Enhancing Product Recommendations

Product Discovery is the most critical area for a PM. But, it is largely misunderstood. Teams waste time and energy delivering ideas that do not work and do not drive the expected outcomes. Product Discovery 101: 1. Why do we need Product Discovery? „The first truth is that at least half of your ideas are just not going to work” - Marty Cagan, Inspired I’d argue that Product Management is, at its heart, about managing risk. And for every product, there are 5 risks that can materialize: - Value. Will it create value for the customers? - Usability. Will users figure out how to use it? - Viability. Can our business support it? - Feasibility. Can it be done (technology)? - Ethics. Should we do it? Are there any ethical considerations? What will happen if we throw random ideas into the Product Backlog? This is agile "learning by delivering."This approach results in a waste and rework. So, we would like to understand: - How can we come up with better ideas? - How can we validate those ideas before the implementation? And the answer is Product Discovery. --- 2. When Does it Happen? Continuous Discovery and Continuous Delivery. Those two streams run in parallel: - The goal of Product Discovery is to discover the product to build. - Product Delivery aims to deliver that product to the market. Product Discovery results in a validated Product Backlog. In particular, high-risk assumptions are tested before the implementation. --- 3. Who's Responsible? Some say that the Product Manager decides what to build, and Engineers and designers should focus on how to build it. Have you heard that before? It hurts my ears because Product Discovery is not a task for a single person. Make sure that a Product Designer and at least one Engineer are included. This will help you build a shared understanding and stay open to different perspectives. And if we believe that customers don't know how to solve their problems, why should a Product Manager know it? Product Managers may be tech-literate, but they are not tech experts. --- 4. What’s inside Continuous Product Discovery? There are two groups of activities: - Exploring the Problem Space to understand and define opportunities (problems, needs, desires). My favorite default approach to mapping opportunities is using the Opportunity Solution Tree, as defined by Teresa Torres. - Exploring the Solution Space to explore possible solutions, formulate testable assumptions, and run experiments to prove or disprove those assumptions. --- What's the state of Product Discovery in your company? What is one improvement you can implement starting tomorrow? Hope that helps! --- 🎁 P.S. In my free post, I described Product Discovery in depth. No email, no paywall: https://lnkd.in/dNUB__n3 And here you can download all my infographics: https://lnkd.in/d5bHGj5j

You click "play" on Netflix. In 200 milliseconds, a recommendation engine just processed millions of videos. Most ML engineers know these systems exist. Few understand what's actually running under the hood. I spent the last 6 months building a complete deep-dive series on production recommendation systems — from first principles to the exact architectures running at YouTube, Spotify, and TikTok. Here's the complete roadmap: 🎯 Foundation Layer 1️⃣ RecSys Fundamentals — Content-based, collaborative filtering, and hybrid approaches that power every modern recommender 2️⃣ How Recommendation Systems Learned to Think — The evolution from matrix factorization to transformer-based generative agents ⚡ Retrieval & Ranking Pipeline 3️⃣ The 3-Stage Funnel — How two-tower models, vector databases, and cross-encoders work together at scale 4️⃣ How YouTube Finds Your Next Video in Milliseconds — Two-tower retrieval, in-batch negatives, and the engineering tricks that make it work 5️⃣ Vector Search at Scale — IVF, PQ compression, and making 100M+ vector search actually possible in production 6️⃣ From Candidates to Clicks — The complete ranking stack: from 1,000 candidates to the one item you actually tap 🔧 Production Reality 7️⃣ Solving the Cold Start Problem — Contextual bandits, meta-learning, and LLMs for new users and items (how Spotify, TikTok, YouTube do it) 8️⃣ Beyond Ranking — How diversity, freshness, and business constraints turn a ranked list into a product-ready feed Every post includes: → Production architecture diagrams → Real code examples (PyTorch, Faiss, ranking models) → Case studies from actual systems → The engineering tradeoffs that matter Full series: https://buff.ly/GKEvulv If you're building RecSys or joining a team that does — this is your blueprint.

