Analyzing Purchase Patterns in E-commerce

🛒 How Basket Analysis Can Drive eCommerce Growth: A Bangladeshi Scenario As eCommerce continues to grow rapidly in Bangladesh, businesses are dealing with more and more customer data. One of the most valuable and often overlooked ways to make sense of that data is through basket analysis. Whether you’re working at a platform like Daraz, Chaldal, Pickaboo, or even running your own online shop, basket analysis can help uncover what products people are buying together. These insights can help you make smarter decisions when it comes to marketing, product placement, bundling, and personalized offers. 🔍 What is Basket Analysis? Basket analysis (also known as market basket analysis) is a method used to find associations between products based on customer purchase history. For example: - What do people usually buy with rice? - Are customers who buy smartphones also buying covers or screen protectors? - Are snack items more popular during weekends? By identifying patterns like these, eCommerce platforms can: - Increase average order value - Run more effective cross-sell campaigns - Deliver personalized recommendations - Make better inventory decisions 🧺 Real-Life Example: A Case Based on Chaldal While analyzing data from Chaldal, one of Bangladesh’s largest online grocery platforms, we noticed something interesting. Many customers in areas like Dhanmondi and Mirpur were buying instant noodles and tomato ketchup together, especially during the evening. This pattern suggested a common need: quick dinner solutions, likely for students or working professionals. Based on this insight, we tested a few simple strategies: - Introduced a combo offer with noodles and ketchup - Showed both products in the “Frequently Bought Together” section - Ran targeted push notifications in the evening with a message like “Need a quick dinner? Grab our Noodles + Ketchup combo now!” The early results were promising: - Better product visibility - More engagement during evening hours - A small bump in basket size for repeat users We’re still monitoring the data, but it’s a great example of how even small insights can be turned into smart decisions. 💡 Final Thoughts You don’t need AI or complex tools to start using basket analysis. A simple SQL query or spreadsheet analysis can help you uncover product relationships that lead to real business value. #eCommerce #BasketAnalysis #DataAnalytics #DigitalBangladesh #CustomerInsights #BusinessGrowth #SQLforBusiness #OnlineGrocery #MarketingStrategy #StartupBangladesh

Half our marketing budget targeted women 25-34. Our highest converting audience? Men 45-65 buying gifts. Discovered this by accident when analyzing order patterns from last Diwali season. These gift-buying men were completely invisible in our targeting strategy. Weird pattern we noticed: ⤵︎ They never used discount codes ⤵︎ Always chose express shipping ⤵︎ Bought our highest-priced items ⤵︎ Had near-zero return rates Our acquisition cost for this segment was 4X lower while average order value was 3.2X higher. Instead of ignoring this insight, we rebuilt our entire holiday strategy around it: ↗︎ Created "gift concierge" landing pages with curated selections ↗︎ Added gift wrapping and personalized message options ↗︎ Developed email sequences specifically for gift occasions ↗︎ Built lookalike audiences based on this high-value segment These changes increased our holiday revenue by 142% year-over-year while reducing marketing spend by 17%. The most profitable audience segments rarely match your brand's imagined customer avatar. Data reveals who's actually buying, not who you think should be buying. What hidden audience segments are you overlooking?

Take a guess: Which two products cross-sell best with Coke at the major US pharmacy chains? Keep reading and learn how I did the analysis. I worked on trade spend optimization at Coca-Cola with a simple goal: generate more revenue from the same shelf space. Here's exactly how I did it. The most impactful analysis I worked on at Coca-Cola had nothing to do with dashboards. The solution? Market basket analysis. We wanted to understand what people were buying with Coca-Cola. In real stores, with real purchasing behavior. One result still sticks with me. In two major pharmacy chains in the US, two of the top products purchased alongside Coca-Cola were Jim Beam and Jack Daniels. Was that your guess? Not likely. Most people say chips, chocolate, any of your other favorite junk foods. I found this really ironic, given these are pharmacy chains, not liquor stores. That insight mattered. It told us something simple but powerful. If people are buying these products together, placement matters. So we tested a small change. We placed Coca-Cola closer to those products in-store. That one decision unlocked tens of millions of dollars in revenue. Not because we sold more products overall, but because we made it easier for customers to buy what they already wanted together. That’s the real power of market basket analysis. The technique itself is relatively straightforward. The impact comes from knowing what the results mean, how to interpret them, and how to act on them. This week in an advanced Next-Level Tableau class, we recreated that kind of analysis using grocery store purchase data. The GIF shows how quickly these patterns jump out when you visualize them properly. Once you see relationships instead of rows, you stop asking "What sold?" and start asking "What should be next to what?" That’s where analysis starts to change outcomes. I’d love to hear, what kinds of use cases could you apply market basket analysis to?

