Business Intelligence in Manufacturing

Our experience of working with manufacturing is that there is a lot of data, it is spread between inbuilt reports, business intelligence systems and a lot of spreadsheets.

In general, manufacturing is already an efficient part of the economy, not least because it faces global competition in one way or another. We rely on a percentage point or two of improvement every year just to stay in the game. The operational people we speak to within our clients’ businesses are switched on to efficiency of delivery and are often focus on the efficiency of product changeovers.

The world is changing in terms of environment and technology, but many things stay the same. Here, I look at the constants and the changes of Business Intelligence within manufacturing.

What stays the same

Some, perhaps most, of the key performance indicators (KPI’s) remain constant. 30 years ago, working on ratios of downtime to production time, doesn’t feel that different to working on the systems today. We certainly have better tools to hook out the data and report on it and the graphs are prettier, but the math’s is similar.

There is a debated phrase “What gets measured gets managed” erroneously often attributed to Peter Drucker.

One reason it is debated is because sometimes, what gets measured gets gamed.

We have been involved in trying to mitigate this gaming, although it is largely a management issue. Our input is to try and ensure that the data warehouse tries as far as possible to represent the real business meaning of data, so it can be shuffled into reports. But most teams can be e a little bit guilty of this attempt to put things in the best possible light. The more we can stick to the ‘facts’ the better.

How to ‘police’ this well much is in the world of good management, letting the figures throw light on the subject without seeing them as the whole truth and certainly not as a stick to hit people with, probably helps. But it also means making that there is a pressure on us to prove the figures.

We tend to think that a good manufacturing enterprise system needs to run for at least a decade. Even then if you have five main systems it means you are having to replace one every two years. Business Intelligence (BI) systems, even if they stay the same, must constantly move to hook onto new data. At the same time the nuance of meaning from one system to another can change as manager manage.

What Changes in Business

Our BI practice is influenced by the ‘great thinkers’ of BI, in particular the seminal work “The Data Warehouse Toolkit” by the Kimball group, understandably not a household name outside the BI community. The office copy of the book has a post-it note stuck on the front saying “thanks be to Kimball”. The book suggests a ‘Publishing Metaphor’ which suggests we need to understand the business, deliver high quality data, sustain the BI environment to maintain trust and to present the data of interest in a consumable way.

What is of interest changes. For example, a few years back we worked on a shop floor data report that presented production in a way that drew attention to the speed of production. The factory was operating at capacity 24/7 and keeping an eye on speed was key to keeping things moving. The balance between moving faster and risking downtime, to slowing down and having an easier life was front and centre. Years later these are different time, smaller runs, no longer at capacity, the focus has changed, alongside this you would expect the reports produced are changed, even over the same data.

In order to get the year-on-year efficiency gain, manufacturers need it is clear that operations will change even if the systems stay the same. There is bound to be some rework, squeezing more value out of the original investment in BI and reporting.

What Changes in Technology

Technology moves onwards, ‘new technologies’ as such are rare, but there are certainly ‘new versions’ of existing technologies. Where once most of the reporting we did used Microsoft Analysis Service alongside SQL Server, supported by SSRS (SQL Server Reports Service) or some equivalent, now more of the work is in Microsoft Power BI. Usually, the data is still in SQL Server (of MySQL or Oracle) but sometimes it is now in the cloud in a Power BI Dataverse, reducing the need of IT capabilities of the client, but admittedly adding to the bills paid to cloud providers as they do more of the heavy lifting.

The hardest part of data, in our experience, is marshalling it. Getting it out of the source system, working out how to equate an order in the logistics system to an order in the sales system for example (we call this ETL (Extract Transform Load)). Increasingly, this process does not hook directly into databases but onto published API’s of various newer systems. Modern tooling is setup for this.

Once you have the data in a data warehouse there are lots of ways to report it. This is probably where the tools have gained most ground. Solutions that have previously only possible for large industrials are now available and accessible to SME’s. Using Power BI, Tableau or Qlik (Power BI being our weapon of choice) to provide the front end functionality.

Machine Learning

The newest area of technology is the world of Machine Learning or more excitedly AI. This is really a form of statistical analysis that has been available to the biggest businesses for decades. However, this is now affordable to smaller businesses and increasingly with some off the shelf functionality built into modern tooling. Power BI will now try to bring to your attention trends that might be of interest.  Being able to exploit these opportunities is still dependent on having a solid base of data to work with.

What Should you do Next

Members of our data practice are of course excited by these new things. But we recommend that you first look at what value there is in your existing setup. Have neglected reports broken? Have updated systems not led to updated data imports? If you’ve invested five or six figure sums to this point, then some clever people have probably done some clever stuff. We recommend you look at what you already have. This unromantic piece of maintenance is sometime the best value.

After that, we suggest you think about what ‘ideal’ looks like in terms of your management reporting requirements and identify where the gaps are.

If you have a stack of data already cleaned and ready to roll it is worth looking at whether you can exploit this further by deploying a modern tool such as Power BI against it to put dashboards and reports together or talk about whether there are ways Machine Learning can give new insights.