Why does manufacturing analytics not live up to expectations?

We think there are two driving forces here.

The first is lowered expectations. The granddaddy of manufacturing analytics was SCADA, and SCADA is ancient.  

The second factor is that getting insights from complex systems is hard. Really hard, like self-driving car hard. Your job as a plant supervisor is not an easy one, and completely automating that process with all your know-how is not something easily done.

‍

Many analytics companies work around this complication by limiting the scope of the problem to such a narrow one that the benefits either marginal or only specific to the use case and not the overall system.

As an aside here, many companies treat these “analytics products” the same way they would a new machine. An independent component of the entire system.

So let’s look at a manufacturing analytics system to see what you can, and what you should be able to do.

What is manufacturing analytics?

Manufacturing analytics covers the use of data from a range of sources, including operations, machines and sensors, and other data systems such as ERPs to optimize output.

In many cases, it’s not for lack of data that companies have been slow to adopt these techniques, but rather a lack of data integration. Generally speaking, data is spread out across a wide range of sources and sensors.

What should you be able to do with manufacturing analytics?

There are three well known use cases for manufacturing analytics:

• Predictive maintenance
• Operational efficiency
• System optimization

Predictive maintenance is a method of identifying machine downtimes and failures before they occur. It is the successor to preventative maintenance (and its predecessor, reactive maintenance). Whereas preventative maintenance requires scheduled downtimes to perform repairs, predictive maintenance uses data from a machine to predict when a potential failure will occur.

Operational efficiency refers to techniques focused on improving yield and/or throughput for a plant. Techniques here will focus on identifying optimal configurations and settings for machines to ensure that waste is minimized while maximizing total uptime. Many manufacturing analytics tools focus on making key performance indicators (KPIs) more easy to evaluate and review. For example, OEE is a common KPI that is commonly automated with manufacturing analytics and included in manufacturing dashboards.

System optimization refers to techniques that improve systemic aspects of a business, such as supply chain management, sales mix, production planning, etc. They often include components of demand forecasting and inventory management.

Many tools are available to help manufacturers perform some aspect of these use cases, but in our opinion they often miss the mark. Meh.

Why does manufacturing analytics often miss the mark?

At its core, manufacturing analytics should surface insights to help companies identify opportunities for improvements - reducing costs and increasing throughputs.

The key concept here is surfacing insights. If the tools don’t do a good job of identifying insights in a timely manner, then people can’t make improvements.

This is worth repeating. Manufacturing analytics tools must:

1. Identify meaningful insights (cost reduction and/or increased throughput)
2. Share them with the right people
3. Do this in a timely manner

Timeliness is perhaps one of the most important components. Arguably in the long run, getting average insights faster is more important than spectacular insights after the fact.

So how can you meet all three criteria?

With a data-driven culture, of course.

The  data analytics maturity model allows us to grade companies based on their level of data proficiency. Generally speaking, it is better to walk before you can run, and with data the same holds true.

The four stages of data analytics maturity are as follows:

• Descriptive Analytics - Reporting on what has happened.
• Diagnostic Analytics - Understanding why something happened.
• Predictive Analytics - Predicting what will happen.
• Prescriptive Analytics - Looking to see what we should do.

Companies looking to perform advanced manufacturing analytics use cases often make the mistake of jumping ahead to predictive and prescriptive, when they haven’t yet implemented the necessary infrastructure to perform the basics.

Manufacturing analytics vendors often cheat here - they either narrow the scope of their offering so much as to be of marginal benefit for the larger business, or they offer mediocre insights that are nothing more than just copy-pasting raw data from the plant.

There’s a much better way.

Check out Manufacturing Insights. Or better yet, start with the basics and get a Data Coach under your belt. It’s an independent assessment of your data proficiency with a roadmap for improvement.

Manufacturing analytics, like all data projects is a journey, not a product. Make sure you work with a partner that knows how to guide you on this journey.