Maybe it was a chocolate bar perfectly placed next to the checkout, or a kitchen gadget you didn’t know you needed until you saw it paired with the blender you were buying.

These aren’t random events. They’re the result of a smart, behind-the-scenes data strategy called Market Basket Analysis (MBA) — a technique retailers use to uncover the secret connections between what we buy and why we buy it.

What is Market basket analysis?

When customers buy bread, they often end up buying butter too. This buying pattern can be described with a simple rule:

IF {bread} THEN {butter}

In simple terms, MBA is a data mining technique used to identify patterns in customer purchasing behaviour — helping businesses group products that are frequently bought together. This insight is used in everything from recommendation engines and product bundling, to store layout design and marketing strategies.

Market Basket analysis using Apriori algorithm:

Market Basket Analysis uses simple association rules like IF {A} THEN {B} to predict which products are likely to be bought together.

There are several algorithms used to perform this analysis, including AIS, Apriori, SETM, FP-Growth, and Eclat. Among these, the Apriori algorithm is one of the most popular because it is easy to understand and implement.

Three key measures used in Market basket analysis are

1. Support: Frequency of items appearing together
2. Confidence: Likelihood of item B being bought when item A is bought
3. Lift: How much more likely items are bought together compared to being bought independently

To better understand Market Basket Analysis (MBA), imagine a small grocery store where customers purchase everyday items like bread, butter, milk, and tomatoes. In this example, the goal is to determine how often customers who buy bread also purchase butter.

1. Support — It measures the proportion of transactions in which both bread and butter are purchased together. It is calculated as:

This indicates that 60% of all transcation include both bread and butter.

2. Confidence — represents the likelihood of purchasing butter given that bread has been purchased.

This implies a 75% probability that customers who buy bread will also purchase butter.

3. Lift— assesses the strength of the association between bread and butter, determining how much more likely these items are bought together than would be expected if they were purchased independently.

Lift< 1, the products are not frequently bought together by consumers.

Lift> 1, the products are frequently bought together by consumers.

Lift = 1, the purchase of one product does not affect the purchase of the other.

A lift of 1.25 means customers who buy Bread are 25% more likely to buy Butter compared to the average. Since Lift > 1, this is a positive association.

🛒 Market Basket Analysis using Apriori Algorithm

Let’s break down how MBA works in practice using Apriori Algorithm.

Step 1: Install apyori and all the required libraries.

!pip install apyori

import pandas as pd
import matplotlib.pyplot as plt
from collections import Counter
from apyori import apriori

Step 2: Begin by installing the apyort

df = pd.read_csv("dailytransactions.csv")

Step 3: Convert each transaction row into a list of non-null, trimmed items

transactions = []
for _, row in df.iterrows():
    transaction = [str(item).strip() for item in row if pd.notna(item)]
    if transaction:
        transactions.append(transaction)

Step 4: Extract Rules and Convert rules to Data Frame

rule_data = []
for rule in results:
    for stat in rule.ordered_statistics:
        if stat.items_base and stat.items_add:
            rule_data.append({
                'LHS (If bought)': ', '.join(stat.items_base),
                'RHS (Then also bought)': ', '.join(stat.items_add),
                'Support': round(rule.support, 4),
                'Confidence': round(stat.confidence, 2),
                'Lift': round(stat.lift, 2)
            })
df_rules = pd.DataFrame(rule_data)

Step 5: Sort and Display Top 10 Rules by Lift

top_10_rules = df_rules.sort_values(by='Lift', ascending=False).head(10)

print("\n Top 10 Association Rules (Sorted by Lift):\n")
print(top_10_rules.to_string(index=False))

Step 6: Final Results

plt.figure(figsize=(12, 6))
plt.barh(
    [f"{row['LHS (If bought)']} → {row['RHS (Then also bought)']}" for _, row in top_10_rules.iterrows()],
    top_10_rules['Lift'],
    color='coral'
)
plt.xlabel("Lift")
plt.title("Top 10 Association Rules by Lift")
plt.gca().invert_yaxis()
plt.tight_layout()
plt.show()

Summary

This practical walkthrough demonstrates how Apriori helps uncover hidden patterns in customer purchases. Retailers can use these rules to:

• Recommend items (e.g., “Customers who bought bread also bought butter”)
• Bundle products for promotions
• Organize shelves for convenience and sales optimization

Market Basket Analysis, powered by algorithms like Apriori, transforms raw transaction data into strategic insights — ultimately enhancing the shopping experience and driving revenue growth.

