What a product identifier actually identifies
A barcode on a pack identifies a specific saleable configuration: a brand, a variant, a size, a pack count. It does not identify a product in the sense a marketer means, and the gap between those two ideas is where a great deal of retail analysis goes wrong.
A brand manager asking "how is the product doing" is asking about something that exists across several identifiers, and the data answers at the identifier level unless somebody has done work.
What does the identifier actually cover?
A saleable unit as the manufacturer defined it. Change the pack size, the flavour, the count, or in many cases the packaging itself, and you have a different identifier. Two items that a shopper would describe as the same product frequently are not the same item in the data.
That is a feature rather than a defect. The identifier is precise; the imprecision is in the question.
Where does this bite in practice?
Trend analysis across a pack change. A brand that moved from one count to another will show the old identifier declining to zero and a new one starting from zero, and a report that treats them as separate items will describe a collapse and a launch that never happened.
The same applies to a reformulation, a rebrand, or a packaging refresh that triggered a new code. None of those are market events and all of them look like market events in a naive time series.
What is the standard behind the codes?
Globally, the identifier system is administered through GS1 and the underlying identifier is a GTIN. The allocation rules are published, and they are the reason a pack change frequently requires a new code rather than merely permitting one.
Knowing the rules tells an analyst when to expect a discontinuity rather than discovering one in a chart.
How do you analyse across identifiers?
By building a product hierarchy: a grouping that maps identifiers to the brand, variant and size concepts the business actually uses. That mapping is maintained work rather than a one-time exercise, because new identifiers appear continuously.
Somebody has to own it. A hierarchy that nobody maintains produces exactly the discontinuity it was built to prevent, and it produces it silently.
What about items with no barcode?
They exist in small-format retail — loose goods, in-store prepared items, and anything rung on a department key. Those transactions carry a value and not an item, which means they are invisible to item-level analysis and present in the total.
An analysis that reconciles to a store's total sales has to account for them. One that does not will show a gap and may conclude something about items when the explanation is a keying practice.
How does private label complicate it?
Private label items may carry retailer-assigned codes rather than manufacturer ones, and the same physical product can appear under different codes at different retailers. For a brand measuring its competitive set, that fragmentation understates a competitor that is actually one competitor.
It is one of the reasons private label share is harder to measure than branded share, and worth stating explicitly whenever it is reported.
What should a buyer ask a provider?
How identifiers are grouped, who maintains the hierarchy, how frequently it is updated, and what happens when a code changes. Those four questions expose most of the ways a time series can lie without anyone intending it.
Ask also what share of measured sales carries no item identifier. That figure is rarely volunteered and it bounds what item-level analysis can claim.
Frequently asked questions
Is every product change a new code?
Not every change, and the allocation rules define which changes require one. Size, count and formulation changes commonly do; a purely cosmetic label refresh may not.
What happens to the old code?
It stops selling as stock depletes, over a period that varies by channel and by how fast the trade cycles. That tail is why a switch looks gradual in data and abrupt in a plan.
Can two products share a code?
They should not, and occasionally do through error or through reuse after a long gap. Duplicate or reused codes are a data quality issue worth checking when a series behaves strangely.
How do I handle a competitor that rebranded?
The same way as your own rebrand: map the codes together in the hierarchy and note the date. Without the note, next year's analyst will read the discontinuity as a market movement.
Does this affect distribution measures?
Directly. Distribution is measured per item, so a brand that splits one item into three has changed its distribution figures without changing its shelf presence. That is one of the more common ways a distribution number misleads.
What about multipacks?
They are separate items with their own codes, and comparing a multipack to a single without normalising volume compares two different things. Equivalised units exist for exactly this.
Should a brand simplify its item count?
That is a commercial question rather than an analytical one, though a smaller, stable item set is easier to measure and easier to explain. Complexity has a real measurement cost that is rarely counted.
Who should own the hierarchy inside a company?
Whoever owns the reporting, with input from whoever owns the products. Splitting the two is how a hierarchy stops reflecting the business within a year.
How often should a hierarchy be audited?
Quarterly for new codes and annually end to end. New items appear continuously and each one either lands in the right group or quietly creates a gap between the data and the business view.
Working with retail data? See NRS Insights data solutions or contact NRS Insights.