Researchers from Sejong University have introduced SimDiffRec, a novel approach that leverages diffusion models for contrastive sequential recommendation, addressing critical limitations in existing data augmentation techniques. The Core Innovation: Traditional recommendation systems rely on random augmentation methods like noise injection and item masking, which often damage contextual information. SimDiffRec introduces three key technical breakthroughs: 1. Semantic Similarity-Based Noise Generation Instead of random Gaussian noise, the system computes similarity scores between item embeddings using dot products, then selects the top k semantically similar items and averages their embeddings as structured noise. This preserves contextual integrity while enabling meaningful augmentation. 2. Diffusion Confidence-Based Positioning  The system calculates confidence scores for each position by taking the highest probability from the softmax distribution of predicted item restoration. Augmentation is performed only at positions where the diffusion model shows high reconstruction confidence, ensuring structural consistency. 3. Hard Negative Sampling Integration The approach constructs hard negative samples by selecting items at specific probability ranks (ksample) rather than random selection, enabling the model to learn subtle discriminative features through more challenging contrastive pairs. Technical Architecture: The forward diffusion process uses deterministic noise injection: zt = αtzt−1 + βt · noise, where semantically similar noise replaces random sampling. The reverse process employs standard denoising with trainable rounding to map continuous embeddings back to discrete items. Performance Impact: Extensive evaluation on five benchmark datasets (Beauty, Toys, Sports, Yelp, MovieLens) shows consistent improvements over existing baselines, with particularly strong gains in Hit Ratio and NDCG metrics. The method outperformed both classical approaches and recent diffusion-based recommendation models. Why This Matters: This research tackles the fundamental challenge of data sparsity in sequential recommendation while maintaining semantic consistency - a critical advancement for e-commerce platforms, content streaming services, and personalized recommendation engines operating with limited user interaction data. The combination of semantic-aware augmentation with confidence-based positioning represents a significant step forward in making recommendation systems more robust and contextually aware.

Should product discovery be done differently for Platforms vs. Products? Let's Explore! Product discovery is crucial, but it's not one-size-fits-all. While there's no shortage of materials on consumer product discovery, the unique challenges of platform product discovery often go unaddressed. Here's what makes platform product discovery distinct: 👨💻 𝗙𝗼𝗿 𝗠𝗶𝗱𝗱𝗹𝗲 𝗨𝘀𝗲𝗿𝘀: Whether they're internal teams or external developers, these users need tools that enhance their ability to innovate on your platform. Does your platform equip them to efficiently meet end-user needs? 🚀 𝗙𝗼𝗿 𝗘𝗻𝗱 𝗨𝘀𝗲𝗿𝘀: It’s all about the final experience. Does the platform serve the end goals of the ultimate customers effectively? 🤝 Dual-Layer Discovery: Platform products require a two-pronged approach: • 𝗘𝗻𝗱 𝗨𝘀𝗲𝗿 𝗙𝗼𝗰𝘂𝘀: Dive into understanding their specific goals and needs. What fundamental capabilities should your platform have to serve these needs?    • 𝗠𝗶𝗱𝗱𝗹𝗲 𝗨𝘀𝗲𝗿 𝗙𝗼𝗰𝘂𝘀: Assess whether the platform provides the necessary tools for these key developers. Is it easier and more cost-effective for them to use your platform than starting from scratch? Combining insights from both user types ensures your platform is not just functional but preferred choice. How do you approach product discovery for platform products? Do you find it differs significantly from consumer products? Share your thoughts and strategies in the comments!👇 #ProductDiscovery #ProductManagement #PlatformProductStrategy

