New buyers and existing buyers
A campaign that raises sales has done it by recruiting buyers, by increasing purchase frequency among existing ones, or by increasing quantity per purchase. Those are three different commercial outcomes and they imply different next steps.
In a channel without shopper identifiers, separating them is indirect and partially possible.
Why does the distinction matter?
Because recruitment and frequency have different long-run values. A campaign that recruits buyers has built something that persists after the flight; one that increased frequency among existing buyers may see the gain fade when the media stops.
A brand deciding whether to continue a programme needs to know which happened.
What can be observed without identifiers?
Store-level penetration proxies: the number of baskets containing the item, and the units per containing basket. Those two decompose the sales change into incidence and quantity.
An increase driven by more baskets containing the item is consistent with recruitment or with higher frequency; one driven by units per basket is quantity.
What does the decomposition not settle?
Whether more baskets means more people or the same people more often. That distinction genuinely requires identification, and it should be acknowledged rather than assumed.
The honest statement is that incidence rose, which is a real finding, and that the mechanism behind it is not observable here.
Is there any indirect evidence?
The persistence of the effect after the flight is suggestive. A gain that holds after media stops is more consistent with recruitment; one that decays with the media is more consistent with a frequency effect that needed the prompt.
That is weak evidence and it is available at no extra cost, which makes a post-period worth having for this reason as well as for pantry effects.
What about category new buyers?
If the category's total baskets rose in test stores, the campaign brought people to the category rather than only moving them between brands. That is the strongest possible outcome and it is directly measurable.
It is also rare, and claiming it without the category measurement is the kind of claim that will be checked.
How should this be reported?
As a decomposition with its limits stated. Incidence, quantity, and the explicit note that the mechanism behind incidence is not identified.
What would change the answer?
Access to identified shopper data, which some retail environments have and most of this channel does not. Where it exists, the question becomes directly answerable.
What should a brand do with a frequency-driven result?
Treat it as a result that depends on continued media, and plan accordingly. That is not a failure — a category bought on impulse may be genuinely media-dependent — and it is a different business case from a recruitment result.
Frequently asked questions
Can penetration be measured in this channel?
Basket penetration can, which counts baskets rather than people. Household penetration cannot without identifiers, and conflating the two is the most common error in this area.
Does a post-period really help?
It gives suggestive evidence about persistence at no additional cost, which is more than nothing. It is not a substitute for identified data.
What if incidence and quantity both rose?
That is a strong result and the mechanism still needs care in description. Report both and avoid asserting recruitment.
Does this differ by category?
Considerably. Impulse categories are more media-responsive and more likely to show frequency effects; planned categories are more likely to show genuine switching.
Can I infer new buyers from a competitor's decline?
That indicates switching rather than recruitment, which is a third outcome again. Category total is what distinguishes recruitment from switching.
Is loyalty data available anywhere in this channel?
Rarely, and where a network offers it the provenance and consent basis deserve scrutiny. Absence of identifiers is a structural property of independent retail.
What about tender type as a proxy?
It is a payment method, not a persona. Inferring shopper characteristics from how somebody pays is analytically unsound and should be declined rather than refined.
How should the limitation be worded?
Plainly: incidence rose, and whether that is more people or more trips is not observable in this data. That sentence protects the finding.
What is the most useful decomposition to report?
Baskets containing the item, and units per containing basket, in test against control. Two numbers, and they carry most of the available information.
### How should the finding be phrased in a report?
In terms of what was measured: baskets containing the item, and units per containing basket. Those are observations, and the interpretation that follows should be labelled as interpretation.
A report that says incidence rose and names the possible mechanisms without choosing between them is more credible than one that asserts recruitment.
What would it take to answer the question properly?
Identified shopper data, which most of this channel does not have. Where a brand has it through another route, combining sources may be possible, and the provenance and consent basis of any such data deserve scrutiny.
### What should a brand do with an ambiguous mechanism?
Plan for the more conservative interpretation and test the difference where the decision depends on it. Assuming recruitment when frequency is equally consistent with the data overstates the durability of the gain.
Where the distinction genuinely changes the business case, a design with a post-period long enough to observe persistence is the practical way to narrow it.
Does a category effect settle it?
It distinguishes recruitment into the category from switching within it, which is one of the two ambiguities. It does not distinguish more people from more trips.
How should this be worded for a retailer?
In category terms, because the retailer's question is whether the category grew. That is directly measurable and it avoids the ambiguity entirely.
Should this analysis be run on every campaign?
Where the data exists it costs little and it adds interpretation to a bare sales figure. It is most valuable on launches and least on routine activity. General measurement practice is discussed by bodies such as Geopath.
What is the honest single sentence?
Incidence rose and quantity did not, or the reverse, with the mechanism behind incidence not observable in this data.
Considering NRS Digital Media? Start at nrsdigitalmedia.com, read the media deck, or contact the team.