Analytics in Manufacturing – Takeaways from the recent webinar
What is Analytics?
In manufacturing/engineering applications data driven models (developed from company’s data, usually need to be built) complement first principle models (scientific principles and laws, often can find “off the shelf”).
IT Perspective to Analytics
Roles in Analytics
Tips for BAs
There are a different set of requirements for the analytic model development and the business application. Analytic models need to be:
Tools for this is the analytical workbench. Most of the time the analytical workbench is not related to the application workbench.
- Data driven models are not necessarily generic - analytics solutions for the same “problem” will always have to be localized and you need to plan accordingly.
- Source data – include requirements that assess the usability of the source data upfront
- Is there enough data to build a model?
- Is the quality good enough?
- Can we use it?
- Analytical Model Development - include requirements that support the model types and that automate development as much as possible.
- Availability of math libraries and visualization aids.
- Automation of standard activities (noise filtering, outlier removal, pattern recognition, etc.)
- Different model types require different techniques.
- Interim Data – include requirements that enable storage of interim and master reference datasets.
- Model development generates more data! Basic data, cleaned data, master data.
- Are there scientific, company or legal requirements to keep interim data?
- Analytic Model Validation – include requirements that validate the model from a scientific point of view.
- Do we have scientific test cases and test data?
- Can we reproduce historical behavior and events?
- Is the output in a useable format for end users?
- Deployment – extremely important one - include requirements that ensure business readiness.
- Is the infrastructure up to it?
- Is the business user ready for it?
- Are the support groups ready for it?
- Runtime – include requirements that work up the runtime input data.
- Do we need filters to avoid responding to outliers?
- Do we need time synchronization?
- What about latency and speed of calculation and data I/O?
- Health System – include requirements that monitor the model health during operations.
- Model updates – include requirements that allow continuous improvement for the models.
Marcel shared that as analytics becomes a wide-spread IT and business competency it systematically creates value in high potential areas across the value chain. “Applications and services using analytical principles as a backbone are delivering hundreds of millions of dollars per year.”
Thank you, Marcel, for this engaging and informative webinar. Thank you Kunal Joshi of the IIBA Business Analytics SIG committee and to Terri Lynn Rodrigues and the team at IIBA for making this webinar possible.
Submitted by Meri Gruber, IIBA Business Analytics SIG Co-chair