Lately, I’ve been thinking about why choosing something to watch often takes longer than actually watching it. You open a streaming app, scroll for a while, watch a few trailers, switch genres, and sometimes end up rewatching something familiar. For years, most recommendation systems have been optimized around predicting what a user is most likely to click or watch next. In large-scale systems, this usually involves generating candidate content using embeddings and retrieval systems, ranking those candidates using machine learning models trained on engagement signals, and then presenting results to the user. But even well-optimized systems struggle with something fundamental: 𝐡𝐮𝐦𝐚𝐧 𝐢𝐧𝐭𝐞𝐧𝐭 𝐜𝐡𝐚𝐧𝐠𝐞𝐬 𝐪𝐮𝐢𝐜𝐤𝐥𝐲. A person’s watch history reflects what they liked in the past, but it does not always capture what they feel like watching in the moment. In reality, discovery often feels less like ranking and more like a conversation. Preferences are 𝐜𝐨𝐧𝐭𝐞𝐱𝐭𝐮𝐚𝐥, 𝐟𝐮𝐳𝐳𝐲 and 𝐞𝐯𝐨𝐥𝐯𝐢𝐧𝐠. This is where 𝐬𝐞𝐪𝐮𝐞𝐧𝐜𝐞-𝐛𝐚𝐬𝐞𝐝 𝐦𝐨𝐝𝐞𝐥𝐢𝐧𝐠 𝐚𝐧𝐝 𝐠𝐞𝐧𝐞𝐫𝐚𝐭𝐢𝐯𝐞 𝐀𝐈 start to change how recommendation systems can be built. Instead of treating every interaction independently, modern approaches can model user behavior as a 𝐬𝐞𝐪𝐮𝐞𝐧𝐜𝐞 𝐨𝐟 𝐚𝐜𝐭𝐢𝐨𝐧𝐬 𝐰𝐢𝐭𝐡𝐢𝐧 𝐚 𝐬𝐞𝐬𝐬𝐢𝐨𝐧. Transformer-based models are particularly well-suited for this because they can learn patterns across sequences of behavior. These systems can begin to understand how preferences shift during discovery rather than simply predicting the next click. In production environments, this often leads to 𝐡𝐲𝐛𝐫𝐢𝐝 𝐫𝐞𝐜𝐨𝐦𝐦𝐞𝐧𝐝𝐚𝐭𝐢𝐨𝐧 𝐚𝐫𝐜𝐡𝐢𝐭𝐞𝐜𝐭𝐮𝐫𝐞𝐬 that combine retrieval systems with generative or reasoning models: 💡Real-time user events feed feature stores that support both offline training and low-latency inference. 💡Embedding-based retrieval systems (e.g., two-tower models + ANN search) reduce millions of items to a few hundred candidates in milliseconds. 💡Ranking models score these candidates based on click probability, watch time, and completion likelihood. 💡Session-aware embeddings and sequence models capture short-term intent shifts during browsing. 💡Re-ranking layers enforce diversity, freshness, and exploration under strict latency constraints. Recommendation systems are gradually moving from 𝐩𝐫𝐞𝐝𝐢𝐜𝐭𝐢𝐧𝐠 𝐛𝐞𝐡𝐚𝐯𝐢𝐨𝐫 to 𝐮𝐧𝐝𝐞𝐫𝐬𝐭𝐚𝐧𝐝𝐢𝐧𝐠 𝐢𝐧𝐭𝐞𝐧𝐭, and from 𝐫𝐚𝐧𝐤𝐢𝐧𝐠 𝐜𝐨𝐧𝐭𝐞𝐧𝐭 to 𝐠𝐮𝐢𝐝𝐢𝐧𝐠 𝐝𝐢𝐬𝐜𝐨𝐯𝐞𝐫𝐲.

🚀 How do you ensure your customers see what they want to see — not just what you want to show? With AI and ML becoming core to ecommerce (both B2B and B2C), product discovery is getting a lot of attention. And rightly so. But here's the truth: most recommendation engines fail not because the models are bad, but because the first two steps were never right. Let me explain. Many product managers (especially in fast-paced orgs) jump into building rec engines with a "let's plug in collaborative filtering and see how it goes" mindset. But without clearly defining what type of recommendation makes sense for your use case — and how it ladders up to a business metric — you're setting yourself up for rework. Here's how I approach it when working with teams: Step 1: Business Understanding: Start with the why before touching the how. ◾ What are you recommending? Products? Content? Users? Services? ◾What does success look like? Higher CTR? More revenue? Better retention? ◾Where will it show up? Homepage, PDP, cart, email, app banner? ◾What constraints exist? Does it need to be real-time? Can it be batched overnight? Without alignment on this, even the most advanced ML model will fall flat. Step 2: Choose the Right Recommendation Type: Now comes the how — but it should be tailored to your product + user journey. ◾Content-based filtering: “You liked this, so you’ll like these similar items.” ◾Collaborative filtering: “Users like you also bought this.” ◾Hybrid models: The best of both worlds — widely used in ecommerce and streaming. ◾Knowledge-based systems: Rule-driven, useful when personalization is constrained (e.g., insurance, banking). Let me make this concrete with a simple example: Imagine you’re building a recommendation module for a first-time visitor on your site who hasn’t logged in. If you apply collaborative filtering, it’ll fail — there’s no past data to compare. But if you use content-based filtering on the item they’re browsing and pair it with trending items, you instantly make the experience better. It’s not about which model is smarter. It’s about which makes sense for the scenario. Let’s be honest — your recommendation engine’s success doesn’t start with machine learning. It starts with product thinking. #AI #ProductManagement #Ecommerce #Personalization #RecommendationEngine #ProductStrategy I write about #artificialintelligence | #technology | #startups | #mentoring | #leadership | #financialindependence   PS: All views are personal Vignesh Kumar

