CURRENT SITUATION
Instacart Grocery Basket is an online grocery store operating through a website and mobile app. With good sales, stakeholders want to use customer information to learn about their purchase behaviors and trends to make effective targeted marketing and increase sales further.
GOAL
Perform an exploratory analysis of the presented data to generate patterns and trends based on customer profiles. Discover insights that would help in creating effective targeted marketing campaigns.
DATA PREPARATION
During this stage, using Phyton I carried out the following:
-
Cleaned, wrangled, and checked the consistency of all five data sets.
-
Combined and merged all data sets into the final one to use for analysis, as it contains all the required information in the same set.
-
Derived new columns and aggregated some data to obtain customers' profiling and segmentation. This was a critical summary stage for my future statistical analysis. Some of the code used in this step can be found here.
All processes of merging, aggregation, and deriving new variables were documented in Population flows which you can see below.
ANALYSIS & VISUALIZATION
To provide answers to key questions of marketing departments were carried out:
Time Analysis
-
Identified the busiest days of the week and hours of the day for orders.
-
Determined the time of day when customers spend the most money.
Products Analysis
-
Compared grouped products by price ranges.
-
Analyzed the frequency of orders by departments.
Customer Profiling
-
Created and performed exploratory analysis on each customers’ profiles based on the distribution among brand loyalty, family status, income, age, etc.
Customer Analysis
-
Analyzed how ordering habits differ based on brand loyalty, age, family status, frequency of orders, and geolocation.
The final data set was analyzed using visualization in Phyton (Matplotlib and Seaborn), which helps me to see patterns and trends more clearly and draw conclusions from my data analysis.
SOME OF INSIGHTS
-
The busiest order days are Saturday and Sunday. The peak of order times ranges from 9 am to 4 pm.
-
The Top 5 departments: produce, dairy eggs, snacks, beverages, and frozen.
-
The Loyalty status mainly is dominated by Regular customers, who account of 49% of all orders.
SOME OF RECOMMENDATIONS
-
Targeted ad campaigns should be scheduled outside of working hours, ideally after 7 pm on weekdays and prioritizing weekends.
-
Giving discounts or promotions to new customers on their 10th order and loyal customers on their 40th order.
-
Prioritize ads for items in the top 5 departments.
-
Ads targeting for most dynamic segments are Families (married customers with more than one dependent) in the South of the USA.
RETROSPECTIVE
High Points
I enjoyed merging the data sets into a single dataset and creating clear and informative visualizations with Python.
Project Challenges
Working with a dataset containing over 34 million rows posed technical issues, as the code execution took longer. However, implementing RAM optimization techniques resolved the problem, eliminating the need to divide the dataset into smaller samples.
The price data set and geographical data about the customers were fabricated, and there were no other options but to use the available data as is.
Future Improvements
I would create and analyze customer profiles such as “gluten tolerance” or “pet owners” to identify which ones would be more lucrative for targeted marketing.