Bloomreach’s Approach to AI and Hyper-Personalization

Xun Wang
Xun Wang

In the realm of ecommerce, personalization is not just a trend but a cornerstone of modern digital strategies. As consumers increasingly shift their activities online, their expectations for a tailored shopping experience that caters to their individual preferences have intensified.

However, achieving personalization at scale is a formidable challenge for many ecommerce brands. Effective personalization involves intricate data analysis, generating numerous content variations, and continuous testing to discern what truly resonates with consumers. The complexity lies in the data analysis and the limited capacity to generate and manage diverse content variations efficiently.

Recent advancements in generative AI and large language models (LLMs) are transforming this landscape by providing limitless opportunities to personalize the entire shopping journey at an individual level. However, it’s critical to recognize that not all AI solutions are created equal. In this discussion, I will delve into the limitations of generic AI models and outline Bloomreach’s strategic approach to refining personalization through AI.

The Limitations of Generic AI Models

Generic AI tools — such as ChatGPT — are built on expansive foundational models trained on diverse internet data and fine-tuned on general human preferences through techniques like reinforced learning with human feedback (RLHF). These tools typically operate by generating responses based on direct user prompts.

While impressive in general applications, these generic models often fall short of delivering the true personalization needed in specific ecommerce scenarios. For instance, consider the different needs of a “searcher” who knows exactly what they want versus a “seeker” who has a vague idea. A seeker might only know they need to find a dress for a wedding in Hawaii but without specific details in mind. A foundational model could provide generic advice or suggest a range of popular dresses, but this lacks the nuanced assistance provided by a skilled shopping assistant (see Appendix A for a sample output from GPT-4 Turbo SOTA).

Woman receiving irrelevant product results from generic AI models like ChatGPT

Bloomreach’s Dual Approach to Ecommerce Challenges

At Bloomreach, we recognize the limitations of off-the-shelf AI models for specialized ecommerce applications. Instead of building our models from scratch, which is resource-intensive, we strategically utilize existing foundational models. We enhance these models by fine-tuning them with our aggregated commerce data and employing commerce-specific prompt engineering. Additionally, we integrate these models with our real-time data APIs during model inference to incorporate individual user preferences.

Hyper-Personalization for Commerce

Rather than relying solely on generic foundational models, we leverage and fine-tune these models using extensive commerce data gathered from our industry experience. This approach allows us to develop proprietary, industry-aligned commerce models through supervised fine-tuning and curated datasets on top of foundational models.

This method enables us to rapidly scale and effectively serve a diverse range of businesses with various needs at a fraction of the cost of developing bespoke generative AI models. We complement these fine-tuned models with our advanced AI architecture, which includes techniques like retrieval augmented generation (RAG). This architecture combines industry-specific and customer-specific data, enhancing the accuracy and personalization of responses.

Using fine-tuned AI to provide hyper-personalized product recommendations

Introducing Bloomreach Clarity

A prime example of our advanced AI application is Bloomreach Clarity, our conversational shopping assistant. Clarity employs an agent-based architecture that dynamically generates workflows based on consumer intent, utilizing system-level reasoning from LLMs. The underlying models, based on Google’s Vertex AI but fine-tuned with our proprietary datasets, are augmented with RAG architecture to incorporate real-time customer data and an accurate, relevant product catalog. This integration results in deeply personalized shopping experiences that are consistent across all channels.

We have designed these innovations with the ecommerce professional in mind. Our systems include built-in app controls for visibility into our AI architecture, such as search ranking diagnostics and customizable settings for tone of voice, third-party ranking signals, and precision relevancy adjustments to align with brand preferences.

To learn more about Bloomreach Clarity, sign up for the waitlist

Bloomreach is at the forefront of reshaping ecommerce personalization through advanced AI solutions. By leveraging sophisticated AI technologies and innovative approaches, Bloomreach is setting new standards in delivering hyper-personalized shopping experiences that meet the evolving expectations of today’s consumers. 


