Quick data puzzle: two product names, one item or two?

Consider these two fictional rows in a product file:

• Harbor Lime Sparkling Water — 12 × 12 fl oz
• Harbor Sparkling Water, Lime — 144 fl oz

One item or two?

The arithmetic invites a quick answer. Twelve cans at 12 fluid ounces each equals 144 fluid ounces. The brand, product type, and flavor also appear to match. It would be easy to call the second row a differently written version of the first and remove a duplicate.

That would be premature.

The two rows might describe the same 12-can carton, with one source recording count and per-unit content while another records total content. They might also describe a 12-can carton and one large container. Or one row may contain incomplete or incorrect package information.

A name helps a person recognize a product. It does not, by itself, fully define the business object being counted.

Start with what “one item” means

The right answer changes with the question.

For an assortment analysis, both rows might belong to one product family: Harbor lime sparkling water. If the task is to count flavors or assess brand coverage, rolling verified package variations into that family may be appropriate.

For shelf pricing, inventory, procurement, or fulfillment, the relevant object is usually a consumer package or an orderable trade item. A 12-can carton and a single 144-fluid-ounce container should remain separate, even though their total stated volume matches. They may have different prices, barcodes, shelf locations, shipping needs, and replenishment patterns.

GS1 defines a trade item as a product or service that is priced, ordered, or invoiced somewhere in the supply chain. That definition is a useful reminder: a commercial item is not necessarily the same thing as a flavor, a product name, or a total quantity of liquid.

Before matching records, write down the level you intend to count:

• Product family: lime sparkling water
• Consumer package: a 12-can carton
• Trade item: the package configuration a buyer can order or be invoiced for

This small decision prevents a common mistake: treating a valid family-level match as proof that two package-level records are duplicates.

Compare the fields that carry meaning

Normalize product names so that they can be compared. Lowercase them, standardize obvious punctuation, and recognize that word order can vary. But do not let the normalized name make the decision for you.

Instead, inspect five fields or field groups.

1. Brand or owner. Is “Harbor” the same brand in both records, rather than a similar-looking name or a retailer label?

2. Product type and variant. Both rows say sparkling water and lime. That is stronger evidence than a broad category match, but formulation still matters. “Citrus” and “citrus + aloe,” for example, are not automatically the same hand soap.

3. Count and net content. Check the math, units, and whether the content refers to each unit or the whole package. Equivalent volume is useful evidence. It is not proof of equivalent packaging.

4. Packaging or selling level. Is the record for one bottle, a carton of cans, a case, or a retailer-created bundle? The same underlying beverage can appear at several levels.

5. A stable identifier. A valid GTIN, supplier SKU, manufacturer part number, or category-specific identifier can be stronger evidence than wording. Confirm that it belongs to the unit being counted. An identifier for a case should not be used to identify a single can.

GS1 standards call for distinct identifiers when separately traded package quantities differ. That makes sense operationally: a unit, a six-pack, and a 12-pack can contain the same product but be different orderable items.

A practical disposition, not a forced verdict

After the comparison, give each pair a clear disposition. Four outcomes cover most cases:

• Same item: Merge only when the evidence supports the same counting level and package.
• Same product family, different item: Keep separate records, while linking them to a shared family.
• Different item: Preserve the distinction because the variant, formulation, package, or brand differs.
• Unresolved: Seek another source, request clarification, or flag a likely data error.

For the Harbor puzzle, “same product family, unresolved package-level identity” is often the honest first answer. The equal volume is a clue, not a verdict. A package image, supplier catalog, or identifier may settle it. Until then, merging could erase a meaningful distinction.

This approach also handles less ambiguous examples. “Northline Cold Brew Coffee, Original, 4 × 11 fl oz” and “Northline Cold Brew Coffee, Original, 11 fl oz” are likely one family and two distinct packages. “Evergreen Hand Soap, Citrus, 12 fl oz” and “Evergreen Hand Soap, Citrus + Aloe, 12 fl oz” are likely different variants despite matching brand, format, and size.

Keep the evidence with the decision

Do not overwrite the original product names after making a match. Keep the raw name, a normalized comparison name, the disposition, the evidence used, and the date checked. If a supplier corrects a listing later, those fields explain why the earlier decision was reasonable and make revision easier.

Identifiers deserve care as well. They are powerful when valid and tied to the right packaging level, but they are not magic. A record may lack one. A retailer bundle may have its own identifier. A source may attach a correct identifier to the wrong row.

The puzzle is useful because it resists a universal rule. “Always deduplicate” loses package distinctions. “Always keep both” inflates counts when two names describe the same record.

Define the item first. Then use names as clues, attributes as evidence, and identifiers as confirmation where available.