Testing a Recently Viewed Row on an App Home Screen

Search logs where returning visitors type the same product queries again, or session recordings of them taking long routes back to a product page they already viewed, make a recently viewed strip worth testing on the homepage. If the strip works, these visitors reach product pages by a shorter path and total purchases rise. If visitors find the product and still do not buy, the problem is not navigation, so look at the product page before adding a strip.

Purchase conversion rate (CR)

Compare purchases per assigned returning visitor with eligible browsing history.

Also track

• Strip click-through rateUse this only as a diagnostic in the variant, where the strip exists.
• Product detail visitsMeasure total product-page visits, not only clicks from the strip.
• Bestseller clicksReview clicks on bestsellers alongside total purchase conversion.
• Homepage exit rateMake sure exits do not increase.

• StripAdd a strip of products the visitor viewed most recently.
• LayoutKeep placement, product count and dimensions fixed.
• EligibilityExclude purchased and unavailable products using the same predefined rules.
• IncrementalityCheck whether strip interactions lead to additional purchases or redirect existing browsing.
• PlatformReview the same comparison within each supported platform.

1. 1Strip click-through rate exists only in the variant, so treat it as a diagnostic rather than grounds for a launch. If bestseller clicks fall while total product detail visits stay flat, clicks have simply shifted; purchase conversion is the measure that counts. Also confirm the homepage exit rate does not rise, and keep the strip small enough that it does not fill the whole first screen, or the rest of the page gets harder to reach.
2. 2The finding covers this strip layout, the current recommendation algorithm and returning visitors with eligible history; if you change the algorithm in the same test, you cannot tell which change caused the difference. Read each platform on its own, and leave products from private browsing sessions out of the strip.
3. 3Returning visitors with eligible history may be a small pool, so an unclear result often just means too little data. Give the sample size calculator today's conversion rate and the lift worth detecting, then see how many days of eligible visitors per arm it takes to reach the sample per variant.