Introducing: Recommendations Testing

In today’s hypercompetitive digital environment, consumers expect all their experiences to be personalized and relevant, including the recommendations they receive from eCommerce and media companies. Recommendations are an essential tool for businesses to improve conversion rates and customer retention, and serving relevant products and content to each visitor is imperative for success. In fact, Dynamic Yield has found that visitors who engage with personalized product recommendations ultimately yield a 2.8X increase in Revenue Per Visit compared to all site visitors – a substantial uplift for any business attempting to increase top-line revenue.

However, while recommendations already act as a core component of the digital strategy for many businesses, simply deploying recommendation widgets on a digital property does not guarantee success. After a recommendation widget is implemented, a business must ensure it is optimized and that campaigns are operating at maximum performance, a process often overlooked. As recommendations can drive a significant percentage of business revenue, continuous optimization and testing of recommendation widgets for optimal results is a necessity.

• A marketer running a campaign on product detail pages (PDPs) could test to see if a recommendation widget with a “Bought Together” strategy or a mixed strategy (with a mix of “Viewed Together” and “Similar Item” strategies) performed better.
• Merchandisers can test different configurations of merchandising rules in Dynamic Yield and measure the impacts on overall performance.
• Marketers can test different layouts – a recommendation carousel vs “Pinterest”-style – for recommendation widgets on their homepage.
• Marketers could target and test different recommendation strategies for each audience segment – whether they should show widgets with a “Most Popular” strategy or “Similar Items” strategy to paid search traffic on a product detail page.

Learn more about recommendations with Dynamic Yield, below: