Team,
Notes from our Visit with Corning SEMPAC in Corning, NY on 12/08/15.
Please review and edit:
Visit Notes: Corning SEMPAC – Corning, NY – 12/08/15
1. Place, Date & Time:
- Customer: Corning SEMPAC
- Place: Corning, NY
- Dates: Tuesday, December 8th, 2015
- Times: 8:00AM-12:00PM EST
- Meeting Type: JMP Demo Visit
2. Background
- JMP live demo visit requested by contact Dean Neubauer to show JMP and JMP Pro fit for Corning’s new SEMPAC improvement program.
- Outgrowth of several meetings with Dean to best fit JMP/JMP Pro tools into Corning’s SEMPAC flow (Mike and Jian)
- SEMPAC program background:
- Steps are Strategy/Structure, Explore, Model, Predictive Analytics, & Control
- Heavy on Predictive Analytics on historical data, but light on Experimental Design
- Meant to be complimentary to existing Corning Six Sigma Process
- Expansion/Roll-Out:
- Program is still in foundational year period
- Piloted process on a few key projects to date using SEMPAC steering committee members
- Would envision training once a quarter
- Can train up to 25 max per calls (but would realistically expect 10 to 12
- Tools current used:
- Analytics within SEMPAC: Considering JMP for analytics/modeling, and Matlab for some modeling/automation in Control phase.
- Analytics elsewhere within Corning: Excel, Minitab, some JMP
- For Database: Oracle, OSI PI, SQL
- For File Types: Excel, CSV
- Pre-Demo Comments from Steering Committee on JMP
- Dean Neubauer:
- Strengths of JMP would fit in the Explore, Modeling & Predictive Analytics steps
- However doesn’t see a lot of JMP Pro possible use (mostly due to price tag)
- Ashley Johnson-Long:
- JMP looks good for training, but concerned on how it would be structured (number of days, etc.)
- Training Goal: Make it easy to apply
- Already bought Matlab for her Modeling
- Note: Matlab coming in for a three day training around SEMPAC flow
- Casey Wood:
- New to JMP
- May use it for light Visualization
- Most interested in how JMP will work with Corning ETL Strategy & Structure
- Using Custom SQL Database (Codes in C Sharp) and OSI PI
- Seeniang Chen:
- Interested in JMP for Data Visualization and Analytics (especially Model Optimization)
- R user for Modeling
- Most concerned about R and Matlab integration if it would slow down the computing speed
- Karen Maxwell:
- Interested in JMP as a Engineering tool that really gets used
- Software goal is to enable them to get started right away
- Pre-Demo JMP Questions:
- How is Corning SEMPAC difference than their Six Sigma program?
- Claimed that they are working with Corporate Quality and SEMPAC and run parallel with the Six Sigma Program
- Pointed out that SEMPAC is in it’s foundation year
- Said that SEMPAC is more “Applied” than Six Sigma with more Structure and Predictive Analytics
- However noted that there wasn’t any Six Sigma representatives at the session, nor any claimed certified belts among the Steering Committee
- Where does DOE fit within SEMPAC?
- They do not call it out specifically in the process as more modeling focused from historical data
- However they felt the modeling will trigger DOEs in other groups (like the M&E Team)
- Corning
- Dean Neubauer – SEMPAC Creator & Champion
- Karen Maxwell – SEMPAC Sponsor
- Casey Wood – IT ETL – SEMPAC Steering Committee
- Ashley Johnson-Long – SEMPAC Steering Committee
- Seeniang Chen - SEMPAC Steering Committee
- JMP
- Dan Durkin – JMP Beacon Account Executive
- David Schultz – JMP New York Account Executive
- Mike Anderson – JMP New York System Engineer
- Jian Cao – JMP Global Technical Team
- Scott Wise – JMP Global Technical Team
4. Visit Purpose
- Position JMP and JMP Pro for inclusion in Corning SEMPAC training program
5. Customer Needs
6. What Shown
- Introductions (Dan Durkin)
- What is JMP
- JMP Pro Advantage
- JMP Fit for SEMPAC (Scott Wise)
- JMP Value for SEMPAC
- Showed Corning SEMPAC Example Executive Slides (for Karen Maxwell – Limited Time )
- Introduced JMP Corning Custom Starter Menu
- JMP SEMPAC Demo Case (Scott Wise)
- Background (White Polymer Case)
- Step 1: Structure
- Pre-Process: Extract, Transform & Load Data
- Database Entry: SQL Builder
- Q&A Want to see if we plan to have a native connector to OSI PI beyond just ODBC? - Answered that we would check on our position regarding OSI PI.
