Founded in 1931, Maurices is a popular online and in-store retailer with over 900 stores in North America that inspires all women to live their most authentic lives through inclusive sizing (0-24) and affordable pricing. Maurices and Tinuiti partnered in 2020 to maximize digital sales and drive cutting-edge, efficient, and effective new customer prospecting.
Like most retailers, Maurices’ in-store revenue was impacted by COVID-19 with temporary store closures and consumers shifting to digital purchasing methods. Within the critical Q4 holiday period, Maurices needed their digital advertising dollars to work harder, driving increases in both year-over-year revenue and new customers.
To solve these challenges, we deployed machine learning models in a new way. We analyzed Maurices’ website visitor audiences and segmented visitors by likelihood of purchase. By leveraging machine-learning technology, Maurices could gather real-time predictions on a user’s purchase intent and then alter the media bids and creative messaging strategies accordingly.
The machine-learning audience integration was aimed at answering the critical question: Can retailers drive a more efficient return on investment by leveraging machine learning to segment website visitors by their predicted likelihood of purchase?
Step 1: Fueling the model with Maurices’ website behavior data
The first step of testing was to train the predictive sales model. Over the course of three weeks, the model tracked each website visitor’s actions, taking into account over 400 real-time signals. The model tapped into existing Facebook and Google pixels in order to gather the data in real time.
Step 2: Evaluating Maurices’ website visitor segments
After training the model, we evaluated segments for website visitors and split them into cohorts. We then built these cohorts so that they would incorporate seamlessly into Maurice's social channels and prepared our activation strategy.
Step 3: Optimizing bids & the creative messaging strategy
After segmenting website visitors into cohorts, we set out to optimize our media buying. We customized bids and spend levels based on what we had learned about the cohorts. We also simultaneously tailored the creative for each audience.
These concurrent strategies aimed at improving media efficiencies (ROAS and revenue) while maximizing Maurices’ profit from each sale.
Leveraging a real-time purchase intent model led to significant improvements in Maurices revenue efficiencies and new customer growth. After a month of implementing the machine-learning model, Maurices was able to scale new customer acquisition efforts 30% while simultaneously decreasing the cost per acquired customer.
In Q4 2020, the tiered purchase intent segmentation drove a 2x higher return on investment compared to the paid social average. Through creative testing, we learned that high-purchase intent users still convert on full-priced products. We were then able to reserve the highest value discounts for low- and medium-purchase intent users, preserving margin.
New customer acquisition efforts were also more effective with the machine-learning model. Purchase intent segments and lookalikes of the high-intent segments were responsible for driving over 65,000 new customers. On average these customers were acquired at 48% lower cost per acquisition than the average.
Overall, the predictive model helped drive 25% higher social revenue year-over-year and at a 25% higher ROI.