Offline Data-Driven Recommender Systems for Improving Small Business Marketing Strategies

Area of Science:

• Computer Science
• Marketing Analytics
• Business Intelligence

Background:

• Recommender systems enhance user engagement across various sectors.
• Small businesses face high cost per action (CPA) and low conversion rates (CR) with online marketing.
• Personalized marketing campaigns are key for identifying high-value customers.

Purpose of the Study:

• To develop a novel recommender system for small businesses to reduce marketing costs.
• To leverage offline interaction data for identifying customers likely to use discount coupons.
• To improve conversion rates (CR) and decrease cost per action (CPA) in marketing efforts.

Main Methods:

• Utilizing offline interaction data, including store-level coupon and point log data.
• Implementing tailored data augmentation techniques to address cold-start problems and data sparsity.
• Evaluating system performance using metrics such as CPA, CR, and root mean squared error.

Main Results:

• The proposed recommender system significantly outperforms conventional online marketing platforms.
• The system effectively identifies customers prone to using discount coupons, thereby increasing CR.
• Demonstrated reduction in marketing costs through improved efficiency.

Conclusions:

• Incorporating offline data with appropriate augmentation is valuable for cost-effective marketing.
• The novel recommender system offers a viable solution for small businesses struggling with online marketing expenses.
• Data augmentation techniques are crucial for overcoming challenges in offline recommender systems.