Success and Guardrail Metrics for Item-Page Recommendations in Target's Online Store

Target's online store is about to launch a feature that shows shoppers a set of suggested items while they are looking at a product page. You are the product manager for this launch. Define the metrics you would use to decide whether the launch succeeded: what the feature is for, which metric you would treat as the primary measure of success, which supporting metrics you would track, and which signals would tell you the feature is doing harm. ```hint Clicks are not the goal A suggestion that pulls a shopper away from the item they came to buy can look like engagement without adding anything. Think about what the store should gain per visit, and how you would tell the feature's own effect apart from purchases that would have happened anyway. ``` ### Clarifying Questions - What do the suggestions consist of: similar alternatives to the item on the page, complementary items that go with it, or personalized picks based on the shopper's searches and past purchases? - Do guest shoppers with no history see the module too, and does it launch on both the website and the mobile app? - Will the launch run as an experiment in which some shoppers do not see the module, or go to everyone at once? - For this launch, does Target care most about revenue per visit, helping shoppers discover more of the catalog, or a smoother path to the right item? ### What a Strong Answer Covers - A clear goal for the module that serves both shoppers and Target's business, stated before any metric is chosen. - One primary success metric measured across the whole visit, so that sales lost on the original item are counted, not only sales gained through suggestions. - A supporting funnel from module impression to click, add-to-cart and purchase, broken down by shopper type, device and suggestion type. - Guardrails covering page speed, conversion of the original item, returns and the shopping experience. - A launch plan that attributes impact through a controlled comparison and separates launch-week health checks from the longer-term readout. ### Follow-up Questions - Shoppers click suggestions often, but revenue per visitor in the test group is flat compared with control. What could explain it, and what would you check first? - Suppose guest shoppers with no history see only popular items from the same category. Would you judge the module differently for them, and how? - After launch, the business wants to place sponsored products from brands in the module. Which metrics would you add or watch more closely?