The Company

LuisaViaRoma is one of the leading online luxury fashion retailers in the world. LuisaViaRoma offers a curated selection of clothing and accessories for men, women, and children, as well as beauty and home items, featuring collections from over 500 esteemed international luxury brands.

The Challenge

LuisaViaRoma had the clear objective of optimizing its customer acquisition campaigns’ conversion rate, which was low compared to product views or even add to cart. The challenge was derived from the volume of conversions, insufficient for Google’s algorithms to establish an optimal pattern for optimizing the campaigns for this KPI.

The Approach

The project started with a comprehension of LuisaViaRoma business model and examined other additional data sources to be included in the predictive model. Making Science together with Google decided to incorporate new signals from earlier stages of the sales cycle (product view and add to cart), to make predictions about purchase likelihood at those stages. This approach was valid thanks to data availability for both training the model (historical data provided by LuisaViaRoma) and making predictions (data tracking on website). This way, whenever a user visits the website and performs any of these actions, a score is sent to Google in real-time. Thanks to this approach, available data volumes increased significantly, permitting the optimization to sales as requested by LuisaViaRoma.

The Results

With an A/B test with equal investment  (control: tROAS using purchase data / test: tROAS using not just purchase data, but also propensity-to-purchase data), we measured the incremental impact of including this predictive model into the custom bidding of DV360 campaigns:

• Sales: +36%
• Revenue: +106%: this is contributing to LuisaViaRoma’s +27% revenue growth BO for 2023, while also increasing brand consideration
• ROAS: +100%
• CPA: -25%