A kitchen can change its purchasing habits without abandoning a familiar ingredient. On 21 August 2024, Reuters reported a shift towards cheaper sunflower oil in Spain as expensive olive oil strained household budgets. The report described some households buying both products rather than continuing to buy olive oil alone. That distinction opens a more useful business question than which oil has won: how should sellers respond when customers divide their purchases differently?
This analysis takes that dated report as its starting point. It does not estimate the current market, reproduce a household survey or predict that a change in the shopping basket will be permanent. The examples below are deliberately hypothetical. They examine decisions a retailer could face when customers preserve selected uses for one product while choosing another for the rest. The central issue is the allocation of a limited budget, and the way that allocation travels from a kitchen cupboard to a store's ordering system.
A category switch can be a division of uses
Calling a household an olive-oil buyer or a sunflower-oil buyer can hide the most commercially important possibility: it may be both. A shopper could reserve a preferred product for occasions where its taste matters most and choose a less expensive alternative for other meals. That is a possible interpretation of mixed purchasing, not a claim about the cooking practices of every household. Its significance is that a fall in litres bought need not mean the preferred product has lost all relevance.
A retailer that interprets every lost litre as a lost customer may react too aggressively. Removing the less frequently purchased option could make the assortment less useful to someone who still wants it occasionally. Conversely, keeping exactly the old shelf allocation could leave too little space for the product now bought more often. The decision is not simply whether to retain a category. It is how much variety, space and replenishment effort each category warrants under a changed pattern of demand.
There is also a difference between a temporary compromise and a newly preferred routine. A household might switch only while a price gap feels unusually large. Another might discover that a mixed basket suits it even after that gap narrows. A sales total cannot separate these explanations on its own. A merchant would need repeated observations over time, with attention to availability and promotions, before treating one explanation as the basis for a long-lived assortment decision.
Litres, spending and customers answer different questions
Three measures are easy to confuse. Volume describes the amount sold. Revenue describes the money received. Customer count describes the number of buyers recorded within a defined period. None automatically supplies the others. A more expensive product can lose volume while preserving a substantial share of spending. A cheaper product can gain litres without generating the same cash contribution. A shopper can buy less of a product but remain an active customer for it.
The comparison period matters just as much as the unit. A calendar half-year, an agricultural marketing season and a rolling set of recent weeks are different windows. Supplier shipments are also not necessarily the same event as a household purchase at the checkout. Goods may move into distribution before consumers take them home. An analysis that puts these measures next to each other must explain their scope rather than assume that the similar product names make them interchangeable.
A useful measurement checklist
- Identify the reporting period and use the same period for the comparison.
- Separate litres, sales value, transactions and unique buyers.
- Check whether the record concerns supplier deliveries, store sales or household purchases.
- Record missing stock and promotions before interpreting a change as preference.
- Keep product grades and package sizes visible when calculating an average price.
This is why a simple ranking is a starting point rather than a complete explanation. A ranking can show which recorded volume is larger within its own dataset. It does not identify the cause of each customer's choice, the profitability of each sale or the permanence of the change. Those are additional questions requiring additional evidence. A sound business response keeps the useful signal without asking it to carry conclusions it cannot support.
A hypothetical basket makes the trade-off visible
Consider an invented household that buys four litres of cooking oil during a month. Assume, solely for this example, that its preferred oil costs eight currency units per litre and an alternative costs two. Buying all four litres of the preferred product would cost thirty-two units. Buying one litre of it and three litres of the alternative would cost fourteen. The eighteen-unit difference is arithmetic within these assumptions, not an estimate of savings achieved by Spanish households.
The household in this example has not stopped buying its preferred oil. It has reduced its share of the basket from all four litres to one. The seller therefore sees several changes at once: lower volume for one product, higher volume for another and lower total spending on the basket. Counting only whether the household purchased the preferred oil would miss the scale of the change. Counting only litres would miss its continued willingness to buy a limited amount.
Now suppose the preferred oil in the same invented example falls to six units per litre. The mixed basket would cost twelve units, while a four-litre basket containing only that oil would cost twenty-four. A price reduction has narrowed the difference, but it has not eliminated it. Nothing in the arithmetic tells us which basket the household would choose. That choice depends on preferences and budget constraints that the example does not measure. The point is to make the remaining trade-off explicit.
This exercise deliberately keeps the total quantity constant. Real customers could also change how much they buy, use existing cupboard stock or shift the date of their next purchase. Those possibilities would alter the calculation. Holding quantity fixed is useful for isolating the effect of a change in composition, but it is a limitation of the example, not a description of real behaviour. An analyst should state that limitation before using the calculation to guide a decision.
Package size can change the decision at the shelf
The amount paid today and the price per litre are separate considerations. A larger container could offer a lower unit price while requiring a larger immediate payment. A smaller container could cost more per litre but fit within the customer's available spending for that trip. Neither observation proves which format is best. It shows why a category manager should not infer affordability from unit price alone or assume that the largest package will always be the most useful offer.
For a household buying a preferred oil less frequently, a smaller pack might also fit its revised pattern of use. That is a hypothesis worth testing rather than a universal prescription. The store would need to observe whether the smaller format attracts additional purchases or merely moves existing buyers into a less efficient package. It would also need to consider the cost of carrying another item. More choice is not automatically better if it adds complexity without answering a distinct customer need.
