A sporting goods brand achieved 2.38x ($220K) in incremental sales uplift and saved $18,000 in marketing costs using data-driven marketing analytics and machine learning.
INTRO AND CLIENT BACKGROUND
We recently teamed up with a sporting-goods store chain headquartered in Houston, Texas. With multiple locations across the country, they offer a wide range of sports equipment, athletic wear and outdoor gear to fitness enthusiasts and athletes.
With robust loyalty programs and direct marketing, the client's success is deeply rooted in the capture and analysis of transaction-rich data.
Their marketing model revolves around using customer insights to cultivate meaningful brand-customer relationships, positioning them as a leader in the industry.
BUSINESS CHALLENGES
→ Despite investing $65,000 annually in marketing efforts, the brand struggled to significantly increase customer's card spend.
→ Their previous campaigns resulted in a mere 2% increase in their sales through credit cards, far below their growth targets.
→ Unable to identify and target high-potential customers due to lack of leveraging existing customer and sales data effectively.
THE SOLUTION
We dived deep into the business problems and implemented a focused marketing analytics and ML approach:
1. Data QC, EDA and Feature Engineering
→ Consolidated data from different sources such as customer transactions, product info, past marketing campaigns, and loyalty programs into a unified Data Lake.
→ Performed QC (Quality Checks), EDA (Exploratory Data Analysis) and Pre-processing of historical data, covering more then 16M transactions from ~5M unique customers across multiple channels.
→ Utilized multiple feature engineering and selection techniques to iteratively come up with the best possible 51 feature list that has the highest impact on the model performance.
2. Propensity Modeling
→ After a lot of analysis and validations, our team considered two-years of data including 12 months for observation window and 12 months for performance window for model building.
→ Observation window - Historical data to understand customer's behavioural patterns.
→ Performance window - To understand the customers spend in the next 12 months from observation window.
→ Trained & Evaluated machine learning models using bagging and boosting algorithms such as Random Forests, XGBoost and LGBM in order to predict the likelihood of a customer's expected wallet spend in the next 12 months to target them effectively.
→ Utilized evaluation metrics such as MSE, RMSE, MAE, Lift and Gain charts for choosing the best and robust generalized model.
3. Target Audience Selection
→ Worked with the marketing campaign measurement and ops team to select the right target audiences for their campaigns based on propensity scores and rank-wise segments.
VALUE DELIVERED: WHAT WAS THE BUSINESS IMPACT?
The innovative solutions implemented led to tangible benefits across various stakeholders:
Significant Revenue Growth
→ The model enabled the identification of customers likely to increase their spend using brand's credit card.
→ By strategically targeting high-potential audience, the solution contributed to a remarkable 2.38x in incremental sales uplift - $220,000 in additional revenue.
Enhanced Marketing Strategies
→ This solution allowed resources to be focused on the audiences most likely to drive card spend, resulting in reduced marketing cost to company of approximately $18,000.
Strategic Decision-Making
→ The insights gained from the model inform a spectrum of decisions, ranging from promotional strategies to the structuring of rewards programs for future focused targeted campaigns.
TECH STACK
CONCLUSION
This case study underscores the power of AI and machine learning in reshaping marketing dynamics within the retail and e-commerce sector.
By leveraging data-driven insights, our client not only enhanced customer engagement but also witnessed tangible improvements in spend and overall revenue.
The success story stands as a testament to the transformative potential of AI and ML in driving strategic marketing decisions in the financial services domain.
Comments