The goal of this project was to increase revenue and customer lifetime value (LTV) by improving the customer journey.
Since customers with chronic health conditions frequently return to purchase the same or related products, they became a key audience for personalized recommendations. To deliver more relevant suggestions, we introduced machine learning–powered recommendation blocks.
This case study focuses on the design of the internal platform used to configure recommendation algorithms, manage placements, and analyze their performance.
In the Placements tab I added a new section — Recommendation blocks. They can appear on the homepage, listing, cart, and post-checkout.
Depending on the platform, 2 or 4 positions are visible above the fold — these are highlighted. The table shows the position number, recommendation algorithm or specific SKU, and managers needed impressions, clicks, CTR, and buyout metrics. If a SKU is unavailable in a region, a warning indicator with a tooltip is shown.
When entering a SKU, we show that it's unavailable in certain regions — but it can still be selected.
We considered adding drag-and-drop to reorder rows, but managers preferred editing positions manually — it felt more precise and reduced accidental changes.
Clicking 'View statistics' takes the manager to the monitoring tab, where they can view stats for selected positions. They can also explore different metrics in the table and configure which ones are shown.
Managers can drill into per-metric trends and compare periods — the analytics view is reachable both from the statistics table and directly from each placement block.