A practical assortment experiment could retain a basic range while changing a limited number of facings or pack sizes. The comparison should be planned before the change, with a clear measure of success and a defined duration. Otherwise, a favourable week can be mistaken for proof that the intervention worked. The objective is to discover which offer remains useful under the new budget constraint, not to create a retrospective story that justifies whatever the store happened to sell.
Replenishment has to follow the new mix
Changes in demand composition create operational questions even if the total amount of oil sold is stable. A store ordering each product according to an old pattern could repeatedly run short of the cheaper option while holding more of the other than it needs. The combined category total might conceal both problems. Replenishment therefore needs to be examined at the level where shortages and excess stock actually occur, rather than only through the aggregate sales line.
An out-of-stock event complicates interpretation. If customers buy an alternative because their intended choice is unavailable, recorded sales can exaggerate a voluntary preference shift. If they leave without buying, the store may not see the unmet demand in its transaction data at all. Neither effect can be quantified from a headline about category volumes. Recording availability alongside sales makes a later explanation more credible and reduces the risk of using an operational failure as evidence of consumer choice.
The reverse problem appears when a promotion prompts customers to buy ahead. A temporary burst of sales can be followed by a quiet period while households use what they already purchased. Ordering permanently higher volumes after the burst could create excess stock. This is a scenario to consider, not a measured feature of the 2024 episode. It illustrates why the timing of purchases matters when translating a short observation into a longer replenishment plan.
A price cut is an experiment with several possible outcomes
A lower shelf price might bring back some volume, preserve existing buyers, encourage stock-building or simply reduce revenue on purchases that would have happened anyway. These possibilities are not mutually exclusive. The store cannot identify their relative importance merely by observing that more bottles were sold during a discount. It needs a comparison that asks what would plausibly have happened without the change, while recognising the limitations of any available comparison group.
For example, a retailer could compare similar stores over the same weeks, provided their availability, customer mix and other promotions are sufficiently comparable. Alternatively, it could examine a longer sequence before and after a change while documenting events that complicate the interpretation. Neither approach creates certainty by itself. A local event, a competitor's offer or a supply interruption might still affect the result. The quality of the decision depends partly on whether those alternative explanations are acknowledged.
The outcome should also match the purpose of the intervention. If the aim is to keep occasional buyers engaged with a category, a measure of repeat purchasing may be more relevant than a one-week volume peak. If the aim is to reduce excess stock, the time needed to sell existing inventory matters. If the aim is to improve category contribution, revenue alone is insufficient because acquisition and operating costs also enter the calculation. Clear objectives prevent a convenient metric from replacing the actual business question.
What would count as a return?
A return to the preferred oil can mean several things: buying it again after an absence, buying it more frequently, increasing the amount per purchase or restoring its former share of a mixed basket. These are not equivalent events. A customer who adds a small bottle to an otherwise unchanged basket has returned under one definition but not under another. Before announcing a recovery, the analyst should say which definition is being used and why it fits the decision at hand.
There is a similar distinction between a category recovering and a particular brand recovering. A shopper might resume buying a type of oil while choosing a different label or package. A retailer's aggregate category figures could improve while a supplier's position remains weak. Conversely, one brand could gain from another without the overall category expanding. Separating these levels avoids attributing every movement to a broad consumer trend when some of it may reflect competition within the shelf.
The most informative evidence would follow the same clearly defined measures through several purchasing cycles. It would show whether a change persists after a promotion, whether stock availability was adequate and whether the composition of the basket moves again when relative prices change. Even then, the result describes an observed group over an observed period. It should not be converted into a claim that all households have permanently changed their habits or that a familiar ingredient has lost its place in the kitchen.
Close the experiment with an explicit decision
An assortment test is useful only if a decision follows it. Before observation begins, the retailer can record which result would justify retaining the change, which would require further investigation and which would support returning to the previous arrangement. This need not involve an elaborate statistical system. Even a small store can state its question, preserve the starting information and agree on a review date. That discipline distinguishes a testable proposition from an ordinary rearrangement that different participants later explain in different ways.
The review should retain inconvenient observations. Sales of a new pack might rise while the old one is unavailable, rather than because the new format better meets demand. Customers might praise the expanded choice but buy the same quantity. Staff might spend more time managing the display than expected. These are possible circumstances of a hypothetical experiment, not observations about a named retailer. They show why a useful record contains both the outcome and the conditions under which it occurred.
Finally, the decision should remain reversible. A successful test in one store does not automatically transfer to another neighbourhood, season or relative-price environment. Extending the change requires preserving the comparison logic and checking whether its assumptions still hold. Otherwise a cautious local finding becomes a broad rule without fresh evidence. In a category where customers can change quantity, frequency and basket composition together, the ability to revise an arrangement is part of good management rather than an admission that the original idea failed.
The retailer's task is to accommodate a more selective basket
The 2024 report provides a starting signal; the analysis above explains why that signal calls for more than a winner-and-loser interpretation. A budget-constrained customer can remain attached to a product while buying less of it. That creates room for a mixed assortment, carefully chosen formats and ordering rules that respond to actual demand. It also creates a need for restraint: a seller should not remove useful options simply because a short period makes another product look dominant.
The strongest response is a sequence of explicit questions. What changed in the recorded basket? Which part is explained by price, which by availability, and which remains uncertain? What limited adjustment can the retailer test without assuming an irreversible shift? What evidence would justify keeping or reversing that adjustment? This approach does not promise a universal solution. It turns a broad market observation into a set of decisions that can be examined, measured and revised as the customer keeps choosing.