Problems with the data

Every database has flaws: idiosyncrasies, missing data, etc. The worst problem is wrong data that looks believable. How are you going to handle these flaws? First, check your data carefully for missing or obviously wrong values. Then estimate how large each problem is. Missing data on one sku out of 10,000 may be safely ignored; missing data on half of those sku's means the project is dead.

Here are some things to note about the SPR data.

• Sales history
  • Time-of-day is not included and so we cannot resolve our analysis more finely than one day.
  • Sometimes a sku will appear in more than one line within a given order. This is the way the customer submitted the order. (They may have done this to separate out, for their own records, different amounts of the same sku that were intended for different customers of theirs.)
  • There are occasionally entire pick-lines that are duplicated. It is not clear why/how this happened.
• Item Master
  • Some items do not have dimensions.
• Warehouse layout
  • The CAD drawings do not agree in all details with the jpg images. The CAD drawings are in principal more accurate than the jpg images; but they harder to update and therefore less likely to be accurate.
  • Some sections of shelf that appear in the drawings do not appear as storage locations in the Item Master. This is not a flaw: It may be that these sections are used to store other things, such as overstock, shipping containers, totes, printer supplies, etc.

Copyright © John J. Bartholdi, III and Steven T. Hackman. All Rights Reserved.

This is material to supplement our textbook Warehousing & Distribution Science. See warehouse-science.com.

Last revised: 27 February 2003