From: Scott Cleland (gt8987b@prism.gatech.edu)
Date: 06 Oct 1999
Time: 13:46:50
Subject: Order Fulfillment Rejects
The Situation: An order fulfillment center (for a drug company) repacks individual items (bottles of pills) for delivery to customers. Each order can consist of many different types of pill bottles. After each order is picked, the box will be weighed and compared to the expected weight of the order. A diverter then sends any boxes not within tolerance to a blind verification station. The acceptable tolerances are listed as: 0-5 lbs +-a lbs 5-10 lbs +-b lbs 10-15 lbs +-c lbs 20+ lbs +-d lbs
The Question: What should a,b,c, and d be set to in order to ensure less than 99% of orders shipped are correct?
The Data: We know the weight of every (600+ items) individual item.
The Current Approach: We are sampling boxes. We weigh the box, itemize its contents, and verify the order is correct. If it is incorrect, we correct it and re-weigh the box. After enough sampling, we will have a spreadsheet like so: Order Weight ExpectedWeight Correct/Incorrect ------------------------------------------ 1 6.05 6.15 Correct 2 1.50 1.90 Incorrect 3 8.08 8.09 Correct Once we have all of this data, we can just arbitrarily choose a,b,c, & d until we reach a level where less than 1% of orders that pass are incorrect.
The Desired Approach We’d like to use some statistical method (like I learned in Stats I class but have since forgotten) to set the tolerances.
Any input will be appreciated. More information will gladly be provided if neccessary. Thank you,
Scott Cleland