GenAI Solution for Effective Upselling and Cross-Selling

The Client

Our client operates a vendor platform that connects restaurants with corporate offices, hospitals, universities, retail locations, and distribution centers, serving more than 10 million meals per year. It also offers popup events, café services, and food delivery perks for companies.

Challenge

Our team needed to implement a feature within the client’s existing food delivery app to increase average order value and enhance the customer experience by intelligently suggesting relevant items to upsell and cross-sell. Traditional methods like manual recommendations or basic collaborative filtering were considered too time-consuming and costly to develop and maintain.

Solution

Leveraging the power of generative AI, specifically large language models (LLMs) like GPT-4o, GPT-4o-mini, and Gemini Flash, First Line Software implemented a system to dynamically generate personalized upsell and cross-sell suggestions within the client’s mobile app. This system analyzes the user’s current order and generates recommendations and explanations for why the items might be appealing to the existing order. The system also collects user feedback on these suggestions to continuously improve its performance and personalize the experience further.

Key Features

Dynamic Recommendations

AI generates upsell and cross-sell suggestions in real-time based on the user’s order and preferences.

Explanations

The AI offers clear insights in natural language into why certain items are recommended, giving users more control and helping them make informed decisions, ultimately enhancing their shopping experience.

Cost-Efficiency

LLMs offer a cost-effective alternative to traditional machine learning models for upselling and cross-selling. LMs are more adaptive, require less manual data input, and offer faster results than other models. It speeds up time to market and reduces the cost of development.

Personalization

The system learns from user feedback to refine recommendations and cater to individual tastes.

Scalability

The GenAI solution can be easily scaled to accommodate a growing user base and menu items. It also allows for integrating the exact solutions into different parts of the sales process, not just in the app but also on the website and other platforms.

Key Benefits

Contextual Cross-selling

When users add an item to their order, the AI suggests a complementary item leveraging the AI’s understanding of food pairings and preferences. The AI can go beyond basic pairings to suggest diverse cross-sell options, such as adding a side dish or recommending a specific drink.

Explanatory Upselling

When a user selects an item, the AI suggests an upgrade, while explaining the value proposition of the upgrade, highlighting the extra items included. Building trust and transparency by explaining why an item is suggested, the AI builds trust with the user and increases the likelihood of conversion. This transparency is crucial for upselling, as users need to understand the value they’re getting for the extra cost.

Location-based Suggestions

The system uses location to suggest items based on the current weather. For example, on a cold day in Seattle, the AI may recommend warm and comforting dishes. Restaurants could upload their sales data to the platform, allowing the AI to analyze order patterns and identify popular pairings. This provides valuable insights for menu optimization and targeted promotions.

Implementation

Outcomes

Increased Average Order Value

Early results indicate a potential increase in average order value due to the AI’s ability to suggest relevant and enticing add-ons.

Enhanced Customer Experience

The personalized recommendations and explanations provide a more engaging and satisfying user experience.

Cost-Effective Solution

Implementing LLMs for this task is highly affordable, with ongoing expenses remaining relatively low. For one particular client, A/B testing of the system powered by LLMs was conducted for just $16, demonstrating its cost-efficiency even in real-world scenarios.

Rapid Implementation

The GenAI solution was implemented within a few sprints, allowing for quick deployment and faster return on investment.

Continuous Improvement

The system’s ability to collect and learn from user feedback ensures ongoing optimization and personalization.

Future Considerations

Hyper-Personalization

Incorporating user preferences, order history, and even location-based data (e.g., weather) to further personalize recommendations.

Monetization

Exploring opportunities to monetize the AI’s capabilities by offering premium features to restaurants or leveraging collected data for market insights.

Integration with Predictive Models

Combining LLMs with traditional predictive models to enhance accuracy and provide even more targeted suggestions.