Retail supply chain analysis
jvr001
New Altair Community Member
We run a retail network across 3 distribution centers, 10 regional cross-dock 'depots' and 1,110 stores. We track goods at a SKU level using bar-code scanning - from receiving (from suppliers) at distribution center to eventual delivery at store.
Challenge: often when stores receive their goods they indicate that the quantity (or actual item, or size) differs from their original order. In other words, either the order was picked wrong at the DC, or pilferage anywhere in the network, or an administrative mistake was made at time of order capture at the store.
We can track at SKU level: when it was scanned in/out at numerous paths/nodes in the distribution network, also who picked and handled the item last.
My idea is to use data mining to look for patterns/correlations within the data set matching all 'wrong orders'. Are the errors correlated to certain pickers, checkers, security personnel, channels, nodes, stores, time of day (e.g. start or end of shifts), store personnel, etc.? Would like to have a prediction of the likelihood of an error happening when certain variables correlate.
My question: do you think data mining is a valuable method for this scenario? Have any of the members dealt with a similar set of challenges/requirements?
Your assistance much appreciated.
Challenge: often when stores receive their goods they indicate that the quantity (or actual item, or size) differs from their original order. In other words, either the order was picked wrong at the DC, or pilferage anywhere in the network, or an administrative mistake was made at time of order capture at the store.
We can track at SKU level: when it was scanned in/out at numerous paths/nodes in the distribution network, also who picked and handled the item last.
My idea is to use data mining to look for patterns/correlations within the data set matching all 'wrong orders'. Are the errors correlated to certain pickers, checkers, security personnel, channels, nodes, stores, time of day (e.g. start or end of shifts), store personnel, etc.? Would like to have a prediction of the likelihood of an error happening when certain variables correlate.
My question: do you think data mining is a valuable method for this scenario? Have any of the members dealt with a similar set of challenges/requirements?
Your assistance much appreciated.
Tagged:
0
Answers
-
Hi,
well, without knowing the available data and its quality it's of course hard to say but I would at least say: it's definately worth a try. Assuming that there actually _is_ any correlation and it was captured by the describing features, data mining will certainly be able to get it and describe it to you. Although I cannot remember that we already have worked on exactly a scenario like this, it reminds me of several successful projects which were quite similar and where also human work was involved a lot, e.g. in production scenarios. But be cautious: taking factors like "Who was the picker / checker / ... at a specific time" into account might be difficult depending on you legal (and ethical) standards.
Cheers,
Ingo0
