Search Funnel · ScaleSKUs MCP Academy #13

Your product is appearing in search, but purchases are not keeping up. Should you increase bids, change the advertised product, or investigate the offer first? Search Query Performance can help locate where the pattern weakens. In this lesson, we will compare two walnut products on the same query and reporting week. We will separate visibility from engagement, and evidence from explanation.

Start by asking for actual impressions, clicks, cart additions and purchases at product level. Request market shares separately. That simple distinction prevents a common mistake: treating a product's share of market activity as if it were its own conversion rate. The stronger product records forty-four thousand eight hundred and thirty-five impressions, one thousand seven hundred and forty-six clicks, one thousand and fifty-four cart additions, and two hundred and fifty-one purchases.

The comparison product receives forty thousand eight hundred and thirteen impressions, five hundred and ninety-five clicks, one hundred and ninety cart additions, and thirty-eight purchases. Both have substantial visibility on the same query. Yet their downstream activity differs sharply. That makes buying more exposure an incomplete answer. First investigate why the second product attracts much less engagement despite a broadly similar number of appearances.

The leader holds two point three five percent of query impressions, six point one six percent of clicks, and five point nine percent of purchases. The comparison product holds two point one four percent of impressions, two point one percent of clicks, and less than one percent of purchases. Each percentage uses its own market-wide denominator. Purchase share is the product's purchases divided by total purchases associated with that query.

It is not purchases divided by the product's clicks. Use those shares to compare competitive position at each stage, while keeping raw counts beside them so the size of the evidence remains clear. Here is an important follow-up: do these counts track the same people through every step? The answer is no. These are aggregated query and product activity counts.

Subtracting purchases from cart additions does not give us a verified number of abandoned carts. A person can interact more than once, and the report is not a session-level journey. Ratios between the counts can be descriptive signals, but they should not be narrated as measured cohort conversion. Our follow-up corrects that interpretation. The defensible conclusion is weaker downstream engagement relative to visibility, not a claim about exactly where individual shoppers left.

The first investigation concerns product and offer fit. Verify the actual variant and pack size before comparing the value shoppers saw. Then check effective price per quantity, displayed offer, delivery promise, ratings and availability during the same period. The comparison product's retail context includes ninety point two percent Buy Box, versus one hundred percent for the leader. That gives us a useful hypothesis, not proof of the cause.

Current availability alone cannot explain a historical week. Ask for dated evidence where it exists, and assign a manual check where the connected data cannot confirm what shoppers actually saw. The evidence does not justify removing the second product immediately or making the leader the sole owner of the query. Those would be targeting decisions, and the current analysis has not isolated their likely effect.

Likewise, adding another exact keyword or increasing a budget would not resolve an unverified product-fit problem. Keep the recommendation precise: investigate the weaker product's offer and attributes first, then decide whether a controlled exposure test is appropriate. This protects useful coverage while the explanation is checked. A strong diagnostic workflow tells the team which action to delay as clearly as it tells them what to investigate.

Finish with a short handoff containing five things: the query and dates, the two product identifiers, the count and share differences, the next evidence check, and the decision that check will inform. Give the follow-up an owner and review date when assigning it in your team's workflow. Do not fill missing evidence with an attractive story. For practice, choose a product with reasonable visibility but weaker downstream share, compare it with a relevant sibling, and ask what observation could change your explanation.

The goal is not a prettier funnel. It is a better next step, grounded in the actual information available. Keep this checklist beside your next review. Compare the same dates, confirm the metric denominators, write down the unanswered question, and only then decide whether a targeting experiment would answer it.