Digital transformation failures are all too common in manufacturing, with failed deployments, costly projects with no return, or initiatives that weren’t adopted. Much has been written about mis-transformation, often focusing on change-management strategies. And while change strategy is important, there are other reasons why transformations fall short. At Sight Machine, we have helped companies digitally transform their plants, and we understand the elements that led to their success.
From blockchain to virtual reality, the latest tech trends pique the interest of companies facing fierce competition. Consultants charge high fees for strategic advice or operational assessments but may not define a vision for generating value on the shop floor. The result: missed opportunities, selection of the wrong technology, and failure to generate ROI. At Sight Machine, we have observed that companies succeed by focusing on four essential elements of digital transformation:
- Building a consistent data architecture
- Having well-defined use cases
- Understanding the business case
- Leveraging champions to drive the transformation
Making IT data OT ready at scale: an-often incomplete view of required data architectures
We encounter many manufacturing and consumer goods firms that have successfully deployed two to five use cases at great expense over a couple of years. But scalability remains the biggest challenge they face in their digital transformations and is often linked to the data architecture and choice of software.
Understanding the data ecosystem needed to leverage production data can be difficult, especially for non-IT people. Many wrongly believe that analytics is the complex part, while in fact there are many simple-to-use solutions on the market. Another common belief is that visualization tools can exploit SQL databases or data warehouses through data pipelines. But with that approach, data pipelines or models must be built one by one. And as manufacturing data ecosystems constantly change, these pipelines require intensive maintenance. If a manufacturer owns dozens of plants with hundreds of machines, transformation at scale requires tremendous resources to maintain these pipelines.
To address that challenge, Sight Machine focuses on a forgotten layer of required data architectures: we call it the sense-making layer, which provides a common data foundation. It sits between connectivity and visualization/analytics and is the key to scalability. Instead of focusing on machines, it connects to main data sources to build a real-time, contextualized, and structured data foundation. What distinguishes Sight Machine’s platform is the following:
- It is an actual product, with pre-built manufacturing-specific data transformations: it can be configured without coding, and avoids deployments that turn into costly data science projects
- It builds a single, unified data pipeline for each company, as opposed to multiple ones for each use case or machines that would require intense maintenance
- It takes a data-first approach, with data at the center and software built around drawing insights from data, rather than building point solutions or applications for each single use case or machine on the shopfloor.
- It creates a data foundation that contextualizes data using common data models. These allow you to explore data from different perspectives (cycle, batch, product) without building custom models
These specific points are what make a solution such as Sight Machine scalable. It contends with changing data ecosystems agilely and cost efficiently.
Don’t wait for perfected data before starting the journey
Some organizations try to reduce risks by establishing a state-of-the-art IT ecosystem before starting to exploit existing data. An effective data architecture will never be frozen but constantly evolves as technologies emerge. Sight Machine’s recommendation here is to get started with existing data to initiate the digital-maturity journey. Waiting for a hypothetical readiness can only delay productivity gains that could be unlocked by leveraging available data. There is no perfect data ecosystem to begin with; the key is to iterate and improve upon experiences and getting started is the quickest path to value.
Pilots are built with the mindset of planning for failure
Another way that risk avoidance impacts digitalization is through pilots. The manufacturing industry has a deep pilot culture. Most pilots, though, lead to disappointment, as the results do not match expectations. Why is that?
Pilots are usually seen as low-cost, low-investment, low-risk trials of a technology. It is essential to understand that most pilots are built with the mindset of planning for failure—not success. The conventional wisdom is when a pilot fails, it won’t cost the organization much. Pilots are therefore constructed with insufficient scope, which reduces the chances of achieving meaningful results. You could create a pilot, for instance, to identify quality issues of a specific asset on a production line. But what if the defects on that asset are linked to machine performance upstream, or to an external factor such as humidity? You could explore data without gaining insight — the scope of the pilot being too narrowly constructed.
A pilot designed to test an analytics solution delivers at best some interesting visualizations, rarely any insights. It will hardly allow for other critical aspects to be explored, such as platform maintenance, scaling up, data-model accuracy, or efforts needed to develop a data-driven culture; these require much longer duration, larger scopes, and larger volume of data. Even worse, a poorly scoped pilot can engender internal opposition, as the lack of value emboldens change-resistant stakeholders to oppose digitalization.
The best performing manufacturers don’t do pilots: They deploy technologies that deliver value, engage appropriate resources, and garner leadership support.
Don’t be led astray by a technology-first approach
An engineering culture can also lead manufacturers to have what we call a technology-first approach, where there is more interest in a technology than in the benefits it can generate. A manufacturer, for instance, buys a technology without first clearly defining a need, which results in a deployment without the right objectives, context, team, or location. Operational benefits then become hard to identify, causing leaders to question the technology.
An example we encounter at Sight Machine is with requests such as “do you do Artificial Intelligence?” A better way to frame the question is: “Can you help us reduce machine downtime or increase throughput?” The incentive is to buy a capability, not resolve specific issues, leading to potentially acquiring the wrong solution. While Sight Machine does offer descriptive statistics, statistical inference, and AI-based applications, we’ve seen that our users often improve OEE by an average of 2 to 5 percent using non-AI applications, demonstrating that success is not a technology question. Rather, success is based on a clear understanding of how software and data can address specific use cases.
In a similar way, IT and OT departments sometimes do not collaborate, and the purchasing and decisions are made by the IT team. In this case, software is assessed from a data and architecture perspective, with data engineers looking for the tool that fits their requirements, without considering the shopfloor and operational end-users. Customized solutions or toolkits, while offering great functionalities, may however not offer the fastest path to value. Rather, solutions that have out-of-the-box features could save millions of dollars and years of development. This technology-first mindset illustrates that manufacturers are deluged with Industry 4.0 reports, buzzwords, and overly simplistic descriptions of the latest “magic” they must adopt to be successful.
How to increase chances of success
Of course, before starting a digital transformation, you need to set up your organization for success. Building a cross-functional team with both IT/OT and operations, getting the blessing of executive leadership, allocating budget and resources, defining a vision, and having the right people to champion projects are all important. Beyond these, based on Sight Machine’s experience, here is a list of other success factors:
- Start from the use cases: what problems do you want to solve?
- Avoid pilot purgatory: fully engage resources and expertise in implementing digital solutions, focused on solving problems.
- What is the business case? Have you assessed the losses or gains you intend to decrease or increase?
- Are you willing to take risks?
A risk-free transformation doesn’t exist, so it’s better to consider possibilities and opportunities than to fear failure or change. The most successful approach is to embrace risks as part of the digital journey and gain experience along the way. Failure and setbacks are part of the path to digital maturity, and you should not allow the fear of failure to deter you from reaching the ultimate goal: digital transformation. In the end, that is our era’s gold rush—our El Dorado.
