develop functionality required to fulfill the task at hand and drive machine learning. As opportunities are identified, whether algorithmically or in the boardroom, your scaffold will be built out further to add revenue streams to your organization through additional service offerings made possible by insights gleaned from incoming data. The edifice is completed, and the business transformed, with the enablement of new revenue streams for your customers.
When your connected product services enable your customers to save more lives, serve more hamburgers, transport more goods, or better fulfill their mission in any way, you become deeply embedded within their value chain and become set up for long-term success in an increasingly connected world.
The Road Ahead
The journey from traditional manufacturer or industrial operator to enterprise service provider requires diligence and expertise. With proper planning and intelligent design the opportunity (or as many are saying, the imperative) can be approached in a step-by-step manner to minimize risk and deliver increasing benefits through the entire value chain.
Advice from Experience
At Bright Wolf, we’ve helped our Fortune 1000 clients deliver successful global deployments of scalable, secure, and maintainable systems. Let us know how we can help you too.