We had several conversation with UPM's presentatives on defining the scope of the problem. We started with defining the value stream of the production line from the logging of order to the delivery. Our focus started to sharpen on the claims and to the link between it and customer behavior during the planning phase of the production.
What it does
KUPI is a claim funnel which aims to provide model for identifying orders that are in danger to generate customer claim. It will provide set of parameters that are found to have in certain combinations greater potential to escalate to a claim. KUPI also provides different stages where unwanted performance indicators are tracked and tries to offer possibility to react fast on certain customer behavior or let the order go on till the next set of indicator parameters.
How we built it
Hypothesis was formed based on the feedback from UPM representatives and after that used IBM Watson to explore the data set to verify our assumption. After that necessary sql queries were created for the data set to uncover the story from data.
Challenges we ran into
We had no previous experience with IBM Bluemix or Watson. We also tried to use R to analyze data but soon found out that we were not up for the task.
Accomplishments that we're proud of
We are happy to be alive
What we learned
We learned that we are too old to stay awake for the weekend. We also found new gratitude towards our comfortable beds back at home. On top of that we learned a lot about IBM Bluemix and Watson. Also a lot of about how paper is made and how the value chain of UPM works.
What's next for KUPI - Key Unwanted Performance Indicators
KUPI might have some useful ideas that will be taken into consideration by UPM. To be honest everyone from UPM were very nice to us but their knowledge of their own business is so wide and deep that our contribution does not stand a change against that.