Product Ownership · ScaleSKUs MCP Academy #12

A good keyword is only half the decision. If you sell several products that match the same search, which one should receive the opportunity? In this lesson, we will compare walnut products for the search walnuts without shell. We will use query-level evidence, check the advertising structure, and build a test hypothesis. Start with a straightforward question: which of our products should own this search, and why?

Ask for the same complete week and the same query across every product. Otherwise, you can accidentally compare one product's broad retail success with another product's performance on a specific customer need. The first candidate ends in D zero six nine. For September twentieth through twenty-sixth, its impression share is two point three five percent, click share is six point one six percent, and purchase share is five point nine percent.

It records two hundred and fifty-one purchases associated with this query. Those measures answer different questions. Impression share tells us how much query visibility the product received. Click and purchase shares describe its share of those respective activities. The growing shares are evidence of query affinity. They do not prove that advertising caused the purchases, or that this product will win every future test.

The secondary product ends in C five three E. It receives two point two percent impression share, close to the primary candidate, but only one point two eight percent click share and one point five one percent purchase share. That represents sixty-four query purchases. Now compare the retail context. The secondary product has a higher unit-session percentage, but that is units per session, not the percentage of visitors who placed an order.

It also covers all the product's retail traffic, not just this query. A stronger overall retail metric cannot automatically overturn weaker evidence for the particular search we are evaluating. Both products already sit in the broader walnut phrase structure. That is useful: this is not a completely unknown product-and-query relationship. But the ad group contains six products. A search-term report aggregated at that level does not divide every order between the six products for us.

Ask which mappings are directly available and where attribution stops. Also check the actual product attributes. The synced titles identify NutriGrove Walnuts, but pack size is unverified. Do not turn an ad-group name into a product fact. Before isolating an owner, confirm that the product really matches the shopper's intended variant and quantity. Our follow-up makes the decision more useful: treat this as a provisional test, and separate what we know from what remains unknown.

The primary candidate has stronger query-specific evidence. The secondary remains a comparison candidate. Neither observation proves incremental advertising lift. Establish a baseline across several complete weeks, including query shares, available ad delivery metrics, retail performance, price, availability and Buy Box. Keep the scope stable. If you change price, targeting, creative and product allocation together, the result becomes difficult to interpret.

The aim is a test with a clear question, a defined product pair and a measurable outcome. When the test is appropriate, evaluate more than whether the chosen product gained sales. If one of your products grows while another loses the same amount, you may have moved demand within your own portfolio. Track combined walnut performance and the query-level result alongside each product.

Compare exposure before comparing efficiency: a product given much more visibility is not a fair control by default. Watch stock and offer changes during the period. Use a suitable experiment if you need a causal claim. An observational comparison can guide a decision, but it should not be described as proof of incremental sales. The final output should name the query, primary candidate, secondary candidate, supporting evidence, unknown attributes and the conditions for a test.

In our example, the evidence supports a provisional primary candidate, while ownership stays unchanged until product fit and the baseline are checked. That is a practical result, not an unfinished analysis. Your team now knows what to verify and what would justify the next action. Try this workflow on one high-value search with two or more relevant products. Ask why the recommendation could be wrong.

The answer should identify a specific missing fact or measurable condition, not simply repeat the ranking.