Stop guessing which discounts to test. Your data already knows. Most brands test discount tiers without looking at what customers already buy naturally. This can potentially be a big mistake. Here’s how to analyze your data so you can test the right discount tiers: Step 1: Pull your data Look at how many customers buy 1 vs 2 vs 3 vs 4+ items. Don't assume it's a smooth dropoff. Step 2: Rate each tier by signal strength • 10%+ buying that quantity = test it • 3-10% = worth testing • 1-3% = probably not worth it (and maybe risky) • Under 1% = skip it Let’s take the example of a wallet brand we helped with analyzing their data: • 88% buy 1 wallet • 4% buy 2 wallets • 7% buy 3 wallets People were skipping 2 and jumping straight to 3. When we dug deeper into those 3-wallet orders: • 50% bought same wallet in 3 colors • 30% bought for family ("dad + 2 sons") • The rest were gifts or small retailers A lot of brands would test a 2-pack discount first. But… Only 4% naturally buy 2 wallets. If you discount 2-packs, you risk pulling those 7% who buy 3-packs down to 2-packs. You might actually lose money. This pattern shows up in a lot of different industries. • Socks: 1 pair or 3+ pairs, rarely 2 • Wine: 1 bottle or 6-pack, not 2-3 • Books: 1 book or vacation stack of 3+ TLDR: Check where demand naturally clusters in YOUR data. Let your actual customer behavior show you which discount tiers deserve tests and which ones are likely not worth testing.

A few months ago, a marketing team at an e-commerce platform was struggling with customer churn despite running aggressive discount campaigns. The assumption was that offering more discounts would improve retention, but after SQL-driven analysis, the real issue turned out to be low repeat purchase rates among first-time buyers. Reducing Customer Churn with Data Analytics 1️⃣ Identifying At-Risk Customers We analyzed repeat purchase behavior to find the drop-off point. SELECT customer_id, COUNT(order_id) AS total_orders, MIN(order_date) AS first_order_date, MAX(order_date) AS last_order_date, DATEDIFF(day, MAX(order_date), GETDATE()) AS days_since_last_order FROM orders GROUP BY customer_id HAVING COUNT(order_id) = 1 AND DATEDIFF(day, MAX(order_date), GETDATE()) > 30; 🔹 Insight: A large percentage of first-time buyers never returned after their initial purchase. 2️⃣ Finding the Root Cause of Low Repeat Purchases We compared product categories and delivery experiences of repeat vs. non-repeat customers. SELECT product_category, COUNT(DISTINCT CASE WHEN repeat_purchase = 1 THEN customer_id END) AS repeat_customers, COUNT(DISTINCT CASE WHEN repeat_purchase = 0 THEN customer_id END) AS churned_customers, AVG(delivery_time) AS avg_delivery_days, AVG(customer_rating) AS avg_rating FROM orders JOIN customer_feedback ON orders.order_id = customer_feedback.order_id GROUP BY product_category ORDER BY churned_customers DESC; 🔹 Insight: Customers who purchased from low-rated categories (e.g., fragile items, late deliveries) were less likely to return. 3️⃣ Improving Customer Retention with Targeted Offers Instead of random discounts, we personalized retention campaigns based on customer behavior. SELECT customer_id, CASE WHEN last_order_category = 'electronics' AND days_since_last_order > 30 THEN 'Offer 10% discount on accessories' WHEN last_order_category = 'fashion' AND days_since_last_order > 45 THEN 'Send personalized style recommendations' ELSE 'No action needed' END AS retention_strategy FROM customer_behavior; 🔹 Insight: Instead of blanket discounts, category-specific retention strategies performed better. Challenges Faced One-time buyers made up a large chunk of new customers, leading to low retention. Poor delivery experiences negatively impacted repeat purchase rates. Generic discounting strategies weren’t increasing loyalty. Business Impact ✔ 12% increase in repeat purchases by improving category-based retention strategies. ✔ Better allocation of discount budgets, leading to a higher ROI on marketing spend. ✔ Enhanced customer experience, reducing negative reviews and churn. Key Takeaway: Not all churn is due to pricing—delivery quality, product experience, and personalized engagement play a bigger role in long-term customer retention. Have you tackled churn problems with data? Let’s discuss!

