Angelo Hulshout is an experienced independent software craftsman and a member of the Brainport High Tech Software Cluster.

31 March 2022

Angelo Hulshout has the ambition to bring the benefits of production agility to the market and set up a new business around that. Always have a good look at what you want to learn and improve.

After last month, when I wrote about it, we collected a lot of data, from our ERP and MES systems, from our PLCs, and also some additional notes from factory workers. That’s great, isn’t it? Now we can start improving our production facility, be more effective, reduce waste and energy consumption and be best in class.

But wait, where do we start? This pile of data is no more accessible than the paper notebooks and spreadsheets we collected in the past. Which items do we need to look at to start improving? Did we just collect all this data to only make our problems bigger and even less easy to find? Why did we spend all that money on getting sensors and a data gateway installed? We aren’t closer to becoming more effective and sustainable at all – we even waste more energy now.

This could be the outcome of a naive (not-so-)smart-industry project in a company that’s just making its first steps into putting the digital world to good use. A horror scenario for a plant manager and his board – spending time and money on becoming more modern and then being left with only more problems to solve. It’s something that may happen when blindly following the advertising of IoT and smart data that comes by on the internet and in magazines. As always, technology by itself isn’t the answer, not even when going for 98 or 100 percent digital solutions. We always need to have a good look at what we want to learn and improve.


We don’t need all data for everything, and we don’t need to use all the data we have. To guide us, we need metrics or key performance indicators we want to improve on. I’ve never been a big fan of KPIs, as they’re often not as key as we want to believe, and as such become unnecessary overhead. Let’s adopt the term PPI to get rid of that nasty taste: production performance indicators. It fits more with the domain I’m talking about and it makes explicit what’s key: production performance.


Free webinar ‘Modernizing your code base with C++20’

As many production tool chains now adopt C++20 features, the potential this brings is unlocked. What advantages can recent versions offer to your code base? In this webinar we’ll look at the great improvements C++ has gone through and how features like concepts and ranges can transform your code. Register for video access.

Of course, production performance still has multiple dimensions, like finance, time and resource usage and waste production, so we still have to make choices. I prefer to look at three items when it comes to production performance in general, which can be fine-tuned for different industries: efficiency, effectivity and sustainability.

Efficiency is something most people will grasp immediately – it’s about the optimal use of available capacity and resources. Squeeze as much product out of a system as possible. This is an important PPI, but also a pitfall. I’m not going to go over all the details, but we already know since the 1980s that too much focus on efficiency leads to local optimizations and overall failure to meet goals. Using every machine to its maximum capacity introduces bottlenecks and starvation at the same time within a single production facility. So, efficiency is important but should be measured across the full production line. There’s a lot to gain from data provided by each machine, each production step, and by finding the optimal efficiency across the line.

Effectivity is closely related to efficiency, but not the same. It’s about using all parts of the production line in the most effective way, making it contribute in the best possible way to the overall system and, if you will, its efficiency. Also here, we have to look at the individual parts of a production system as well as at the whole. In an earlier article, I used an example where a company scrapped about a ton of steel every week due to a poor setup procedure for one of its machines. Because of this bad setup, that machine’s effectivity was 1,000 kg of steel (and its processing time) away from what it should be.

Sustainability ties directly into this. A sustainable factory minimizes waste production and resource consumption. For those in the world of lean manufacturing: indeed, both of these things classify as waste. Wasting materials isn’t sustainable – it ruins production, costs money and in a lot of cases also negatively affects the environment. Similarly, wasting 1,000 kg of steel per week also means that the energy used to process that material (and correct the machine setup) is wasted.


For each factory we try to help, we try to identify which things they’d like to improve in these areas. Sometimes it’s very specific, like “use less electricity” or “lose less time on production errors/fault/mistakes.” Once we’ve identified those things, which are more quantifiable than efficiency or sustainability themselves, we can start determining which data we need to gather to be able to analyze what part of the production is a candidate for change. Sometimes we need data from individual machines, sometimes a correlation between production planning and production output, or material purchasing and production execution.

When we have this clear, we can identify for each item what we want to analyze. Sometimes so-called time-series data from devices is sufficient, sometimes we may also want to match this with MES or ERP data (production plans, logistics flows) to find out what’s not obvious on the surface. Digital dashboards help visualize the data related to selected PPIs and combinations of data that can be used to create new insights.

Take the case of a dosing system that was leading to an undesired waste of material (effectivity and sustainability are involved here). Neither the data coming from the machine itself nor from the other systems gave an inkling of what was going on. Then someone realized that the problems always occurred on Friday morning between 10 and 12, and only when a certain type of ingredient was dosed, one that flows quite easily through dosing systems.

That helped a lot, but not enough – the dashboard and available data provided no clue as to what was causing the problem. Only when someone went on to a piece of data from a completely different source, the issue became clear. That piece of data was the work schedule of the truck cleaning facility next to the production location. It turned out that on Friday morning, a larger amount of trucks came in for cleaning than during the rest of the week, causing vibrations that affected the dosing system. Enough to cause overdoses, not enough to cause complete failure.

Needless to say that this piece of data was only added after someone went on-site and noticed the vibrations. This proves both the need for and the limitations of dashboards: they can bring correlations to light, but only to a limited extent. We’d very much like to see some form of artificial intelligence help us with this – but also here we have to deal with the fact that while correlations may be detected, the current state of the art in AI may not necessarily be able to identify the underlying cause either.

Main picture credit: Arie Wubben on Unsplash

Edited by Nieke Roos