What changed, and what remains unproven
Kevel announced on 2026-11-17 that Soriana Media in Mexico selected its advertising infrastructure. Initial placements are live; additional formats are planned for 2027. This announcement does not establish incremental sales.
The practical question for a brand is not simply whether an advertisement can be connected to a purchase. It is whether the advertising changes a commercial outcome enough to justify its cost, without making shopping harder. Those are different questions. A transaction may occur after an exposure even when the shopper already intended to buy. A retailer may also earn advertising revenue while the advertised supplier sees little additional contribution. A useful review must keep the objectives of retailer, supplier and shopper visible rather than compressing them into a single attractive performance number.
The remainder of this article is an original proposed workflow for a hypothetical retail-media pilot, not a description of Soriana's implementation or a promise about Kevel features. It deliberately starts with a narrow decision and a limited comparison. That is more useful than treating infrastructure selection as proof that a channel works for every product. The same reasoning can help a team assess an existing advertising placement without buying new software, changing customer permissions or assuming access to data it does not already have authorization to use.
Write a testable question before buying the placement
Choose a product and shopping situation that can be described without a large collection of audience labels. For example, a hypothetical household-goods supplier might ask whether a clearly identified sponsored placement increases completed purchases of an available product in eligible store areas. Specify the item, offer, placement, service area and observation window. Do not quietly replace the question with whether people clicked. The commercial decision is whether to continue, change or stop that specific placement, not whether the marketing team can produce a larger engagement report.
Write down the intended mechanism. Perhaps the placement makes an otherwise overlooked product easier to find. Perhaps it explains a useful difference between package sizes. Those hypotheses imply different creative and different checks. A discovery message should not hide the shopper's original search results, while a comparison message should not imply a saving that disappears when unit prices are considered. Ask a colleague to describe the proposed message using only the information a shopper will see. If they cannot explain the offer correctly, revise the material before allocating more traffic to it.
Also name the decision owner and the operating constraint that could stop the test. The brand may own creative, the retailer may own placement, and store operations may own availability. Agreement on these roles does not require a complicated governance program. A short shared brief can specify who approves the launch, who records an interruption and who can pause delivery. Keep the initial scope small enough that these people can investigate an unexpected result. A test that nobody can explain is not automatically valuable because it reaches many shoppers.
Establish the comparison and protect its meaning
A credible comparison begins before the campaign runs. Where the operating setup allows it, an analyst can consider random assignment of eligible units. If store groups are used instead, explain why their prior patterns, assortment, availability and local trading conditions are reasonably comparable. Do not present a convenient selection of strong stores as a neutral control. The design needs to match where exposure and purchases can actually be observed. If shoppers frequently move between the groups, discuss how that overlap could weaken the distinction between the test and comparison conditions.
Record the baseline period, the campaign period and the definition of a completed purchase. Distinguish orders placed from orders fulfilled, and decide how cancellations and returns will enter the review. Freeze the item mapping so that a packaging change does not silently become a different product midway through the analysis. When an item is unavailable, record the affected time rather than letting missing opportunity look like a weak advertising response. These operational details are not a substitute for statistical design, but without them even careful analysis can answer the wrong question.
Agree in advance on how to handle exceptions. A store closure, an unrelated promotion or a major price change may make a period difficult to compare. Do not remove an inconvenient observation simply because it lowers the apparent result. Preserve the original data scope, document the reason for any exclusion, and show how the decision changes with and without it when appropriate. If the available design cannot support a causal conclusion, label the result as descriptive or exploratory. Honest limits are more useful than a precise-looking percentage attached to an unsuitable comparison.
Check the shopping and operational experience together
Before launch, walk through the complete shopper task on the actual surfaces included in the pilot. Check that the advertisement is identifiable, the product information is accurate, the price matches the destination and the item can be obtained under the stated conditions. A useful placement should help a shopper make a suitable choice, not create a detour that obscures ordinary navigation. Keep a record of failed destinations and confusing substitutions. A campaign can appear successful at the click stage while sending people into unavailable stock or a checkout they cannot complete.
Define guardrails that reflect these risks. They might include product unavailability, cancellations, complaints about the offer and repeated failed attempts to reach the intended item. Choose a small set the operating team can genuinely monitor, rather than collecting a long list nobody reviews. A guardrail breach should trigger investigation or a pause according to the agreed plan. Do not wait for the final media report if the offer has become inaccurate. Correcting a broken shopping path is a separate responsibility from deciding whether the campaign generated additional demand.

Original explanatory visual by Karim Elganainy for the proposed pilot, not a platform interface or reported campaign result. News context: Kevel's partnership announcement.
The diagram is a checklist of connected responsibilities, not evidence that any particular software supplies every step. A retailer and a supplier should be able to point to the person responsible for each handoff. If a problem cannot be routed to someone who can act, adding another dashboard will not repair the operating gap.
