Tips For Interpreting Consumer Purchase Patterns

I remember years ago working with a coffee brand, and we discovered some fascinating insights from analyzing customer buying behavior. We had two types of purchases: subscriptions and one-time buys. When we dug into the data, we found a significant pattern. Only 18% of one-time buyers made a second purchase. But if they did, there was an 85% chance they’d order a third time, and the repeat order rate stayed high after that. This showed us a major bottleneck. The founder initially wanted to focus all incentives on attracting first-time buyers, but the data told a different story. We saw the value in driving that crucial second purchase. So, we overhauled our approach: 1. Revamped Fulfillment Kits: The first order kit included incentives for a second purchase. 2. Updated Email Campaigns: Emails were tailored to encourage a second buy. The results? We boosted the second purchase rate to nearly 30%, leading to a significant increase in overall sales and customer lifetime value (LTV). Even with pushing more people into that second order, we only saw a small dip in the number of people who went from a 2nd to a 3rd order, moving from 85% to 83%. This experience shows the power of slicing your data by cohorts to uncover bottlenecks and then addressing them directly. Sometimes, the biggest gains come from focusing on the steps beyond the initial sale.

There are several key ways to boost sales on Amazon. One of the most impactful: Target your customers when they are most likely to buy again. To do this, you need to understand your customer’s repeat purchase behavior. Here’s how we do this for our clients: At Cartograph we built an analytics dashboard to help our clients better understand the repeat purchase behavior of their customers. The data was being produced by the customer and ingested by the brands, but it wasn’t being leveraged to its greatest potential. So Cartograph build tools to analyze, visualize, and act on this data. What we discovered was different from what we expected. We segmented repeat purchases into 4 categories: 1. How many people repurchase within 0-15 days? 2. How many people repurchase within 16-30 days? 3. How many people repurchase within 31-60 days? 4. How many people repurchase within 3 months? For some product categories, we created additional cohorts for as far as 6 months beyond the initial date of purchase. When repurchases were grouped into cohorts and spread across an x-axis, a strange curve emerged in the graph. It was a bimodal distribution: The first peak occurred within two weeks of the initial purchase, then a trough, and then a second peak between month 1 and month 2. When we began to connect other brands to the Repurchasing Dashboard, we saw the exact same results with varying intensity — Two peaks, one trough. After weeks of research, we were able to piece together the cause of this behavior. Take, for example, a tube of toothpaste. A single tube will last about 3 months before it has to be replaced. Therefore, you would expect to see a single slope that grows exponentially as it approaches the 90-day mark. But, there are a shocking number of repeat purchases between days 0 and 15. As it turns out, the most likely (and most common) explanation for this consumer behavior is: - Stocking up with a larger supply of the product - Deciding to order more of that same item (or a slight variant of that item) for yourself or for friends/family. - Clicking “Buy Now” twice in a row (instead of adding several items to a single cart) That’s why, for every single product, we saw two large repurchasing spikes: One at the beginning and one towards the end. That’s what makes this so interesting — From our analysis, many of those curves have bimodal distributions of varying shape and intensity. In fact, we found that the most common repurchasing behavior came 30-45 days after that first transaction. This allows us to help our clients by tailoring our look back windows with DSP to target customers when they are most likely to buy again. As a marketer, the more familiar you are with that curve, the better equipped you’ll be to answer those three vital questions: “What products should I be promoting?” “When should I be promoting them?” “Who should I be promoting them to?”

I've spent 10 years figuring out how to predict repeat customer purchases online. Here’s how to do it right and get to 95%+ accuracy: If you want to understand your repeat customers and predict their behavior, it all starts with cohort analysis. This sounds fancy, but it’s just grouping customers based on the date they made their first purchase. From there, you can build a clear picture of what’s happening in your business. Here’s the step-by-step process: 1. Assign customers to cohorts. Start by grouping customers by the month (or week, depending on your volume) of their first purchase. This will be the starting point for tracking retention and repeat purchase behavior. 2. Establish a baseline retention curve. Most customer behavior follows a predictable pattern: orders gradually taper off over time. Plot this out to create a baseline curve—a starting point to measure future cohorts against. 3. Weight for recent behavior. Here’s the thing: the customers you acquired last month are much more relevant to forecasting than the ones you acquired three years ago. Weight your analysis to focus on recent cohorts to get a more accurate picture of what’s next. 4. Segment by customer type. Not all customers behave the same way. You might notice early customers were all over the place—some subscribing, some buying once. Breaking this down by type (e.g., subscribers vs. one-time buyers) makes the data a lot more actionable. 5. Adjust for seasonality. Timing matters. A customer you acquire in October is probably going to shop again in November because… Black Friday. That doesn’t mean they’re inherently “better,” but you need to account for these factors when predicting future behavior. 6. Predict orders, not people. Instead of predicting how many customers will come back, focus on the total number of orders a cohort will generate. Then multiply that by your average order value to get to revenue. Trying to count subscribers, then adjust for churn, reschedules, or payment failure will create lots of inputs to manage and ultimately leads to precision without accuracy. 7. Keep it fresh. The most accurate forecasts come from constantly updating your data. Monthly refreshes are usually the sweet spot—they let you capture new trends without bogging you down with constant updates. Sounds like a lot of work? It doesn’t have to be. Drivepoint does all of this out of the box. Want to see how it works? We can ingest your Shopify and Amazon data into actionable retention and revenue forecasts and show you the results. Link in the comments to book time if you want to learn more. 🚀 #CohortAnalysis #Forecasting #Shopify #Amazon #DTC