I’ve joined 100s of discovery calls as a buyer—only 1 in 10 was run to help me vs the AE. Most companies treat discovery as just a stage and rely on weak BANT. That’s why AEs get stuck at ‘Level 1’ thinking Disco = Qualification and NEVER win 6-figure deals. Here’s my breakdown of Discovery Levels 1–3 (and the exact steps to finally break free): LEVEL 1 - Thinks: “Discovery is for me—are they worth my time?” - Uses BANT as box ☑ qualification (and to get managers off their backs).  - Asks a few surface questions (“Why are we here?”, “Do you have budget?”). - Disco isn’t leveraged throughout the sales process; it sits in call-1 notes. - If it is used, it’s only for ‘guilt-trip’ moves—“I guess solving X is not a priority?”. - Deals lost/stall/heavily discounted; disco never influenced the buying process. LEVEL 2 - Thinks: “Discovery is for prospects; to help THEM see why they should buy”. - Levels up from BANT to *real* disco using GAP Selling, MEDDIC, SPICED, etc. - Dives deeper than Need—to Impact, Priorities, and Rout Causes of Problems. - Disco is leveraged to tailor demos—give problem-solving tours vs generic tour. - Deals close at OK win rate. Buyers feel understood and see high $$$ problems. LEVEL 3 - Thinks: “Discovery IS the sales process—shaping buying and sales decisions”. - Framework agnostic; digs deep, letters only used for consistency & forecast. - Every call is about shaping the ultimate buying moment—the Business Case. - Every call gets documented; discovery continuously builds their Account Plan. - Doesn’t use a bank of questions; but prepares points to explore (e.g. urgency). - Leverage MAPs to discover & align on how to best run the process—together. - Disco drives unstoppable business case narratives that unlock $6-7 fig budgets. —— Stop treating discovery like just another task. Turn it into your entire selling strategy. Business Case > BANT Qualification. Mutual Plan > “What’s your Timeline?” Discovery as a Service. That’s how you go from chasing… To being a partner that gets trusted with 6-7-figure budgets. Always be discovering.

How do Strategy, OKRs, and Discovery reinforce each other. Strategy without customer evidence becomes abstract and theoretical. Teams struggle to prioritize, creating vague statements like "Make customers happy through a better UI" that justify almost any action. 𝗗𝗶𝘀𝗰𝗼𝘃𝗲𝗿𝘆 provides necessary context to ground strategic choices in reality, while 𝗢𝗞𝗥𝘀 transform these choices into measurable actions, ensuring Strategy is both evidence-based and executable. 𝗢𝗞𝗥𝘀 𝗧𝗿𝗮𝗻𝘀𝗹𝗮𝘁𝗲 𝗦𝘁𝗿𝗮𝘁𝗲𝗴𝘆 𝗶𝗻𝘁𝗼 𝗧𝗮𝗻𝗴𝗶𝗯𝗹𝗲 𝗔𝗰𝘁𝗶𝗼𝗻𝘀: Useful OKRs turn strategic directions into measurable, actionable priorities for teams. They translate broad strategic choices into specific changes in customer behavior and business impacts. 𝗘𝘅𝗮𝗺𝗽𝗹𝗲: Your strategy may focus on “𝘉𝘦𝘤𝘰𝘮𝘦 𝘵𝘩𝘦 𝘭𝘦𝘢𝘥𝘪𝘯𝘨 𝘱𝘳𝘰𝘥𝘶𝘤𝘵𝘪𝘷𝘪𝘵𝘺 𝘴𝘰𝘭𝘶𝘵𝘪𝘰𝘯 𝘧𝘰𝘳 𝘳𝘦𝘮𝘰𝘵𝘦-𝘧𝘪𝘳𝘴𝘵 𝘵𝘦𝘢𝘮𝘴 𝘵𝘩𝘳𝘰𝘶𝘨𝘩 𝘤𝘶𝘴𝘵𝘰𝘮𝘪𝘻𝘢𝘵𝘪𝘰𝘯,” but OKRs specify exactly what measurable change (e.g., increasing weekly active usage of collaboration features from specific types of customers) will demonstrate tangible progress. OKRs without strategic boundaries become generic, ineffective, and misaligned. Product Strategy provides the specific choices you want to measure, while Product Discovery ensures your OKRs are grounded in reliable evidence about customer problems. 𝗗𝗶𝘀𝗰𝗼𝘃𝗲𝗿𝘆 𝗙𝘂𝗲𝗹𝘀 𝗠𝗲𝗮𝗻𝗶𝗻𝗴𝗳𝘂𝗹 𝗞𝗲𝘆 𝗥𝗲𝘀𝘂𝗹𝘁𝘀: Discovery identifies user needs and customer behaviors worth targeting through OKRs. When Discovery reveals specific user pains, OKRs can turn the solving of them into measurable user Outcomes. 𝗘𝘅𝗮𝗺𝗽𝗹𝗲: If Discovery highlights onboarding friction reducing customer retention, OKRs can aim at measuring improvements in onboarding flow adoption rates, dramatically enhancing the contribution to business goals. Discovery without direction leads to endless exploration. Teams can feel overwhelmed, believing they must investigate every problem. 𝗢𝗞𝗥𝘀 𝗮𝗻𝗱 𝗣𝗿𝗼𝗱𝘂𝗰𝘁 𝗦𝘁𝗿𝗮𝘁𝗲𝗴𝘆 provide necessary structure, narrowing Discovery’s focus. 𝗦𝘁𝗿𝗮𝘁𝗲𝗴𝘆 𝗦𝗲𝘁𝘀 𝗕𝗼𝘂𝗻𝗱𝗮𝗿𝗶𝗲𝘀 𝗳𝗼𝗿 𝗣𝗿𝗼𝗱𝘂𝗰𝘁 𝗗𝗶𝘀𝗰𝗼𝘃𝗲𝗿𝘆: Product Strategy clarifies relevant market opportunities, preventing Discovery from drifting into misaligned or superficial areas. 𝗘𝘅𝗮𝗺𝗽𝗹𝗲: If Strategy emphasizes expanding existing enterprise accounts, Discovery clearly prioritizes problems current customers face instead of exploring new market segments. Visualization by the fabulous Julia Steier!

