Increasing Trust in E-commerce Recommendations

Zomato faced a big problem: How can we turn app browsers into loyal customers? The goal was clear, improve the user experience with personalized restaurant suggestions. But there were a few challenges too: 🔴 Understanding user preferences from massive data. 🔴 Combining multiple data sources for meaningful insights. 🔴 Developing accurate recommendation algorithms. 🔴 Processing data in real time to keep users engaged. 🔴 Building trust in the recommendations to ensure they felt helpful, not intrusive. To tackle this, Zomato used a structured approach: 🟢 Data Collection and Cleaning - They collected user behavior data (searches, clicks, abandoned carts). - They analyzed restaurant details (cuisine types, delivery times, ratings). - Past orders were also analyzed for trends. 🟢 User Segmentation - Users were grouped based on age, location, past orders, and browsing habits. - This helped them identify patterns and preferences. 🟢 Developing the Recommendation System - Combined collaborative filtering (what others like you prefer) and content-based filtering (what matches your past orders). - Fine-tuned algorithms with ongoing testing for better accuracy. 🟢 Implementation and Testing - They rolled out the recommendations and tested them through A/B experiments. - Adjusted based on user feedback and data performance. 🟢 Continuous Improvement - Introduced feedback loops for real-time adjustments. - Regular updates ensured the system stayed relevant to evolving user needs. And, the impact was impressive: ⬆️ 35% more time spent on the app by users receiving personalized suggestions. ⬆️ 28% higher click-through rates, showing better engagement. ⬆️ 22% increase in orders per user per month due to tailored suggestions. ⬆️ 18% boost in retention rates, turning occasional users into loyal customers. ⬆️ 12% higher average order value, leading to revenue growth. ⬆️ 15% jump in monthly revenue, proving personalization works! I see this as the perfect example of using data to deepen customer relationships. It's not just about the tech—it’s about understanding people and making their experience smoother and more personal. 📊 Data is the secret to building trust and loyalty. What do you think? Can other industries learn from Zomato’s success? How can personalization improve your industry? #zomato #deepindergoyal

Exciting breakthrough in E-commerce Recommendation Systems! Just read a fascinating paper from eBay's research team on "LLM-PKG" - a novel approach that combines Large Language Models with Product Knowledge Graphs for explainable recommendations. Here's what makes it groundbreaking: >> Technical Architecture - The system uses a two-module approach: offline construction and online serving - LLM generates initial product relationships and rationales, which are transformed into RDF triplets (Subject, Predicate, Object) to build the knowledge graph - The system employs rigorous validation using LLM-based scoring (1-10 scale) to evaluate recommendation quality and prune low-quality nodes (score < 6) >> Under the Hood - Product mapping uses BERT embeddings and KNN indexing for semantic matching between LLM recommendations and actual inventory - The system caches graph triplets in key-value databases for lightning-fast retrieval during online serving - Supports both item-centric and user-centric recommendation scenarios >> Real-World Impact The A/B testing results are impressive: - 5.19% increase in clicks - 7.59% boost in transactions - 8.56% growth in Gross Merchandise Bought - 10.84% increase in ad revenue This is a game-changer for e-commerce platforms looking to provide transparent, explainable recommendations while maintaining high performance at scale.

