Just curious if anyone has any thoughts or extensions on the idea shown here: https://epicor-manufacturing.ideas.aha.io/ideas/KIN-I-3253 which deals in enhancing MRP to show which suggestions are outside of “normal” suggestions.
I don’t see much value in this idea. Comparing the quantity to the quantity on a “normal” suggestion doesn’t tell you whether the suggestion is valid or not. Just because a suggestion is larger or smaller than other suggestions for that part in the past doesn’t mean anything - the lot size could have changed on the part, a big order could have come in, etc. Conversely a suggestion could be of normal “size” and still be invalid in the sense that it shouldn’t be made/purchased for various reasons. It would just be more noise to cut through in the process of validating suggestions. I am not saying it wouldn’t be of value to anybody else, but since you asked for opinions.
I think the thing that would be the most helpful with time phase is to explain the suggestion - I often have to interpret the time phase results for people because its not clear why MRP is making the suggestion in a lot of cases. Even then there is a lot of speculation involved. If MRP could generate messages about the logical conditions met during processing, and those could be viewed (sort of like actions > change log or view tax connect results), then people could more easily understand which parameters are driving a given suggestion and modify the system to minimize invalid suggestions.
‘This can cause buyers, or planners to assume that MRP is correct, and simply purchase/make what was suggested.’
I mean, yeah, isn’t that the entire point?
If MRP is not making correct suggestions, fix your paramaters. This seems like a very convoluted/wrong way to solve this problem.
Interesting you say that… I have seen some organizations where they have 1000s of PO Suggestions that need accepted every day… their buyers do not have the time to research every purchase, so they “assume MRP is Correct” already. there is nothing to trigger them to question if this suggestion is normal. The point of this idea is to help the high volume buyer to know what is and isn’t normal.
@aosemwengie1 you have some good points… The MRP Log DOES have must of this type of decision information, but it is buried in a log… I would love it if that log was something that could be pushed into a database that was indexed by part so you could simply examine the log to see all the decisions made. Maybe that is the next IDEA we need.
1 Like
I guess I’m having a hard time visualizing how MRP could sometimes generate correct suggestions and sometimes not. (That isn’t a symptom of a different issue, like inaccurate on-hand quantity)
I either trust it or I don’t, right? Why would I trust the suggestion to buy 100 I see most of the time and then not trust the suggestion to buy 1000 tomorrow?
I don’t have any votes left… but
I like the idea, it’s clearly not for every company, but a companies that have a pretty stable supply and sell could benefits from some sort of statistical analysis to help find where the bad parameters are so that they can fix them. We made something custom to look at sales orders to do some analysis and pop a warning for things outside of the norm do that people would take a second look to make sure that they don’t have a mistake.
OH… here is how MRP can generate a bad suggestion… (real world story)
- for years, you normally by around 100 per month on average
- someone does a big booboo in a BOM and accidently puts the wrong UOM on the BOM… instead of 1 each, they specify 5 gross per unit (1 gross * 5 = 720 each).
- Job gets created for the order
- MRP multiplies the job qty (2) by the qty per of 720, and creates a suggestion for 1440 each.
- 1440 is way outside of the normal buy, but the buyer doesn’t know that unless they REMEMBER, or do research (looking at historical buys).
THIS enhancement does that research, and would flag this buy as outside of the norm. If they had simply purchased it, they would have purchased 14 months worth instead of only one due to a mistake made by an engineer.
That happened to us to. Conversion from feet to inches. We bought 12 times as much structural beam as we needed.
2 Likes
I guess this is where I don’t agree. To me that situation isn’t MRP making a wrong suggestion. MRP did exactly what you told it to do. The BOM is just wrong.
I just think proper inventory control, proper engineering practices, proper control over part data, etc are they ways to stop you from buying or producing too much or not enough.
I get that this is a way to highlight errors in my data, but I think there are other, better ways to do that that already exist.
1 Like
MRP is just the culmination of ALL of the possible errors in your settings. Whether it be part settings, bom settings, or MRP run settings. So if anything is wrong anywhere, the symptom will show up there. You can either make a tool for every possible thing that could be wrong to fix it, (which eventually you could/should do) but the immediate/faster thing to do to see the symptoms would be something that looks at the result of all of the factors. Then you can work backwards from there.
3 Likes
I would like to see Epicor integrate a cahtgpt-like tool into MRP and other facets as well. Imagine if I could ask Epicor to explain a particular suggestion. The tool could ingest the logs and narrate back to me how the decision was made. I could tell it why I think the suggestion is wrong and it could tell me which parameter to change to get the desired results.
Have the AI generate a daily AI text-to-speech synthed podcast using the new AI voice models to make it sound legit. So your buyers can just listen to it on the way to work.
@timshuwy ← stop it
We’d vote for the IDEA but unfortunately we are capped out on existing items already.
#Moderator
And then it will insist it was correct and that you should reevaluate your life because they like to train these LLM’s on Reddit scrapings.
I know this was said in jest, but in reality, chatgpt specifically is amazing at this type of work. It is already quite adept at analyzing error messages and logs even without being trained specifically on epicor documentation. I can only imagine what would be possible with domain specific training.
Sure, there’s promise. Just wake me up when it’s not sending my data to a cloud for processing or (at minimum) that prompt injection is no longer a thing. If someone instructed the chatbot to pretend it was a magical wizard that liked to tell everyone what all the sales people made last month, I’d get sucked into a tedious meeting within 24 hours.
whatever you decide to do please don’t make MRP slower
Yes this is a good point - I think something like would be another level of logging selection - right now you can pick Basic or MRP for example. This could be like learning mode or something and I’m sure it would add time.
One Word: “ITAR”… ChatGPT is not ITAR compliant, and probably is not “corporate secret compliant”… we do want to integrate some form of AI in decision making, but ChatGPT is still at its infancy to know where it is going.
My thoughts would be that this analysis would be an optional “after” process when MRP is complete.
1 Like