Dynamic Product Recommendations

Exciting breakthrough in AI recommendation systems! A team of researchers from Meta, UMN, NCSU, and UNC Chapel Hill have developed an innovative framework that significantly improves both efficiency and accuracy of LLM-based recommender systems. The framework introduces two key innovations: >> GCN-Retriever Their solution uses Graph Convolutional Networks (GCNs) to efficiently identify similar users by analyzing interaction patterns in user-item graphs. This replaces traditional LLM-based retrieval methods, dramatically reducing computational overhead while maintaining recommendation quality. >> Multi-Head Early Exit Architecture  The system implements a novel early exit strategy with multiple prediction heads at different layers. By monitoring prediction confidence in real-time, the model can terminate processing early when sufficient confidence is reached, significantly improving inference speed. >> Performance Highlights - Achieved 96.37 AUC on Amazon Beauty dataset - Up to 4.96x improvement in requests per second - Maintains or improves accuracy while reducing computation time - Successfully handles both sparse and dense interaction data The framework addresses two critical bottlenecks in current LLM recommender systems: retrieval delays and inference slowdown. By combining GCN-based retrieval with dynamic early exit strategies, the system delivers faster, more accurate recommendations at scale. This work represents a significant step forward in making LLM-based recommendation systems practical for real-world commercial applications. The framework's ability to balance efficiency and accuracy while maintaining robust performance across different datasets demonstrates its potential for wide-scale adoption.

Want to know how companies like H&M recommend the perfect items? The 𝗧𝘄𝗼-𝗧𝗼𝘄𝗲𝗿 𝗡𝗲𝘁𝘄𝗼𝗿𝗸 is behind it all - here's how it works: It embeds users and items into a shared vector space to enable fast and efficient retrieval of recommendations. The model consists of two parallel networks, known as towers: 1. 𝗤𝘂𝗲𝗿𝘆 𝗧𝗼𝘄𝗲𝗿 (𝗖𝘂𝘀𝘁𝗼𝗺𝗲𝗿 𝗤𝘂𝗲𝗿𝘆 𝗘𝗻𝗰𝗼𝗱𝗲𝗿) This tower encodes user features such as customer_id, age, and other attributes to understand user preferences. 2. 𝗖𝗮𝗻𝗱𝗶𝗱𝗮𝘁𝗲 𝗧𝗼𝘄𝗲𝗿 (𝗜𝘁𝗲𝗺 𝗘𝗻𝗰𝗼𝗱𝗲𝗿): This tower encodes item-specific features like article_id, garment_group_name, and other product attributes. Both towers work independently but generate embeddings that live in the same low-dimensional space. This shared space allows the model to compare users and items efficiently and determine relevant matches. Each tower processes its respective inputs by: - 𝗙𝗲𝗮𝘁𝘂𝗿𝗲 𝗘𝗻𝗰𝗼𝗱𝗶𝗻𝗴 𝗮𝗻𝗱 𝗙𝘂𝘀𝗶𝗼𝗻: User and item features (both categorical and numerical) are converted into dense embeddings. For example, customer_id and article_id are mapped to vectors leveraging an *EmbeddingLayer*, while categorical features like garment_group_name use a one-hot encoding layer. - 𝗡𝗲𝘂𝗿𝗮𝗹 𝗡𝗲𝘁𝘄𝗼𝗿𝗸 𝗥𝗲𝗳𝗶𝗻𝗲𝗺𝗲𝗻𝘁: A feedforward neural network (FNN) with multiple dense layers refines these embeddings, reducing them to a low-dimensional space. This compact embedding helps prevent overfitting and ensures more generalized recommendations. 𝘉𝘶𝘵 𝘸𝘩𝘺 𝘥𝘰 𝘭𝘰𝘸-𝘥𝘪𝘮𝘦𝘯𝘴𝘪𝘰𝘯𝘢𝘭 𝘦𝘮𝘣𝘦𝘥𝘥𝘪𝘯𝘨𝘴 𝘮𝘢𝘵𝘵𝘦𝘳? Because without limiting the dimensions, the model might overfit by memorizing past purchases. (Meaning users will receive repetitive recommendations of items users already own) By balancing collaborative filtering and content-based filtering, the Two-Tower Model provides both personalized and diverse recommendations. The more features added, the more the model leans toward content-based filtering. But to experiment, you can add more features to both towers, measure the results, compare them and repeat. Why? The more user and item features are included, the more the model moves towards content-based filtering. (resulting in more diverse and dynamic recommendations) P.S. Have you built a personalized recommender with the Two-Tower network before?

Something clicked for me this week… I was exploring how AI is changing how we shop! And I realised we’re entering a completely NEW phase of product discovery 😅 Not through Google. ❌ Not through influencers.❌ But through conversational AI.✅ People aren’t just searching anymore. They’re asking. - “What’s the best shampoo for curly hair?” - “Recommend a vegan face wash under £20.” - “What are some eco-friendly travel kits?” And here’s the part that stopped me: And if your product isn’t showing up in ChatGPT or Gemini? You’re not just missing sales. You’re invisible. I have worked in marketing for decades! I’ve seen trends come and go. But this one feels different. It’s not a “maybe.” It’s happening. Quietly. Rapidly. So I’ve started looking into: - How D2C brands can register their product data into AI models - What tools like ChatGPT actually “see” when people ask for recommendations -And how we, as marketers, can show up where the next-gen consumer is asking This shift reminds me of when brands hesitated to build for mobile. Or ignored TikTok at first. We all know how that played out. 😉 This new wave of AI-first discovery needs a different playbook. 👉🏻👉🏻 So here’s a step by step guide for D2C brands to show up in AI recommendations: 1. Structure your product data using Schema.org / JSON-LD 2. List on Google Shopping, Amazon, and Shopify (these feed AI models) 3. Integrate with ChatGPT plugins like Klarna, Shop.app or use your own GPT 4. Write product content that answers real questions (not just keywords) 5. Test visibility by asking ChatGPT or Gemini to recommend your product category 6. Keep optimizing based on what AI can (and can’t) find If you’re building a brand today especially in D2C, it’s time to think beyond search. The new product shelf is a ‘chat box’ :) If you are a D2C founder or a consumer brand and need more insights on to leverage AI first shopping, comment ‘Product discovery’ and I will share the playbook with you. #AI #D2C #Productdiscovery #Search #Shoppingtrends #consumerbehaviour

