Considering industry 4.0? Need some best practices how to lay the analytics foundation?
While preparing for the Hannover Messe 2015 with Industrie 4.0 as a major theme, I stumbled on TDWI’s best practice report on next generation analytics. TDWI specializes on education and research around data & analytics. Their report is based on surveys and telephone interviews with technical users, business sponsors and experts. Most importantly the report collects the wisdom and experiences from many companies on their way to next generation analytics.
Why bother?
The eyes and ears of Industry 4.0 and its cyber-physical systems collects streaming, geospatial and unstructured data from sensors, makes sense of social networks and human interactions. The corner stone for industry 4.0 is the ability to capture and analyze what the sensors acquire, and contextualize it with business data – to drive better decisions. In many cases this is achieved through “next generation analytics”.
What is “next generation” anyway?
TDWI describes next generation analytics to go beyond BI reporting and dashboards – utilizing predictive analysis, text, social or geospatial analytics.
Where to start?
“Visualization for data discovery and/or predictive analytics is often a first stepping stone to more advanced analytics.”
When our data science teams from SAP work with customers, a key step is to understand and explore the data first. Data discovery helps to understand which parameters are most important, define the target variables, and develop a hypothesis.
Several mill products customers I talked to were quite happy with SAP Lumira for data discovery. Feedback was: it is straightforward to use, brings great visualization possiblities, and thus enables normal business experts to explore their data through ad hoc analytics.
Even before you get your hands dirty in predictive analysis, SAP Lumira is a great & simple starting point. Run it on SAP HANA, run it against operational analytics – OLAP and OLTP on one simple platform – you will get even more insights & decisions, and your SAP HANA platform gives you all opportunity to dive into advanced analytics.
Why did other customers use advanced analytics?
The TDWI survey indicates 2 main drivers:
- Support (strategic) decision making
- Understand customers better.
TDWI’s survey was cross industry. Based on the conversations I had, I would expect a slightly different picture in our industries. For building products industries I would agree with “understanding customers”.
For classical heavy manufacturing in mill products, like steel, paper or cement, I have seen a strong focus on operations – manufacturing, quality monitoring, asset management, and energy. In this domain, a large part of the big data is time-series-based, streaming data from sensors, manufacturing execution systems, SCADA and the likes.
In early POCs customers often look multiple years of data and try to develop a model to better understand quality deviations, or unplanned down-times.
As experience grows, queries and advanced analytics get much more (near) real time, including complex event stream processing, real time image processing e.g. for in-process quality monitoring.
Insight to action
Most respondents to the TDWI survey still use analytics to help manual decisions.
Some already have moved into alerting systems where an alert is fired when e.g. a combination of sensor values moves outside the allowed range.
And a few have progressed into proposing or automatically triggering actions.
Automatically triggering actions of course is nothing new in shop floor automation. But what is new, is to make use of advanced analytics and predictions as the trigger – pushing the automation to a higher and smarter level.
Top challenges and how to overcome
The survey respondents’ top challenge was “lack of skills in place to deal with new data, new technology, new analytics”. Over 52% percent were struggling with this – but they also gave clear and simple recommendations.
a) Train your people. Maybe also consider to hire external skill. Bundle your expertise in a COE.
b) Start with a proof of concept.
From your experience I would add:
c) Do not run this as a pure IT-driven project. Let the operations & business experts define their requirements for predictive analytics, and embed them into the project. They have the domain expertise, and they will need to trust the recommendation and prediction to take better decisions.
d) Also look outside the shop floor:
- How do your customers talk about your products and services on social media?
- How to segment your customers?
- Where do you suspect to loose money on fraudulent activities?
In which area do you see the biggest value for advanced analytics? Where to you see the biggest challenges? How did you overcome?