TL;DR: Delivery speed is slow. Suggested parcel lockers for the customers and a trip planner for their safety.

Intro

It's my favourite assignment, since I was allowed to choose my own data science tool. I came across Olist's dataset on Kaggle and decided to use it for the assignment. Olist is a Brazillian e-commerce platform that has been growing rapidly. I was required to find a business problem from the data and I am going to discuss about the delivery system.

The codes are available here.

Data Analysis

Descriptive Analysis (What is the problem?):
The problem identified is that it takes an average of almost 10 delivery days. None of the customers in the dataset were repeat customers. Thus, based on the average revenue of each order, each customer only generated revenue of 50 to 150. For order reviews that mentioned about delivery, it has lower review scores on average, compared to those that did not. In fact, even though in 2018, average delivery days were shorter, reviews scores that are related to delivery become lower.

Diagnostic Analysis (Why did the problem occur?):
It takes an average of 45 minutes to approve an order. The number increased slightly in 2018. Boleto takes an average of 5 hours to be approved and was the second most used payment type. However, the actual delivery process takes up most of the delivery time taken. Order volume, order weight and order distance do not have significant correlations with delivery duration.

Information Analysis

The national courier service in Brazil was reported to be having delays with delivery due to the outbreak. However, Brazil was not only privatizing the national courier service, but it has had problems in the past, especially related to employment.

While public transportation is provided, citizens still choose private transportation, as public transportation development could not match up with the high rate of urbanization. Without private transportion, it takes 60% more time, which is roughly 2 hours, to travel to work in urban areas. 61.1% of the freights in Brazil uses road transportation. Thus, the long delivery duration found in the data analysis could be due to transportation system.

Knowledge Analysis

Last-mile delivery is essential for customer satisfaction and is the most expensive component in order delivery. Researchers have proposed a few solutions regarding last-mile delivery, such as:
1. The integration of first-mile pickup and last-mile delivery
2. The integration of collection-and-delivery points (CDPs) in last-mile delivery
3. Allowing customers to collect their products at parcel lockers

Critical Analysis

I recommended solution number 3 from the knowledge analysis. I could not find which company acquired the national courier service, so controlling the delivery route might not be practical. Parcel lockers not only increase customer convenience, but also decrease delivery distance.

Remember Boleto? Although it takes almost 5 hours to approve, it was still the second most famous payment type by the customers. It is famous because it allows customers to pay through bank wiring, especially for those without credit cards. At the same time, the most used payment type was credit card.

I suggest encouraging customers to use parcel lockers by giving them vouchers of goods and services nearby their chosen parcel lockers. The visualization of the suggestion can be found in the dashboard.

To protect the customers while they collect their orders, parcel lockers should be equipped with security cameras . It should also be convenient for different types of transportation. In fact, it's not all that safe to travel in some areas in Brazil. So, taking inspirations from Graband Amazon, I also included a table of business data that might be needed for a smart trip planner that includes safety features for the customers to collect from parcel lockers.