Designing your manufacturing data strategy

Isaac Sukin

Having the right data is crucial to keeping yields high and manufacturing schedules on track. That’s because the right data at the right time allows engineers to quickly find and fix issues. However, deciding to track data is only the first step. In this post, we’ll cover the other questions you should consider when designing a strategy to collect the data that keeps your team productive.

How do you want to use the data?

This significantly affects the other questions you’ll need to answer to have a successful data program. There are a few different patterns for how data is used:

  • Mission-critical: data that your team can’t wait to use as a key part of their jobs on a daily or weekly basis, like reliability results. It’s clear who will use it. You may not know the specific improvements it will lead to, but you probably know at a high level how it will help.
  • Event-driven: data that a subset of your team looks at as part of researching next steps when an event occurs (for example, when investigating failures in units from the assembly line).
  • Background: data that would probably be useful, so you might as well collect it, but there’s no urgent need to act on it. An example is the temperature and humidity near a glue machine. This type of data tends to get neglected unless a leader specifically assigns the responsibility to monitor the data and take action on it. If that happens, it often leads to useful optimizations or at least lessons about how things are actually working.

In addition to use cases you may already have in mind, you may be surprised at new applications your team comes up with once they have more information to work with.

Many product companies are used to setting up functional performance tests on their assembly lines. It’s important to keep in mind that the difference between tests and data is that tests have to be set up to check specific attributes in advance, whereas data maintains a historical record that can be put to use at any time to solve unanticipated problems.

What do you need to track?

Your team probably has some ideas about possible risks that they would like to monitor. For example, if you’re concerned about a particular assembly step, you probably want to collect signals from before and after that step to help you identify problematic trends. This is a good starting point, but you’ll also need to decide how much richness you need from your data. Functional performance tests are an essential tool for tracking anticipated issues – but the data is not very flexible, especially when the results are stored as a “pass/fail” status rather than specific measurements. For unanticipated issues and digging deeper into the root causes of anticipated issues, images are an underutilized data type that maximizes efficiency by allowing engineers to go “back in time” and virtually disassemble units. In order to make sure all this data is actually useful, you’ll also need metadata like related unit serial numbers and inspection timestamps.

If your team will be directly responsible for all aspects of managing the data, you’ll need to figure out how to structure the information you collect in a way that makes it easy to manipulate later. And how much of it do you need? As the number of units you’ll produce and the number of metrics you want to track increase, so will the complexity of managing and using the data.

How will you collect the data?

Most equipment commonly used in assembly lines generate data of some sort. Sometimes that data goes into a manufacturer’s shop floor system, but most of the time it simply remains on the machines until it’s overwritten. Even when it goes into a shop floor system, engineers at product companies tell us that they have a hard time getting their information back out. Plus, sometimes those systems are too rigid for the desired use case. As a result, if you want to keep structured data without a third-party vendor, you’ll either be negotiating with the manufacturer (and even the largest companies struggle with that) or designing your own collection and extraction system. Such a system must be robust to internet connectivity issues, especially in China. Additionally, designing collection systems is its own art that requires investment to get the highest-value results. For example, Instrumental’s imaging stations have proprietary lighting systems to ensure that key features are illuminated well and evenly.

How will you store, process, and secure the data?

Many hardware companies understandably focus their engineering resources on their product rather than making large investments in building in-house expertise to create and maintain secure big data systems. Managing big data securely is not as simple as just putting engineers’ iPhone images into a shared FTP drive or dumping parametric data into an internal database (especially from a factory in China). It is essential to have both the security expertise and processes on the team maintaining your data to keep properly updated, firewalled, and monitored. Systems that are sufficiently scalable and robust to engineering and infrastructure failures are also expensive to design and implement.

How will you make the data actionable?

Storing the data isn’t enough; you’ll need a way for your team to access the raw data and reports in order to get value. If you’re managing the data on your own, that means you’ll need to invest in a user interface or data visualization software. Think carefully about how you can present data through a user interface in the most productive way. For example, a fast, powerful search feature can determine whether the tool will be used constantly or only in desperation. Good search quality can require a dedicated search infrastructure, which in turn requires its own expense, maintenance, and expertise. Similarly, data lookup is good, but comparison is better. Optimizing for serendipitous insights and discovery requires careful design.

What next?

If all this sounds complex and expensive, don’t worry – we can help. Instrumental’s secure and scalable data infrastructure was created specifically for product companies seeking to end manufacturing delays and keep yields high using intelligent data. Keep your team focused on what they do best, and let ours help them get the data they need to be successful.

Contact us today to get started, or subscribe below to receive more posts like this one.

Related Topics