Why isn’t anyone buying the couches? One of my favorite analysis projects ever was when a national furniture & home goods client brought us in to figure out why high-ticket items like chairs and couches weren’t selling well online. They had all the web data. The heatmaps. The funnels. Every web data report you can imagine, they had looked at. But they still couldn’t make sense of the behavior. So before we pulled a single report, we decided we wanted to know what it felt like to be a customer. So we used the website like real customers, bought products. We visited physical stores and browsed like we were furnishing our own homes. We wanted to feel what it was like to be a customer. Then we started pulling data: - Store locations - Customer ZIP codes - Sales transactions - Survey responses We triangulated customer ZIP, store ZIP, and product SKUs and we started seeing a pattern. Customers within 25 miles of a store were far more likely to buy a chair or couch than those farther away. Why? Well, the theory was that distance can create doubt about how to return items. If you are not close to a store and you buy a couch and if you don’t like it, what does returning it even look like? It's probably a hassle. So instead of pushing them to spend more money on ads or running some CRO tricks, we proposed something pretty simple: "Add clear, human language to product pages explaining how easy it is to return big items like chairs and couches." Almost instantly, the “outer ring” of customers, those 25+ miles out, started buying chairs and couches.

Every Shopify store owner: "What bundles should we offer?" The answer is literally in your store data. You just need to ask. Most people don't know Shopify AI Sidekick can analyze: - Revenue by quantity - Customer purchase patterns - Which combos have lowest return rates - Profit margins per bundle - and much more It'll tell you exactly what to KEEP, MODIFY, ADD, or REMOVE. No guessing. No copying competitors. Just data. Here's the exact prompt that unlocks all of this: [Full prompt below] -------------------- "Analyze my Shopify store data to recommend optimal product quantity combinations and bundles. Analysis Required: 1. Current Performance Review Which quantity combinations generate the most revenue? What are my best-performing vs. underperforming quantity options? How do different quantities affect average order value? Which combinations have the highest profit margins? 2. Customer Behavior Patterns What quantities do customers purchase most frequently? How do purchase quantities differ between first-time and repeat customers? Which quantity combinations have the lowest return rates? What's the relationship between quantity purchased and customer lifetime value? 3. Market Positioning Analysis How do my current quantity offerings compare to industry standards? Are there gaps in my quantity range that customer behavior suggests? Which quantities provide the best perceived value to customers? Recommendation Categories: KEEP ✅ - High-performing combinations to maintain Include specific performance metrics Explain why these work well MODIFY 🔄 - Existing combinations needing adjustments Specify if price, quantity, or positioning needs changing Provide data-driven rationale ADD ➕ - New combinations to introduce Identify gaps revealed by customer behavior Target specific customer segments REMOVE ❌ - Underperforming combinations to eliminate Low-volume options creating unnecessary complexity Combinations cannibalizing more profitable options Key Questions to Answer: - What's my optimal quantity range for different customer types (trial, regular, bulk buyers)? - Which combinations encourage upselling to larger quantities? - How should my quantities align with natural usage cycles? - What quantity strategy maximizes both revenue and customer satisfaction? Output Format: Provide specific recommendations for each category with supporting sales data, customer behavior insights, and expected business impact."