Customer Lifetime Value Analysis for an Online Consumer Goods Retailer

Project Description

Who We Are: 

We are an e-commerce company based in Los Angeles, California. We sell various scented products (candles, bath bombs and beads) to customers throughout the U.S. Our sales are generated primarily through three (3) channels: (i) paid advertising on platforms such as Facebook, Google, Yahoo and Pinterest, (ii) email marketing and (iii) our subscription program which offers a candle each month and certain other perks called the VIP Club.

What We Need:

We are looking data scientists to help us better understand the customer data we possess and how certain dimensions and variables affect the likelihood that our customers will purchase from us again. Specifically, we want the following:

1. A multidimensional lifetime value database broken down by time intervals (1 week, 3 weeks, 1 month, 3 months, 6 months, 9 months) after the initial order and filtered through the following dimensions:

• initial order size
• time between order and shipping label creation
• time between shipping label creation and shipment
• time between send and delivery
• location, time of day order placed
• traffic source from which order generated
• offer/deal for initial order
• city and state where the customer lives
• average email opens before second purchase
• Month of year

We would like to run each of our existing products through this model and would like to understand how candles (our first product) differ from bath bombs (our newest product) in the above respects.  We want you to create a data pipeline that will ingest all of our old data into a format incorporating the above dimensions that is also able to ingest all of our new data. 

2. A lifetime value analysis of our customers that purchased the VIP Club membership as compared to the customers that did not purchase the VIP Club membership.

• The effect that the device used by the customer to purchase (desktop, iPad, mobile) has on LTV.

3. Lifetime value analysis of customers that did not leave reviews as compared to customers that did leave reviews. 

What We Have

Approximately 600,000 sales records.Email open and click-through rates tracked on Klaviyo.

Email open and click-through rates tracked on Klaviyo

Old data format:  multiple CSVs, JSON files keyed by email.New data:

New data:  segment.com raw data files, data warehouse. 

Goal

Using the data, we would like to be able to do predictive analysis on our customers so that we can know, based upon their behavior and characteristics, how likely they are to purchase again.