Manufacturing analytics: Why traditional Buy vs Build analysis doesn’t work

buy vs build
Facebook
Twitter
LinkedIn

Table of Contents

There has been much guidance written on performing buy vs build analysis for enterprise software. I like the simple summary created by the Technology Consulting team at Carnegie Mellon University which you can find here. But for manufacturing analytics projects, the answer is not as straightforward.

The manufacturing analytics software market is in its early phases. There are a myriad of solution providers saying the same things, making it difficult to evaluate off-the-shelf software functionality. Additionally, most products available in the market only do a portion of what’s required (such as data aggregation, analytics, or dashboards), resulting in a lot of build activity even when one is attempting to buy. Finally, most vendor solutions haven’t been widely deployed, so understanding the true scalability and support costs is difficult to quantify.  

In my experience, the most successful projects involve a combination of both build and buy – enabling manufacturers to leverage the innovation and scalability of solutions in the market, while building off of their in-house expertise on production and process analysis.  This is a platform approach — finding a software framework that you can build on top of and greatly accelerate in-house efforts.

Challenges in building and buying

Becoming a data driven manufacturer involves multiple steps, including: machine/IoT data acquisition, data preparation/contextualization, data modeling, analytics, visualization, and integrating output into workflow. These steps are just precursors to the associated change management and business process reengineering required to truly transform an organization.

I’ve worked with a number of manufacturers that have attempted to build out all of these elements internally. These DIY projects are often driven by an internal innovation team, which attempts to solve a single problem as a proof of concept. As digital manufacturing is so new, there is usually no one running the effort who has done a similar project before. Lack of experience leads them to underestimate the length and complexity of the projects, often resulting in frustration and ultimately failure.

But I’ve also seen a number of buy efforts that have led to dead ends. Most of these initiatives involve the use of packaged solutions not purpose-built for manufacturing.  These either provide limited business value or end up in pilot purgatory. Examples of these include:

  • IIoT data ingestion solutions that acquire and collect data but don’t structure it in a way that supports decision making
  • Off-the-shelf BI-tools which can do basic visualization but don’t have a rich capability for analysis of time series data
  • Solutions that focus on an individual asset or process step, but cannot be expanded to integrate the full enterprise

So what’s an approach that can work?

Manufacturers need to leverage the power of the market to create a platform that builds upon industry driven innovation while ensuring that the solution is flexible enough to seamlessly (read, no expensive integrator work required) incorporate their team’s expertise and respond to their organization’s changing needs.

Let me provide some examples of what I mean.

Machine data acquisition

Machine/IIoT data acquisition and data visualization are well defined markets, with numerous offerings from a number of solution providers. It makes sense to leverage the capabilities of the market to ensure that the widest set of machine protocols can be ingested and visualized. I’ve worked with many plant IT teams that have attempted to piece together the data acquisition and integration capabilities for their production equipment. But processes and equipment will most certainly continue to evolve, and corporate acquisitions of new facilities will assuredly introduce a new protocols and standards. Let the market help you stay current.

Data modeling and contextualization

Production data modeling and contextualization is complex area, and most manufacturers don’t have the expertise to build these capabilities in an automated and scalable manner. Much of that upfront work consists of turning raw production data into a snapshot of the production process. A company may have dozens of data sources and thousands of sensors along a production line. In order to turn this raw data into something actionable the data must be prepared, blended, and transformed – optimally in real time.

Even manufacturers that have specialized resources to pilot efforts in this area, will be challenged to continue to invest and innovate on an ongoing basis. Here at Sight Machine, we’ve been working on our patented AI Data Pipeline, which uses AI to automate the creation of digital twins of production processes from raw machine data, for over five years. Our developers, data engineers, and data scientists have a deep understanding of how the latest machine learning techniques can be applied to manufacturing process data and are building upon years of learnings from engagements with the leading global manufacturers.

So where does leveraging your internal capabilities to build make sense?

One of the reasons off-the-shelf solutions have nominal impact is their limited applicability to the unique complexities of a given manufacturer’s production environment. To achieve scale, many solution providers end up focusing on a narrow use case, such as anomaly detection for a given machine. But these types of use cases tend to have limited transformational capability and value.

To deliver the greatest impact, manufacturers need to leverage their team’s expertise in understanding and optimizing production, and incorporate their expertise and capabilities into their digital manufacturing platform. I’ve seen the biggest value delivered from solutions that incorporate this expertise into platforms that ingest, blend, and model data.  This automates the data exploration that process experts would pursue themselves, if they had the time and capability to manipulate the data directly.

Manufacturers need to ensure that the platform they invest in enables them to maintain the IP unique to their processes while taking advantage of the markets IP for automation and scale. Investing in a platform that allows for easy integration of custom analytics and algorithms developed by the in-house team will enable the manufacturer to build competitive differentiation without the need to build out an expensive in-house software development organization.

By combining a manufacturer’s in-house expertise on production and process analysis with the innovation of the market, manufacturers can deliver an impact that will scale with their needs. This is the leverage that a platform approach provides to an internal team, and lets them focus on where their skills and expertise are most valuable.  To learn more about how manufacturers are optimizing the mix of DIY and commercial software for manufacturing analytics visit the Sight Machine use cases page.

Why Sight Machine Uniquely Answers Your Manufacturing Needs Better manufacturing analytics across the enterprise The only AI-enabled platform for digital manufacturing built from deep manufacturing experience, understanding, and a history of results Learn More

Nate Oostendorp

Nate Oostendorp

Co-Founder and CTO at Sight Machine Nathan co-founded Slashdot.org, worked 9 years as an architect for SourceForge.net, and has developed several other successful online communities. Nate has also worked in industrial controls. He holds an MS in Information Science from the University of Michigan and a BS in Computer Science from Hope College and he has contributed to Forbes.

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