Appendix: Output From GPT-4 Turbo SOTA


I need some help to find a dress for a wedding in Hawaii


Finding the perfect dress for a wedding in Hawaii sounds like a fun task! Here are some tips and ideas to consider when choosing your outfit:

### 1. **Consider the Climate**

   – **Weather:** Hawaii is generally warm and humid. Opt for breathable, lightweight fabrics like cotton, linen, or silk to stay comfortable.

   – **Time of Day:** Evening weddings might be a bit cooler, so you might want a light shawl or wrap.

### 2. **Dress Code**

   – **Formality:** Check the invitation for any dress code. Beach weddings are usually more casual, while a ceremony at a resort could be more formal.

   – **Colors and Patterns:** Tropical prints, floral designs, or bright, solid colors work well in Hawaii. Avoid wearing white or black to a wedding unless specified.

### 3. **Style Suggestions**

   – **Women:**

     – **Maxi Dresses:** Flowy maxi dresses are perfect for a beach setting and come in many patterns and colors.

     – **Cocktail Dresses:** If the event is semi-formal, a cocktail dress in a tropical print or a bright color can be a great choice.

     – **Sundresses:** For a more casual wedding, a simple sundress can be both comfortable and stylish.

   – **Men:**

     – **Linen Suits:** A light linen suit in beige, light gray, or pastel colors can keep you cool and stylish.

     – **Guayabera Shirts:** These are traditional Latin American shirts perfect for tropical climates. Pairing one with linen pants can be both elegant and appropriate for a Hawaiian wedding.

     – **Aloha Shirts:** Also known as Hawaiian shirts, these are acceptable at many local weddings when paired with nice pants or khaki shorts for a more casual look.

### 4. **Footwear**

   – **Women:** Consider sandals or wedges that are easy to walk in on sand if it’s a beach wedding. Avoid high heels that could sink into the sand.

   – **Men:** Loafers or boat shoes can work well. If it’s very casual, nice sandals might also be appropriate.

### 5. **Accessories**

   – **Sun Protection:** A stylish hat and sunglasses can protect you from the sun and add to your outfit.

   – **Jewelry:** Keep it simple to complement the relaxed, tropical vibe. Think wooden beads, shell necklaces, or simple gold or silver pieces.

   – **Bags:** A small clutch or a beach bag in straw or a natural fiber can complete your look.

### 6. **Shopping Tips**

   – **Local Boutiques:** If you’re already in Hawaii, consider visiting local boutiques for unique finds that fit the local style.

   – **Online Retailers:** Websites like ASOS, Nordstrom, or Revolve often have a wide range of suitable dresses and accessories for a destination wedding.

   – **Rental Services:** Consider renting a dress from services like Rent the Runway if you want a high-end look without the commitment of purchasing.

### 7. **Consider the Venue**

   – If the wedding is at a luxurious resort, you might lean towards a more formal look. For a beach or outdoor wedding, lighter and more casual attire is usually better.

By keeping these points in mind, you can choose an outfit that not only looks great but also keeps you comfortable throughout the celebration. Enjoy the wedding and the beautiful setting of Hawaii!


Xun Wang

Chief Technology Officer at Bloomreach

Xun leads Bloomreach’s global engineering and operations team.  He is a veteran engineering executive with over 15 years of experience leading engineering teams. Xun is passionate about technology, complex engineering challenges, and building world-class teams.  In the consumer space, he led the team that created the world’s highest quality cloud gaming platform: Geforce Now. In the enterprise space, he led the team that built Medallia’s cloud platform. Xun holds a Computer Engineering degree from the University of Waterloo.

Table of Contents

Share with Your Community

Recent Posts

Maintain an Edge With These New Posts


Subscribe to get our hot takes on ecommerce topics, trends and innovations delivered to straight your inbox.

Life With Bloomreach

Watch this video to learn what your life could look like when you use Bloomreach.