I audited how AI recommends products. The same 8 ranking factors keep showing up. Most ecommerce brands haven't optimized for a single one. Here's what actually determines whether ChatGPT, Perplexity, or Google AI recommends your product, or your competitor's. Factor 1: Intent Alignment → AI prioritizes content that directly resolves specific, conversational buyer questions, not pages built around exact-match keyword targets. Factor 2: Entity Strength → AI models verify brand credibility by cross-referencing your company data across Wikidata, directories, and social platforms.  → Fragmented or conflicting data causes AI to lose confidence and skip recommending your products. Factor 3: E-E-A-T Signals → Experience, Expertise, Authority, and Trust are the foundational filters AI uses to select reliable sources.  → If your product page doesn't show WHO stands behind it, AI skips you. Factor 4: Structured Data & Schema ← start here → Comprehensive JSON-LD markup acts as a direct API for AI.  → Full Product, Review, and Organization schema allows LLMs to instantly extract exact pricing and availability. Factor 5: Extractable Structure → AI engines favor content formats they can easily scrape, parse, and present in synthesized summaries.  → Q&A sections, comparison tables, and quick-summary blocks improve your machine-readability. Factor 6: Factual, Conversational Tone → AI engines are trained to discard subjective marketing fluff in favor of objective, verifiable data points.  → "20% L-ascorbic acid, clinically tested for dry skin" gets cited. "Revolutionary, life-changing serum" gets ignored. Factor 7: Topical Authority → Publishing isolated blog posts fails to build the comprehensive knowledge graphs that AI models look for.  → Centralized hub pages linked to 8–12 hyper-specific supporting articles prove you are the category expert. Factor 8: Cross-Platform Consistency → AI evaluates your brand as a holistic entity across the entire internet, not just the pages on your primary domain.  → Synchronizing product specs, pricing, and messaging across Amazon and Google Shopping solidifies entity recognition. The priority order: → Schema (+94% relevance boost) - do this week → E-E-A-T (2.1× citation lift) - do this month → Freshness (2025–2026 data wins) - make it ongoing Then test: query your products in ChatGPT and Perplexity. Track what gets cited. Target: 2–5x AI visibility improvement. ♻️ Repost so your ecom network sees this before their competitors do.

The 2024 PwC "Voice of the Consumer Report" delivers some sage insights into what today’s Australian consumers want from B2C brands. 👀 🇦🇺 🚧 Spoiler alert: personalised, trusted, secure, and seamless experiences powered by AI are at the top of the list 👇🔎 The Power of Omnichannel: 🛍️💻🤖 While 73% of shoppers still value in-store experiences, 86% want seamless integration between online and offline shopping channels. AI is playing a central role in delivering these personalised experiences, but transparency about how AI-driven data is collected and used is key. Delight comes when customers feel secure while enjoying AI-enhanced, seamless journeys. Value + Delightful AI-Driven Experiences: 💸🔐 67% of consumers prioritise value for money, but today’s value proposition goes beyond price—it’s about personalised, AI-powered, and secure experiences. B2C brands must balance competitive pricing with responsible data protection to create truly delightful customer journeys. Sustainability with Integrity: 🌱 🌏 52% of consumers are willing to pay more for sustainable products, but they expect brands to align their sustainability practices with transparency and integrity. Consumers want brands to back up their green claims with real action. Rising Demand for Privacy: 🔒 🤝 🛡️ A whopping 85% of consumers are concerned about how their personal data is collected and used. With AI driving more customer interactions, privacy-enhancing technologies (PETs) are essential to maintaining consumer trust. Brands that prioritise privacy will gain loyal customers. Expectations for Personalisation:🤖🔏 59% of consumers expect personalised product recommendations, but they also want brands to respect their privacy. AI can deliver tailored experiences, but it’s crucial to use ethical AI that align with consumer expectations for data security. 💡 3 Recommendations for B2C Brands: 1) Combine AI with Privacy-Enhancing Technologies (PETs): 🤖 Using AI responsibly alongside PETs helps brands deliver personalised experiences while safeguarding customer data. This approach builds trust and loyalty. 2) Be Transparent About AI Use: 📝🔍 Consumers are increasingly aware of AI’s role in their shopping experiences. Make sure your brand communicates clearly about how AI is being used and how data is handled. Transparency equals trust. 3) Ethical AI for Delightful Personalisation: 🤖✨ Invest in ethical AI to enhance customer experiences through tailored recommendations and personalised shopping journeys while maintaining privacy. AI can delight, but only when deployed responsibly. 👀 The future of B2C brands lies in delivering AI-driven, delightful experiences that seamlessly blend online and offline channels while ensuring privacy, trust, and sustainability. Brands that lead with ethical AI and transparency will be the big winners in consumer confidence and loyalty. 💪 #Omnichannel #PrivacyMatters #DataSecurity #Sustainability #DelightfulExperiences

