Part 3 - Favorite Brand’s Bra Size Methodology

In this post, we’re diving into Option 2: Using Favorite Brand’s Bra Size to determine the correct breast pump flange size. This approach combines a rule-based system with user feedback to refine recommendations over time, creating a hybrid recommender system that evolves with real-world data.

Why Use a Recommender System Instead of Generative AI?

While Generative AI (GenAI) can create new data patterns, we chose a recommender system because it is more suited to structured decision-making based on predefined rules and user preferences. Below are the key differences:

Advantages of Using a Recommender System

• More Reliable & Explainable – Rule-based recommendations ensure that users understand why they are getting a specific flange size recommendation.

• Faster & More Efficient – Requires less computational power compared to GenAI, which would need extensive fine-tuning and training on large datasets.

• Easier to Validate – Since we are using existing sizing charts and user feedback, we can verify the accuracy of recommendations more effectively.

• Better for Cold-Start Problems – A recommender system can begin with rule-based logic and improve over time with user feedback, whereas GenAI would struggle without a large training dataset.

Disadvantages of Not Using Generative AI

• Limited to Existing Data – Unlike GenAI, which can generate entirely new possibilities, a recommender system relies on available data and predefined mappings.

• Less Adaptable to Rare Cases – Edge cases that fall outside of common size patterns might not be handled as effectively.

• Slower to Evolve Without Data – While the system improves with feedback, it won’t create entirely new solutions like GenAI potentially could.

Ultimately, our approach ensures a structured, accurate, and transparent way to match users with the right flange size without requiring the computational complexity of Generative AI.

How It Works

This method leverages bra size charts from the top 10 brands to provide an initial recommendation based on the user’s preferred brand and size. However, because breasts grow during pregnancy and postpartum, we account for these changes by:

1. Asking users whether their inputted bra size is pre-pregnancy or post-pregnancy.

2. Adjusting pre-pregnancy sizes by an estimated 1 cup size based on average breast growth during lactation.

3. Using post-pregnancy sizes as direct inputs, assuming they reflect the user’s current state.

4. Mapping the adjusted size to the most appropriate breast pump flange size.

Initially, this will be entirely rule-based, meaning that we will directly correlate bra size to pump size based on manufacturer recommendations. However, we will gradually integrate user feedback to optimize and refine the system as we collect data over time.

Cold-Start Problem & Hybrid Approach

Like with our first approach, we’re encountering a cold-start problem—we have no existing user data to validate our recommendations. The advantage of this method is that bra size is a widely known and frequently used measurement, making it easier for users to input their data without extensive effort.

Why This Hybrid Approach?

• Phase 1: Rule-based recommendations using bra size charts.

• Phase 2: Collecting user feedback on whether the recommended flange size was correct.

• Phase 3: Optimizing recommendations using machine learning based on real-world sizing feedback.

This phased approach allows us to start with an effective, structured system while improving its accuracy over time.

Example: Bra Size to Pump Flange Conversion

Below is an example of a bra size chart, showing how we match sizes to breast pump flanges to give an idea of how the system works.

Note: Pre-pregnancy sizes would be adjusted based on an estimated cup-size increase.

User Flow & UI Implementation

To make this system intuitive and user-friendly, the user experience (UX) flow will follow these steps:

1. User Selects Brand & Size

   • Dropdown menu for top brands.

   • Input field for specific bra size.

2. Pre-Pregnancy or Post-Pregnancy Size?

   • If pre-pregnancy, apply 1 cup growth adjustment

   • If post-pregnancy, use inputted size directly.

3. Flange Size Recommendation

   • Output: Recommended pump size based on adjusted or direct input.

   • Confidence score: Based on rule-based mapping.

4. User Feedback Collection

   • Prompt users to confirm if the recommended size was comfortable and effective.

   • Allow manual adjustments and feedback.

Next Steps & Future Enhancements

While this method is easier for users than image recognition, it still has room for improvement:

• Expand brand database to cover a wider variety of sizing systems.

• Refine growth factor calculations using real user data.

• Improve accuracy through ML: Once enough feedback is collected, transition to an ML-driven approach.

• Integrate real-world testing with lactation consultants to validate recommendations.

By continuously improving the system and leveraging user feedback, we aim to provide a personalized, hassle-free experience for moms, helping them get the most out of their breast pump.

In the next post, I’ll discuss how we plan to implement Option 3: Manual Measurements, ensuring an even more customized approach to sizing. Stay tuned…