Three Strategies for Avoiding IIoT Dead Ends


Table of Contents

In “Three Paths to IIoT Dead-Ends” we described how three of the most common approaches to turning manufacturing data into insights lead companies into digital dead ends. The first approach, trying to cobble together a solution out of legacy manufacturing software, typically leads to outright failure. The second approach, the tailored data science project, is costly, not repeatable, and challenging to scale. Finally, IoT platforms focus on acquiring data and offer APIs to work with that data, but leave all the hard work of data contextualization and analytics undone.

Sight Machine’s approach to deriving insight from manufacturing data is based on three pillars – a unified data model for manufacturing; a commitment to open technology; and a strategy of letting the problem and desired ROI drive the technology rather than starting with a technology and attempting to find a problem that suits the tool.

The unified data model — Sight Machine’s digital twin — is designed to solve the problem of context. The critical challenge in turning data into insight is neither the collection of the data nor the visualization of the results. Rather, it is contextualization: turning torrents of heterogeneous digital data into a working model of a manufacturing process.

Sight Machine starts by collecting and preprocessing the data, either directly from sensors or from a system of record (a historian or other common data store). Sight Machine then feeds the data through its AI Pipeline, which contextualizes and combines the disparate data streams into a structure that models the production processes. This structure is Sight Machine’s digital twin, a virtual replica of a customer’s plant or enterprise.

Once the data is contextualized and modeled with Sight Machine’s digital twin, it can be analyzed using Sight Machine’s extensive toolbox of statistical and analytical tools. Alternately the contextualized and modeled data can provide a solid foundation for a data science team to work with. These models not only drive ROI for the first problem, but are available for all subsequent use cases, providing rapid return for all future problems.

The heterogeneous nature of manufacturing problems means we must be open to all types of data and all applications. Limiting what data or devices a software system works with isn’t good for the customer. A true data platform should allow a combination of any relevant raw data, independent of original source or system. It should also be open on the output, to feed into other systems. These tenets were critical to the consumer internet, and will also be central to the industrial internet.

Finally, Sight Machine starts by focusing on the final objective: what is the business problem you are attempting to solve, and what is the nature of the return on investment you hope to achieve? Technology for the sake of technology can be interesting, but ROI is critical to any manufacturing project. We encourage our clients to identify a clear business problem to solve embarking on a digital transformation journey.

We see a lot of companies start with the technology first, spending a lot of money and effort to implement a technology expecting to get great results, but then finding it a struggle, or the reach limited.

We believe if you start with ROI and figure out the context, the tools and analytics will follow from there.

Ryan Smith

Ryan Smith

VP of Product and Engineering at Sight Machine Ryan has expertise in manufacturing, life sciences, and automation hardware and software. He has developed and implemented robotic inspection systems and real-time surgical navigation software at leading life sciences companies.

Curious about how we can help? 
Schedule a chat about your data and transformation needs.