We process 10 billion ecommerce transactions per year at Rokt. That volume gives us a window into something fundamental about consumer behavior: people are drowning in options. Research shows ~60% of consumers now expect personalization, and ~35% of Amazon's purchases come from recommendations. These are signals that choice overload is real and measurable. I think of it as a privilege to accompany shoppers through their decisions and learn how to improve the next experience. But that privilege comes with responsibility: transparency about what data we collect and how we use it, a fair value exchange where the benefit outweighs the privacy cost, and reversibility - making it easy to delete data immediately if requested. The smartwatch analogy is useful here. People welcome biometric data collection when the value is clear (health insights, fitness tracking, sleep patterns). But if that same data gets resold to insurers without consent, trust collapses instantly. The technology hasn't changed - the terms of the exchange have. Trust isn't built through marketing language or privacy policies written by lawyers. It's built through closed networks, no data resale, and consistently delivering value that outweighs the privacy trade-off. At scale, when you have volume and breadth of data, assistance feels genuinely useful rather than intrusive advertising. Jurisdictional variance matters too. EU consumers expect different levels of control than US consumers. You can't design for trust with a one-size-fits-all approach. The paradox of choice is real. Our mission is solving it by recommending what consumers truly need in-moment, but we only earn that role through sustained trustworthiness, not clever positioning.

Trust means a confident belief that something will hold up. Not persuasion. Belief. But most ecommerce operators treat trust like a persuasion problem. Stack the page with badges, a wall of reviews, a clip of someone using the product, and the shopper on the fence finally tips over into buying. The whole theory is that the right proof element closes the gap. It doesn't. Think about the last time you'd already cooled on something in a store and the salesperson kept pushing anyway. It didn't move you an inch closer to buying, and it probably soured how you felt about the place. That's a used car lot. Don't turn your product page into one. Shoppers are smarter than we give them credit for, and we keep insulting their intelligence by pretending otherwise. By the time someone lands on your product page, they've done most of the deciding. The media buying and the creative already did the persuading. They clicked because they want the thing. So a trust element isn't there to convince anyone. It's there to keep you from eroding a decision the shopper already made in their head. Every element on the page either protects that confidence or chips away at it. A confusing shipping line, a review that raises a question you never answer, a guarantee buried three scrolls down. None of those sell anybody. They just hand a ready buyer a reason to pause. Once you see trust as validation instead of persuasion, the page gets easier to build. What goes on it, and where each piece sits, stops being a guess about what might convince someone. It becomes a simpler question: what keeps a sold shopper feeling sold. You're not arguing anyone into the purchase. They already got there. Your job is to not talk them out of it.

When I worked in eCommerce, product recommendation quizzes were a somewhat secret weapon—especially during the holiday gifting season. They were like DIY personalization: a way to guide shoppers to certain products without requiring sophisticated tech (and 71% of consumers expect this). They helped them buy more, buy faster, and with less guesswork. They helped us hit sales goals while collecting valuable information. Fast forward to now, and I’m marketing that same solution to my old self. And I’m still just as bullish about it because I’ve seen the power of it from all sides: ♦ 𝐀𝐬 𝐚 𝐦𝐚𝐫𝐤𝐞𝐭𝐞𝐫 𝐮𝐬𝐢𝐧𝐠 𝐭𝐡𝐞𝐦, they drive sales and improve the user experience. ♦ 𝐀𝐬 𝐚 𝐦𝐚𝐫𝐤𝐞𝐭𝐞𝐫 𝐧𝐨𝐰 𝐦𝐚𝐫𝐤𝐞𝐭𝐢𝐧𝐠 𝐭𝐡𝐞𝐦, they help brands create seemingly personalized experiences that build trust, drive results, and collect data. ♦ 𝐀𝐬 𝐚 𝐬𝐡𝐨𝐩𝐩𝐞𝐫 𝐭𝐚𝐤𝐢𝐧𝐠 𝐭𝐡𝐞𝐦, they just work. I give something, I get something. I trust the output because I provided the input—not AI, not cookies. And that makes me more inclined to *add to cart*. Not many things in this business are a win-win, but I can confidently say that product recommendation quizzes often are. They are simple. They are effective. They benefit both the buyer and the seller. What's not to love? Anyway, we made a whole eBook about it—link in comments.

