A Serverless Amazon Web Services Prototype for Customer Behavior Analytics Using Recency, Frequency, and Monetary Segmentation

Many small retailers record every sale yet rarely study the records. This preprint describes a small serverless pipeline on Amazon Web Services (AWS) that takes one retail transaction file through storage in Amazon S3, cleaning with written rules, SQL analysis with Amazon Athena, customer segmentation, and a web dashboard. Because no suitable open Indian retail dataset with customer identifiers was available, a simulated dataset of 40,263 order lines for 5,000 customers was used. Eleven cleaning rules kept 39,353 rows (97.7%). Athena views scored 4,995 customers on recency, frequency, and monetary value (RFM) and sorted them into seven segments. Champions were 19.7% of customers and produced 49.6% of revenue. Of 26 test cases, 24 passed; the other two were limited by the free account. Athena queries ran in 488 to 904 ms on a 4.5-MB file, and ten users rated the dashboard 3.82 out of 5. The data are simulated, so the results show that the method works, not how a real retailer behaves. This preprint has not been peer reviewed.