Your Product Managers are talking to customers. So why isn’t your product getting better? A few years ago, I was on a team where our boss had a rule: 🗣️ “Everyone must talk to at least one customer each week.” So we did. Calls were scheduled. Conversations happened. Boxes were checked. But nothing changed. No real insights. No real impact. Because talking to customers isn’t the goal. Learning the right things is. When discovery lacks purpose, it leads to wasted effort, misaligned strategy, and poor business decisions: ❌ Features get built that no one actually needs. ❌ Roadmaps get shaped by the loudest voices, not the right customers. ❌ Teams collect insights… but fail to act on them. How Do You Fix It? ✅ Talk to the Right People Not every customer insight is useful. Prioritize: -> Decision-makers AND end-users – You need both perspectives. -> Customers who represent your core market – Not just the loudest complainers. -> Direct conversations – Avoid proxy insights that create blind spots. 👉 Actionable Step: Before each interview, ask: “Is this customer representative of the next 100 we want to win?” If not, rethink who you’re talking to. ✅ Ask the Right Questions A great question challenges assumptions. A bad one reinforces them. -> Stop asking: “Would you use this?” -> Start asking: “How do you solve this today?” -> Show AI prototypes and iterate in real-time – Faster than long discovery cycles. -> If shipping something is faster than researching it—just build it. 👉 Actionable Step: Replace one of your upcoming interview questions with: “What workarounds have you created to solve this problem?” This reveals real pain points. ✅ Don’t Let Insights Die in a Doc Discovery isn’t about collecting insights. It’s about acting on them. -> Validate across multiple customers before making decisions. -> Share findings with your team—don’t keep them locked in Notion. -> Close the loop—show customers how their feedback shaped the product. 👉 Actionable Step: Every two weeks, review customer insights with your team to decipher key patterns and identify what changes should be applied. If there’s no clear action, you’re just collecting data—not driving change. Final Thought Great discovery doesn’t just inform product decisions—it shapes business strategy. Done right, it helps teams build what matters, align with real customer needs, and drive meaningful outcomes. 👉 Be honest—are your customer conversations actually making a difference? If not, what’s missing? -- 👋 I'm Ron Yang, a product leader and advisor. Follow me for insights on product leadership + strategy.

It's alive! My "bilingual" recsys x llm can converse in English AND item IDs! Here it recommends an item and explains why I might like it. <|sid_start|> ... <|sid_end|> is the item ID. The model input and output has the item ID (and English), and the timestamped logs map the item ID back to the human-readable item name. Wait, can't a regular LLM do this already? Where I give it an item and ask it to recommend more items? Yes, but it's doing that based on internet knowledge, not how customers actually behavior (e.g., search, click, purchase, etc.) In this case, I'm finetuning an LLM (Qwen3-8B) on transaction and reviews data (https://lnkd.in/g_4ki38K) to inject user behavioral patterns, creating an LLM-recsys hybrid that can (i) recommend based on user behavior, while (ii) allowing us to shape and explain recommendations via natural language.