Analysis of Consumer Behaviour in an Ecommerce Grocery Business

Table of Contents

1. Project Overview
2. Data Sources
3. Tools Used
4. Data Cleaning and Preparation
5. Exploratory Data Analysis
   • Customer Segmentation
   • Product Purchase Insights
   • Time-related Patterns
   • Cart Analysis
6. Data Visualization
7. Findings
8. Limitations and Assumptions
9. Recommendations to Ecommerce Store and Conclusion
10. References

---

PROJECT OVERVIEW

Bussiness Problem : Improving customer retention and determining product purchase patterns.

Description:

• The dataset contains information about customer orders, including the day of the week, hour of the day, and days since the prior order.

• We can use this data to analyze customer behavior and identify patterns related to purchase of different products.

• The goal is to determine product purchase and re-ordering patterns in relation to consumer behaviour.

• Therefore, the analysis will consists of the following target areas:

  1. Customer Segmentation:
  - What are the distinct customer segments?
  - How can we categorize customers as new or existing?
  -Purporse : To determine the general nature of the Online Store's Users.

  2.Product purchase Insights:
  - Identify which products have the highest sales and re-ordering rate and vice-versa.
  - Purporse : To identify the best performing and least performing products, and draw reasons behind this, to help in improving the business sales.

  3. Time-related Patterns:
  - Explore the order time of the day and day of the week to identify peak shopping times.
  - Calculate which time of the 24hour clock are most orders made, and which ones.
  -Purporse: To help in human resource scheduling and iventory planning.

  4. Cart Analysis:
  - Items per Cart: Explore the number of products added to a cart per order.
  - Cart Abandonment: Identify instances where products are added to the cart but not purchased.
  - Purporse: Determine the rate of product abandonment and how to reduce this.

Data Sources

• The dataset was sourced from a Kaggle accountClick here to view account

• The dataset is from a real E-commerce grocery store called ; 'Hunter's e-grocery for their orders in a given time period in the year 2023 as explained here.

Tools Used

• Cleaning and Preparation: Python programming language was used for data cleaning and preparation. Python is powerful and efficient for handling large datasets.

• Exploratory Data Analysis: Python and SQL were utilized for data analysis and insights extraction.

• Visualizations: Tableau was used to effectively display findings.

• Presentation: Google Slides were employed to create a simplified explanation of the entire process, ensuring easy understanding even for those unfamiliar with data analysis.

Data Cleaning and Preparation

Overview of the data:

The dataset contains the following columns:

- order_id - unique identity of order

• user_id -unique identity of user/customer

• order_number - Number OF THE ORDER

• order_dow -Day of the Week the order was made(either 0,1,2,3,4,5,6) - '0' represents Monday and '6' represents Sunday

• order_hour_of_day - Hour of the day order was made

• days_since_prior_order - Days since prior order ; 0 for new customers, the rest depending on last day

• product_id - unique ID of product that is part of an order

• add_to_cart_order - Number of specific products added to cart as part of the order

• reordered - If the re-order took place ( is in binary of 0 or 1)

• department_id - specific department identity that an ordered product is part of.

• department - name of department

• product_name - name of product

Raw dataset - For detailed information: Click here to view and download original dataset

The following are the steps in my data cleaning process:

Clean dataset - After cleaning and preparation process, the cleaned dataset was also uploaded in my Google Drive: Click here to view and download clean dataset

Exploratory Data Analysis

• 1. Customer Segmentation:

  • In the original dataset, the order_id column has duplicate values.\

  • This is because every order was brocken down to individual products contained in the specific order, leading to a repeat of a unique order.\

  • Thus, I combined the products in respective order batches in order to handle duplicate values.

  • With this, I eliminated data duplication and was able to analyze the different groups of customers.\

    • Click to view the SQL file: [SQL_Consumer_Segmentation] (https://github.com/REVOgati/Ecommerce_Consumer_Behaviour/blob/155c3faa481e21f0e32490816ee21f0323d13587/SQL_Consumer_Segmentation.sql)

  • 2. Product purchase Insights:

    1. Which products are most ordered in specific departments in the dataset? - To determine the most and least purchased products in every department.\

       • I used SQL to group products by departments, and return the total number of orders of each product.\

       -A snapshot of the SQL code and the link:Product purchase insights

    2. Which products are most re-ordered and vice-versa, and why? - To determine which products attract most purchases and vice-versa.