In our journey of building personalized recommendations, we often debate when models should run in real-time vs. batch processing. It completely depends on use case, scalability, and latency that is acceptable. Let me try to simplify it so that you can explain it better to your management - 1) Real-Time Models – When Instant Personalization is Key. This flow is used when recommendations must be generated instantly based on a user’s current actions. Example Use Cases: "You May Also Like" – A user clicks on a product, and recommendations are generated dynamically. Personalized Home Page – When a user logs in, their recommendations are fetched in real time. Dynamic Offers – Based on recent user behavior, a discount or coupon is displayed immediately. This is how it can be implemented if using Amazon Web Services (AWS): 🔹 User Action → A user visits a webpage or clicks on a product. 🔹 API Gateway + Lambda → Triggers an API call to fetch recommendations. 🔹 Model Prediction (SageMaker Endpoint) → If no cached results exist, the model generates new recommendations. 🔹 DynamoDB / Redis Cache → First checks for recent recommendations to reduce latency. 🔹 Response to Frontend → Results are returned and displayed instantly. 2) Batch Processing – Precomputed Recommendations This approach is used when personalization can be precomputed, reducing the need for real-time execution. Example Use Cases: "Your Favorites" (Rule-Based Personalization) – If a user buys from X retailers frequently, precompute recommendations daily. Periodic Email / Push Notifications – Personalized product suggestions for email marketing campaigns. Homepage Personalization (Static User Preferences) – Daily updates to improve page load speed. This is how it can be implemented: 🔹 Daily / Weekly Training Jobs (Glue, SageMaker, EMR) → or you can use dedicated EC2 & Jenkins to process large amounts of data and update recommendations. 🔹 Updated Recommendations Stored (DynamoDB, Redis) 🔹 Precomputed Recommendations Served via API / CloudFront So, if recommendation changes dynamically basis user session, use real time. For predictable updates use batch. Infact, one can use hybrid approach also - Cache precomputed results and fall back on real-time inference when needed. #recommendation #n=1personalisation #datascience #data

I boosted click-through rates by 45% with dynamic product recommendations. A DTC brand struggled with low engagement and click-through rates in their email campaigns. Their emails lacked relevance and personalization, which led to subscribers tuning out. The Problem: Generic product recommendations that didn’t resonate with individual customers. Low engagement and poor click-through rates. What I Did: 1. Implemented Dynamic Content: Introduced product recommendations based on customers’ browsing and purchase history. 2. Segmented Emails: Sent tailored campaigns to specific customer segments based on their interests and past behavior. 3. Tested Personalization Approaches: A/B tested various personalized elements such as product recommendations, email timing, and subject lines. The Result: 45% increase in click-through rates. Higher engagement and stronger customer relationships. Improved conversion rates from emails. Personalized product recommendations are key to driving engagement and conversions. Are you personalizing your emails effectively? #EmailMarketing #DTC #Personalization #CaseStudy #ClickThroughRates #DynamicContent

Amazon Launches “Product Selector” Quiz Tool to Boost Storefront Conversion Amazon has introduced a new Product Selector feature for Brand Stores — a quiz-style interactive tool that helps shoppers discover products through a guided Q&A flow. Brands can now create up to four simple, image-supported questions that filter down to personalized product recommendations based on customer preferences. Why It Matters For brands with deep catalogs or nuanced product lines, this is a low-lift way to streamline discovery and boost conversion. It mimics the experience of guided selling in retail — narrowing choices through personalized questions that make decision-making faster and more engaging. Key Features 🔹Quiz-style interface with up to 4 questions and 6 answers each 🔹Visual answer options using high-quality lifestyle or product images 🔹Dynamic filtering to ensure product matches only show if they align with selected responses 🔹Custom branding tools: fonts, colors, button styles, and intro messaging 🔹Fallback options if no match is found (up to 4 alternative products or redirect to a store page) Use Cases for Sellers 🔹Prime Day prep: Direct traffic to budget-friendly or best-match deals 🔹Gift guides: Recommend by occasion, recipient, or preferences 🔹Category filtering: Help customers navigate large or complex catalogs 🔹Routine product discovery: Recommend products based on needs or usage style How to Build It 🔹Select up to 50 featured products 🔹Map answer choices to relevant SKUs 🔹Use clear, benefit-driven language and avoid overlap across answers 🔹Optionally include product images with clean composition and high contrast 🔹Preview and test in Brand Store Builder before publishing This tool is now live inside the Brand Store builder under the “Add new section” menu. Brands looking to enhance product discovery and reduce friction during browsing should start experimenting with Product Selector ahead of major sales events. ⬇️ Tap the link to read more about this update! https://lnkd.in/dupMQDft