I’m writing my master’s thesis under the title of “An Integrated Framework between Enterprise Resource Planning (ERP) and Machine Learning (ML) to Enhance Decision Making” And I will be extracting the data from Epicor ERP.
The main idea behind my Topic is – I have followed a research paper using Machine Learning in demand forecasting, utilizing historical data (Sales & Return (RMA) of a Finished Manufactured good) to predict future sales and make informed inventory decisions.
My new Approach is to make a detailed sophisticated analysis on the BOM not only the FG. Scenario: we are in a Manufacturing Company, we are creating a Job Entry for a Part (Manufactured), Part no = FG. I have added a BOM of 5 Parts to manufacture this FG. Part no as follows: P1, P2, P3, P4, P5. At the End of 2023, the CEO observed a decline in sales ratios and aims to conduct a more comprehensive analysis for the upcoming year, 2024.
Old Analysis: We will extract all Sales & RMA of all the Finished Goods we have manufactured through 2023, we will feed the Machine Learning with this amount of data, and upon Analysis ML will give us a better decision of what to manufacture again and what not to…
New Analysis( My Topic): I will not only look into the FG , But instead I will turn my focus on the BOM , I will try to gather as many records as possible from Epicor that will help me identify if any of the Materials I’m using is the reason behind the issue (Lower Sales) - because Ideally I may be using a Part in the BOM which is having a problem or for example for I’m purchasing this part from a bad supplier with a bad quality.
What I have reached so far – The field I can take on all the parts included in the BOM would be:
Inspection Records
DMR
RMA
Non- Conformance
Supplier Rating
these records will give me a better insight about the Quality of all Parts I’m using in the BOM.
My Query is – are there any more records from Epicor I can extract other than the mentioned above 5 that would help me?
Thanks!!
That sounds very interesting and I can see where you are going. We have a similar need in that most of our Quality claims are because of some unseen or unobserved fault in the raw materials we use. Sometimes it’s because of the assembly process = read: skill level of the assembler.
You have the correct tablesets for quality data, so I might suggest the following extra data points:
Laborers in the process / time on the job - this would be labordtl and empbasic tables
Any Part substitutions made for raw materials (you didn’t mention if this is possible or not in your scenario)
“Seasonality” on the sales cycle - eg. our FG are durable and vary in their replacement cycles. Some years the same FG may sell more than other years, but sometimes the customer will break the item and require it sooner that on a normal cycle. it’s possible to skew the analysis in the wrong direction for something that is outside the control of the company.
RMAs - you have this, but make sure it’s both for FG returned to the company and parts returned to vendor for material defects
OrderHed/JobHead - delayed start/finish that causes a job stoppage or lost order.
Thanks so much Mike!!
This is very helpful for me!
In your List - Point 1,3,4,5 are in my scenario, I need to do more research regarding Point no 2.
Thanks for opening up my mind with the “Seasonality” because defiantly it will also affect the decision.
Very Welcome!
I know that a number of us nerds here would surely like to read your work once it’s concluded. Most of us have been asked by executive management at one time or another if AI/ML would apply in some manner -but most of us have no spare time to investigate.
I can understand the unavailability of Time
Thanks, Mike for the very beautiful compliment, I would surely be happy and welcoming if someone reads my work! and very much Thanks for the help!
As Mike said, LaborDtl and EmpBasic would be an important component on this analysis. We’ve had a few issues with Order Entry as well (users misunderstanding the order entry process leading to errors), so that may be a piece to consider as well. Purchase Orders may be one more piece to consider. If receiving P1 and P2 is taking much longer than expected, FG production will be delayed.