  • 3. Time-related Patterns:

    • At what time of the 24hour day, are the highest orders made? - To help in human resource allocation\
    • At what day of the week are the highest orders made?\
    • What time of the day, is the most purchased products mostly ordered?
    1. I grouped the items according to the values of the 24hour day to determine the time with highest orders.\
    2. Since 0 represents Monday, and 6 represents Sunday, I used SQL to return number of orders for every day.\
    3. I used an SQL code that returns the most ordered product, and its different purchase time periods.

  • 4. Cart Analysis:

    • For every ordered item, I subtracted the number in the cart - order number to determine number of abandoned items and multiplied by 100.
    • This in order to get the percentange rate of cart abandonment.
    • I then stored the result in a new column 'abandoned_items' and counted the total number per product.

• Data Visualization

  • Visual representations are available on Tableau Public.

Findings

• Consumer Segmentation

  • There 103,761 unique orders made.
  • Approximately 8,405 entries with value 0 in days_since_prior_order, indicating new users.
  • Majority of customers are returning users.

• Product purchase patterns - The most top six most ordered products are 'fresh fruits', 'fresh vegetables','packaged vegetables' 'fruits', 'youghurt', 'milk' and 'packaged cheese' respectively.
  - These six fall under the departments of either 'produce' or 'dairy eggs', each having three items.\

    - The least ordered products are 'frozen juice', 'shave needs', 'beauty', 'first aid', 'eye ear care' and 'kitchen supplies' respectively.\
    - Four of these items fall under the department 'personal care' with 'frozen' and 'household' having each one.
  

• Time-related patterns 1. - The hours 10,11,14,15,13,12 have the highest number of orders.
  -The hours 3,4,2,5,1,0 have the least number of orders.
  - The above data can inform human resource allocation in the recommendations section below.\

    2.  - The highest number of orders are made on Friday while the least being made on Monday.\
    	- This result may be attributed to factors only unique to the specific grocery store as much cannot be explained based of a global point of view.\
    	
  
    3.	- The most ordered product is fresh fruits.
  
    	- The time period 10am to 3pm has the highest number of orders for fresh fruits\
    	- while the time period of 12am to 5am has the least number of orders for fresh fruits.\
  

• Cart Analysis - All results were either less than 1% or a negative for the percentage rate of cart abandonment.
  - This means that the percentage of users abandoning items they add to Cart is extremely low.
  - Therefore, this is not a problem for this Ecommerce Store.
  - An assumption may be, since most users are returning customers, they are well aware of the satisfying quality of the products they order.
  - Or, due to essential nature of grocery items being sold.

Limitations and Assumptions

1. There were many unique USER_IDs that had more than 1 occurence of the value 0 in the days_since_prior_order column.\
	-I assumed they may have made more than 1 order in their first day as the only possible explanation. \
	- I used filtering and grouping techniques to handle this and get the correct number of new users in the period.\

2. The data is collected over a long period of time, however the limit is set to 30 days.\
	- Therefore, those users that only bought once in this period are given the value 30 in the 'days_since_prior_order' column.\
	- Therefore, those with 30 in the said column are not new users, but already established users,that have not ordered in a long period.\

Recommendations to Ecommerce Store and Conclusion

1. On products insights\

   • Given that the departments 'produce' or 'dairy eggs' have the highest orders, I would recommend measures to ensure consistency of quality of these products as well as their availability.\
   • Given that the departments 'personal care', 'frozen' and 'household' have the lowest orders, measures should be put in place to promote products of these departments.\
   • They might include improvement of product quality, research from users, advertisements or discounts.

2. Time-related

   • I would recommend that there should be a significantly higher number of human resource allocation of the employees in the period 10am - 3pm due to the higher number of orders.
   • Thus, there should be a smaller group at the hours 12am to 5am due to the lower number of orders.
   • This helps to balance the number of human resource against traffic in order to ensure optimality in performance.

3. Consumer Segementation

   • 8% of all the customers are new users while the rest are exisiting users.
   • This means that a very huge number of customers are returning.
   • I would recommend surveys to understand reasons for returning customers.

4. Cart Analysis

   • The store does not have a major problem with cart abandonment as the average rate is at a percentage of less than 1%.
   • This might be due to the essential nature of most of its products as groceries and household items are always very readily needed.

References

• Visualizations : Path to my Tableau Dashboard for this project : Ecommerce Store Viz

• Data Source :

  • The dataset was sourced from a Kaggle accountClick here to view account

• The dataset is from a real E-commerce grocery store called ; 'Hunter's e-grocery for their orders in a given time period in the year 2023 as explained here.