Insight Discovery – Deep Analysis of Customer Behavior

Our client, an Amazon seller with multi-million dollar sales, consistently analyzed the market and launched new products in the creative goods sector. One approach to finding new products was the introduction of bundles, which are sets of combined items. These bundles could consist of related products that customers often use together, or a product paired with essential supplies for initial use.

As order data accumulated, the idea emerged to conduct a deep analysis of this data to identify patterns in what products customers frequently purchase together or subsequently order, with the goal of creating bundles based on these insights.

Initially, the task was defined as follows: analyze all orders over several years to identify potential bundles.

1. Identify products frequently purchased together. Highlight the most popular pairs and trios, and generate a report showcasing which products are often bought together, their total quantities, and their percentage of overall orders.
2. Identify products that are purchased sequentially. Often, customers do not buy two or three items at once but make additional purchases later.

During our testing, we discovered additional insights and proposed creating extra reports:

1. Initial Product Analysis: The first product purchased influences the overall perception of the brand. This report shows the likelihood of customers returning and making another purchase after buying a specific initial product. It highlights which products perform best and, conversely, which ones underperform and damage the brand’s image.
2. Repeat Purchases: Some products are purchased repeatedly by customers, and the proportion of such customers can be significant. This insight can be used to create subscription services for products or bundles containing multiple identical items.
3. Customer Flow Between Categories: Sometimes, after becoming familiar with products from one category, customers start purchasing products from another category within the same brand.

These reports proved valuable not only for the Research and Development department in launching new products but also for the Marketing department in understanding customer psychology and developing advertising campaigns.

We developed an automated Python script that extracts and groups all orders from various sales channels (Amazon, Shopify, Retail), analyzes all products purchased by a single customer within a single order or across multiple orders, and generates several ready-to-use reports. The script is implemented as an Airflow DAG, a reliable data pipeline management system, and automatically generates reports on a monthly basis.

The processed data is presented in a custom report format in Tableau, making it accessible for viewing or export.

The insights gained allowed the Research and Development department to launch new composite products and enabled the Marketing department to improve advertising strategies, ultimately enhancing the client’s business performance.