If you're serious about GEO / AEO, you need to take "Trust Surface Area" seriously. Try this. Trust Surface Area is the total footprint of credible, consistent signals about your brand that an AI model can find, connect, and cite when answering real customer questions. In plain English, Trust Surface Area is the reason some brands get cited and recommended over and over, even when competitors have similar products. It is not a single metric. It is the combined surface of: • Verified facts about who you are • Third-party validation that you do what you claim • Repeated, consistent positioning across multiple independent sources • Clear ties between your brand entity, your category, and the problems you solve In classic SEO, you could sometimes “rank” with good on-page work and a backlink profile. In AI search, the model has to recommend. That means AI search is like a judge that is thinking about recommending you. Recommendation requires confidence. Confidence requires evidence. • For a large regional tutoring company, over 70% of their citations can come from their website, which means their website has the largest trust surface area.    • For a leading beverage brand, I've seen over 80% of citations come from product pages and third-party roundup articles.    • For B2B SaaS in the governance niche, I've seen third-party review sites comprise over 60% of citations, Different industries have different trust surface areas. Here's my advice to you if you're starting from scratch. 1. Determine the evidence an AI needs to confidently recommend you, and where does it expect to find that evidence? 2. Standardize your identity signals (your name, slogans, product positioning, etc.) on your home page, product pages, LinkedIn, YouTube. 3. Turn offline trust into public proof. Publish case studies that include context: industry, constraints, timeline, before and after state. Try to name names. 4. Create structured and in-depth mid and bottom funnel content. If you're an e-commerce, make sure your product pages are rich with product info, who it's for, FAQs, etc. For B2B SaaS, include use cases. What do you think?

Incredible Shoptalk conversation this week with Jason Del Rey from The Aisle, Ashye Marcus from Stripe and Kristina Elkhazin from Klarna on agentic commerce and what brands need to do to get their products recommended in this fast growing discovery channel. Carla DeSantis from PwC said it well: Brand sentiment and awareness used to be the center of the show and the vehicle through which products were brought to market. Now, we’re seeing that flip with the role of brand sentiment being disrupted and product-level data and attributes becoming the most important (yet previously neglected) component of getting your product recommended in AI. Key takeaways: 1/ The 3 problems brands & retailers face in discovery: - AI can’t see your products at all (unreadable / non-existent catalog feed) - AI can see your products but doesn’t recommend you (low-quality, low trust attribute data) - AI can see you and recommends you, but in the wrong contexts & conversations (wrong attribute data) 2/ What is causing this: - Existing catalog/merchant schemas are basic and generic. - Different product types (e.g., face wash vs socks vs peanut butter) require very different attribute data. - LLM-driven discovery needs deeper category and intent-specific product data than what current feeds provide. 3/ The solution: - Know what category contexts you’re trying to win in (what products are you being compared to and considered against) - Infuse that category specific data into your product schema (introduce the information needed for products to show up in the conversations you care about) - Get that data out consistently, everywhere (put it in all of the channels you sell across the internet) 4/ Lastly, this doesn’t just apply to brands. Retailers need to make sure that when a product is recommended, their store is the place the consumer is being sent to buy it. This requires the same “golden catalog” data. If your feed is full of garbage, AI will ignore you as well. 

Buyers aren’t just searching for products. They’re searching for proof that you understand their world. And if your search bar can’t keep up? They’re already losing trust. Too many B2B ecommerce sites treat keywords like “filter” as universal. But context changes everything. This is why I've added "contextualization" as the fifth of the "5 Cs of Product Content". An HVAC tech, an auto mechanic, and a wastewater engineer will all type "filter." But what they need? Completely different. So let’s talk about context. Because that’s where B2B #ecomauthenticity lives. There are two kinds: 🔹 Long-term context The stuff that doesn’t change often—industry, facility type, buyer role. This should already be shaping the search results. It’s table stakes. 🔹 Short-term context This is where most B2B sellers fall short. A special job, a spec change, a short-term substitution—they’re hidden from most sellers, so they are harder to account for in the digital experience. Historically - they live in your salespeople’s heads. Here’s a real-world example: A cutting tools distributor finds out one of their aerospace customers is moving from jobs focused on aluminum components to those made of carbon fiber. That shift changes the game: Different edge geometry. Different coatings. Entirely different SKUs. If your ecommerce site keeps recommending the same old tools? You’ve just told your customer: we’re not paying attention. But if that insight flows from sales into your CRM and powers your search & merch logic? This is why having a system that makes it easy for your outside sales reps to track, report, and log any new short-term contexts is so critical in the new model of #b2becommerce. Because if you can do that... Now you’re showing relevance. Now you're showing expertise. Now you're building trust—before the buyer ever clicks “Add to Cart.” This is what authenticity looks like online: Not just having products on your site. But knowing which ones matter, to this buyer, right now. That’s what younger B2B buyers are looking for. Not claims. Not slogans. Signals.