A row whose family name disagrees
On this page11
One container on this shelf has a family name that contradicts its other fields
The product is named 500 cc Clear PET Packer Bottle. Its material field reads PET. Its family name reads HDPE Packer Bottle.
Two fields agree with each other and the third does not
The product name and the material field both say PET. The family label says a different polymer, and it is the only field that does.
And a separate family called PET Packer Bottle exists on the same shelf
Also holding exactly one row. So the shelf carries two packer bottle families, one named for each polymer, and the HDPE-named one contains a PET bottle.
Which settles what the family field is for
It is a grouping label applied to a product line, and a product line can be named after the polymer it mostly uses. It is not a per-row material statement and should not be read as one.
The material field is the one to rely on
Across this shelf it takes exactly three values: 117 PET, 78 polypropylene and 56 glass. No row records HDPE in the material field, notwithstanding the family name on that one row.
Which also corrects a possible reading of the three-value material field
Three values on this shelf is a routing outcome and not the shape of the catalog. Across all 2,150 products the material field takes twenty seven values.
Including the polymers this shelf does not have
HDPE on 251 products, LDPE on 65, PVC on 52, and beyond the plastics, Metal on 91 and Steel on 28. So a specification calling for HDPE has somewhere to go in the catalog, just not on this shelf.
That distinction is the one this library keeps returning to
A curated shelf cannot show what it excluded. Three materials here is a statement about this shelf, and reading it as a statement about supply is the recurring error.
One contradictory row out of 251 is a small defect and a useful one
Small because it affects one row. Useful because it demonstrates what the field is, which saves anybody from building a material filter on top of the family name.
And it was found by counting families rather than by looking for errors
Which is the usual way these turn up. Enumerating a field to the end surfaces the rows nobody has looked at, and the eleven single-row families are exactly those rows.