Work through a small illustrative calculation
All numbers in this example are invented for teaching and are not campaign results, benchmarks or a recommended sample size. Suppose a pilot uses 8 stores split into 4 test stores and 4 comparison stores. In a baseline window, the groups record 200 and 180 completed units respectively. In an equally defined follow-up window, they record 230 and 190. The test group's change is 230 - 200 = 30 units; the comparison group's change is 190 - 180 = 10 units. Subtracting these changes gives 30 - 10 = 20 units as a simple adjusted difference.
That arithmetic is not sufficient to prove an advertising effect. It assumes the comparison is suitable and other relevant differences have been addressed. It says nothing by itself about uncertainty, shopper overlap, price changes or whether the pattern would persist. An analyst would need to assess the design and data before treating the difference as an estimate of causal impact. The example is useful because it shows why the test group's increase alone is not the entire story. Reporting all 30 units as advertising-driven would ignore the movement visible in the comparison group.
Suppose, only for this scenario, that the 20-unit adjusted difference were usable for planning and each additional unit contributed 5 currency units after the specified variable costs. The associated contribution would be 20 × 5 = 100 currency units. If the pilot media cost were 150, the limited balance would be 100 - 150 = -50 currency units. This excludes other unlisted costs and does not measure total retailer value. It demonstrates why more attributed revenue, or even a positive unit difference, does not automatically establish a worthwhile supplier investment.
Keep attribution, customer data and business value separate
An attribution report follows an agreed rule for linking exposure or interaction to a transaction. A test asks a different question about what changed relative to a comparison. Neither should be renamed to make the presentation sound stronger. Record the attribution window and matching definition if those reports are used, and explain which shoppers or transactions are not represented. Do not assume that the observable portion of purchasing describes the entire customer base. A tidy report can still omit unlinked purchases, unavailable identifiers or activity outside the measured scope.
Retailer-controlled information does not create unrestricted permission to share customer records with a brand. Use only the authorized data, purposes and recipients for the pilot. Prefer the least detailed information that can answer the agreed question, and involve the responsible privacy team when the proposed use changes. Aggregation can reduce detail but should not be described as a universal guarantee of anonymity. The practical aim is to make the test useful without collecting personal information simply because a platform or an analyst could technically process it.
Keep the supplier's economics distinct from the retailer's media revenue and the shopper's experience. A placement may be valuable to one party and less useful to another. A shared review can show these perspectives separately without pretending they are interchangeable. If a pilot improves product discovery but does not yet justify a larger budget, preserve that finding as a limited learning outcome. Do not manufacture a single success label that conceals disagreement about costs, service quality or the actual decision each participant needs to make.
Review the evidence before expanding the pilot
At the review, begin with what happened operationally: whether the intended placement ran, whether the product remained available and whether the comparison stayed usable. Then examine the commercial result and its limitations. Separate observed facts from explanations the team considers plausible. If a result is inconclusive, specify whether the problem is insufficient information, a compromised design or a genuinely small effect that the available evidence cannot distinguish. Do not describe every uncertain result as a failure of the creative, and do not promise that a larger budget will automatically resolve the uncertainty.
A useful decision record states what will happen next and why. Continue the same test only if that is appropriate to the original evaluation plan; revise the design if a specific defect makes it uninformative; stop if the offer or economics no longer meet the agreed conditions. Avoid repeatedly changing the outcome measure until something looks favorable. Preserve the creative version, product scope, comparison rules and interruption log so a later reviewer can understand the result. Without that record, an apparently successful pilot may become impossible to reproduce or challenge responsibly.
Expansion should test whether the original conditions still apply. A different store format, product category or service area may change availability and shopper needs. Treat the next phase as another decision with explicit assumptions, not as an entitlement earned by one positive chart. My practical preference is to scale the clarity of the decision process before scaling the spend. That means making the evidence and stop conditions easier to understand, rather than making the report longer or adding more channels to a pilot whose basic mechanism remains unclear.
What the announcement should prompt a marketing team to do
The useful response is to prepare a concrete test brief, not to assume that a new infrastructure partnership settles the measurement debate. Start with a shopper task, a suitable comparison, operational checks and a commercial decision that someone can own. The proposed workflow is general analysis; it does not establish availability in the GCC, access terms for a particular advertiser or measured results in another market. A team considering a local program should verify its own setup instead of importing assumptions from a foreign announcement.
Before approving spend, ask whether the planned evidence could genuinely change the decision. If every possible outcome will be described as a reason to invest more, the exercise is not a useful test. Define what would support continuation, what would require repair and what would justify stopping. Keep the distinction between a recorded sale and an incremental contribution visible in the final summary. Retail media becomes easier to manage when its operating promises are specific enough to examine, and when uncertainty can be reported without being treated as an embarrassment.

Comments
No published comments yet.