Basket effects of an in-store campaign
Most measurement of an in-store campaign looks at the advertised item. The basket it sits in is also observable, and looking at it answers questions the item-level view cannot: whether the campaign grew the category, moved volume from a competitor, or changed the value of the trip.
It is one of the genuine advantages of measuring in a channel with transaction data.
What can the basket show?
Whether the advertised item's gain came alongside category growth or alongside a competitor's decline. Those are different commercial outcomes with the same item-level result.
It also shows whether the basket as a whole grew, which is the retailer's question rather than the brand's.
Why does the retailer's question matter to a brand?
Because it determines how the campaign is received by the store. A campaign that grows the category makes the case for the channel to the retailer; one that only moves share between brands does not.
Bringing the category result to a retailer conversation is considerably more persuasive than bringing a brand result.
What is the complement effect?
A campaign for one item can lift a related one. Advertising a snack can lift a drink; advertising a meal component can lift the rest of the occasion.
Where those relationships exist, measuring only the advertised item understates the campaign. Checking a small set of pre-specified complements is cheap and frequently informative.
How should complements be chosen?
Before the analysis, from category knowledge or from co-purchase patterns in the baseline data. Selecting them afterwards, from whatever moved, is the multiple-comparisons problem and it will find something.
Pre-specifying two or three is the defensible version.
What about basket size and value?
Both are measurable and both are the retailer's metrics. A campaign that raises average basket value in test stores has an argument that transcends the advertised brand.
That is the strongest possible framing for a retail media channel and it is rarely measured because brands measure their own items.
What confounds basket analysis?
Everything that confounds item analysis, plus the additional risk of finding spurious relationships across many items. Restricting the analysis to pre-specified measures is the control.
How should it be reported?
Item effect, category effect, and the pre-specified complements, with the primary measure identified. Three views, clearly labelled, and no fishing.
What does this require from the network?
Transaction-level data with baskets, at store level, for test and control stores. A network with point-of-sale visibility can support it; one selling impressions alone cannot.
Frequently asked questions
Is basket analysis always worth doing?
Where the data exists it costs little and frequently answers the retailer-facing question. Where the campaign is purely brand awareness, the basket view may show nothing and the absence is itself informative.
How many complements should be tested?
Two or three, pre-specified. Testing many raises the chance of a spurious finding to near certainty and undermines the credible ones.
Can the campaign shrink the basket?
It can shift value within it, and a campaign that moves a shopper to a cheaper item lowers basket value while succeeding on its own terms. Reporting both keeps that visible.
Does this work for a category not bought in the same trip?
Less well, since the basket view is strongest for same-visit effects. A considered purchase made later is not in the observed basket.
What about substitution from a competitor?
It is observable as a competitor's decline in test stores relative to control, and it is the cleanest evidence of share movement available in this channel.
Should the retailer see this analysis?
Frequently yes, because it speaks to their question. It is the kind of contribution that changes how a brand's media is regarded by the store.
How large does the test need to be?
Basket-level effects are usually smaller than item effects, so more power is needed. Sizing for the basket measure rather than the item measure is the correct approach if the basket is the primary question.
Does this require loyalty data?
No. Basket composition is visible in transaction data without any shopper identifier, which is what makes it available in this channel.
What is the single most useful basket measure?
Category sales in test against control stores. It answers whether the campaign grew the category, which is the question that matters to everybody in the room. Channel context is collected by NACS.
### How should the complements be chosen and recorded?
From baseline co-purchase patterns or category knowledge, written down before the analysis with a stated reason for each. Two or three, no more.
Recording them prevents the post-hoc version, where whichever item moved is described as a complement. That version will find something in any dataset.
Does a basket effect strengthen the case with a retailer?
Considerably, because it speaks to the retailer's measure rather than the brand's. A campaign that raised category sales or basket value has an argument the store can act on.
### What should be reported to the brand team?
The item effect as the primary, the category effect as the retailer-facing measure, and the complements as secondary with their pre-specification noted. Three views, clearly ordered.
Ordering them prevents the common outcome where a secondary finding becomes the headline because it was the largest number in the analysis.
Does a basket effect survive scrutiny?
It does where the complements were pre-specified and the design was controlled. It does not where the analysis searched across many items, which is why the pre-specification matters.
Can this replace an item-level measure?
No, it complements it. The item measure answers the brand's question and the basket measures answer the retailer's.