- File Entry: Excel Wizard
- Explore Missing Values
- Explore Outliers
- Q&A: Asked if more Outlier Exploration techniques were available in JMP? - Answered that some are being considered for future releases. (Action: Dean will send Mike and Team ones he is looking from from the ASTM standards)
- Data Resolution
- Load Data
- Distribution: Outputs & Inputs (with Dynamic Linkage)
- Q&A: Dean wants to move the Distribution Dynamic Linkage down to Explore – Just Distribution for Outlier Detection here
- Step 2: Explore
- Find Relevant Variables
- Multivariate Chart
- Q&A: Dean liked the combination of the Correlation Table with the Scatterplot Matrix
- Q&A: Dan wanted SVM methodologies – Answer: Available over an R Add-In (Action: Send link to Dean)
- “Lite” Anlaytics
- More Intensive Data Visualization
- Targeted Statistical Analysis
- Process Capability (Goal Plot)
- Step 3: Model
- Find Important Variables & Create Models
- Partition – Decision Tree in JMP
- Partition – Decision Tree with IM in JMP
- Fit Model – Stepwise Regression in JMP
- Q&A: Showed analysis for Categorical Outputs (Life Curves, etc.)
- Q&A: Team big on VIP (Variable Importance) measures
- Advanced Statistical Anlaysis
- Continue to Visualize Data
- Step 4: Predictive Analytics
- Machine Learning Procedures:
- Cross Validation Set-Up
- Q&A: Want Cross Validation in JMP? - Answered on complexity and capability provided by the Training Set.
- Q&A: Wanted this moved up to the Modeling phase.
- Partition – Decision Tree in JMP Pro with CV
- Q&A: Dean wanted RMSE instead of R2 in the Decision Split History Graph with CV
- Partition – Bootstrap Forest in JMP Pro with CV
- Other Advanced Models:
- Fit Model – Generalized Regression – Stepwise and Lasso in JMP Pro with CV
- Verify Findings:
- Optimize the Model:
- Model Profiler
- Model Profiler with Optimization
- Model Profiler with Simulation
- Step 5: Control Phase
- Monitor & Reassess:
- Graph Builder – After Results
- Control Charts – After Results
- Reporting & Sharing
- Wrap-Up
7. Customer Feedback
- Lots of Discussion around possible JMP Training
- Two possible trainings
- JMP to Minitab Conversion (JMP Only)
- JMP SEMPAC (JMP or JMP Pro if can work out cost)
- JMP Training Structure (Courses/Workshops/JMP Days)
- JMP 1 – Structure & Explore
- JMP 2 – Model & Predictive Analytics
- JMP 3 – DOE Conversion
- JMP 4 – DOE
- JMP 5 – Reliability (Bill & Leo)
- JMP Pro as being too expensive
- Team encouraged a broader view of possible users from JMP Pro if they agree it has the vital tools they need (particularly cross validation and model comparison)
- Possibility if we can get volume pricing
- JMP Pro only used by select Steering Committee Members now
- Planning on only training 40-50 next year (very much project pulled interest in SEMPAC right now)
- Dean mentioned possibly starting out just using JMP Pro as a teaser in the training.
- Want to keep JMP from being perceived at Corning as only a DOE tool
- Opportunity to get strong JMP and possibly JMP Pro in Corning SEMPAC training
- Offer to get JMP more properly tool managed through Casey’s IT group from all of Corning
9. Next Steps
- Have Corning meet with the JMP Training group to discuss possible ways to provide JMP Training options for SEMPAC
- Meet with Casey’s Manager in IT to get tool management of JMP throughout Corning
- Meeting will need to take place the tail end of January
- Technical Team to follow-up on requested actions:
- Plans for JMP native connections to things like OSI PI instead of ODBC?
- Send Dean the JMP R integration SVM add-in
- Send any info on ways to do cross validation strategies in base JMP
- Get info (ASTM Standard Tools) from Dean to help with any planned additions to Outlier Detection methods in future JMP releases
- Pass on request to look at RMSE instead of R2 in the Decision Split History Graph with CV
Attachments:
- Visit Notes- Corning SEMPAC – Corning, NY.eml (Detailed Notes)
- RE- JMP Training at Corning.eml (Training Opportunity Notice to SAS Training)
- RE - JMP Visit December 8.eml (Dan's Visit Follow-Up with Contact)