***

I hope you find this article helpful. If I’ve missed anything, let me know in the comment section.

What do you think about this article? Comment!!!

Not sure what to read next? I’ve picked another article for You !!!

Cohort analysis | Sentiment analysis

Follow me @ Medium | LinkedIn

Thanks for reading!

Uncovering Shopping Patterns: Market Basket Analysis Using the Apriori Algorithm was originally published in Towards Explainable AI on Medium, where people are continuing the conversation by highlighting and responding to this story.

In this article, we will see an Easy and simple way to analyze the performance of the sales team using the following icons.

Dataset: It contains sales data of 15 employees.

Steps to be followed:

1. Ranking: Formula- RANK(B3,B$3:E$17,0), then copy the same formula to other cells as shown in the picture.

Function Arguments:

Number: Number for which you want to find the rank. In this case, 84(B3)

Reference: Reference table which is B$3:E$17.

Order: If it is 0(zero), the highest number gets Rank-1 and if it is 1(one), the lowest number gets Rank-1.

Output:

2. Select L3 to N17 and change the font to ‘Wingding 3’, then use the following formula.

Formula: IF(H3<G3,”h”,IF(H3>G3),”i”,”j”))

FINAL OUTPUT:

*****

I hope you find this article helpful. If I’ve missed anything, let me know in the comment section.

What do you think about this blog? Comment!!!

Not sure what to read next? I’ve picked another article for You !!!

Cohort analysis | Sentiment analysis

Follow me @ Medium | LinkedIn

Thanks for reading!

Originally published at https://www.linkedin.com/in/shaik-subhani/

Sales data analysis using Excel was originally published in Analytics Vidhya on Medium, where people are continuing the conversation by highlighting and responding to this story.

Sentiment analysis is the process of determining whether a piece of text is positive or negative. It allows businesses to identify their customer’s sentiments towards products or services through reviews and online feedback.

For Example:

In other words, sentiment analysis gives an opportunity to explore the mindset of your customers and study the state of your product or service from your customer’s point of view.

This makes sentiment analysis a great tool for

1. Product reviews
2. Market research
3. Customer service
4. Social media monitoring
5. Reputation management etc.,

Sentiment analysis is a filed of Natural language processing(NLP).

INTRODUCTION

The dataset which we are using contains, restaurant reviews and we are going to use sentiment analysis to find wheater a particular review is positive or negative. If the review is negative we will display the following message.

Thank you <customer first name> for taking out the time to write a review. we apologize for the inconvenience caused. We hope we will get a chance to serve you better in the near future.

If the review is positive we will display the following message.

<Customer first name>, Thank you so much for that awesome review. We look forward to serving you again in the near future.

Steps to be Followed

1. Importing the libraries.
2. Importing the dataset.
3. Cleaning the data.
4. Creating the Bag of words.
5. Training and classification.
6. Confusion matrix.
7. Predicting a customer review.

Step1: Importing the Libraries

Step2: Importing the Dataset

Step3: Cleaning the data

From the above dataset output, we will find some information that does not help in determining whether a review is positive or negative.

For example, words like a, an, the, was, on, etc, doesn't have any impact on the decision. These words are called Stopwords. We also have words like loved, stopped, loving, etc, we will have to convert them to its root form.

Step4: Creating a bag of words

The bag of words is the simplest form of text representation in numbers. To learn more about bag of words — click here

Step5: Training and Classification

We have to split our dataset into training and testing sets, and then we have to apply the classification model.

Classification Algorithms

1. Linear classifier: Logistic regression, Naive Bayes classifier
2. Nearest neighbors.
3. Support vector machine.
4. Decision tree.
5. Random forest etc..,

We can use any one of them, I tried all and surprisingly Naive Bayes gives the best accuracy.

Step6: Classification model

It is a tool, used for evaluation of the performance of a classification model. If you want to learn more about this — Click here.

The accuracy of our model is 73%, which means 73% of the predictions are accurate.

Step7: Final step - Predicting a customer review

I hope you find this article helpful. If I’ve missed anything, let me know in the comment section.

What do you think about this blog? Comment!!!

Follow me @ Medium | LinkedIn

Thanks for reading!

Sentiment Analysis With Bag-of-Words was originally published in The Startup on Medium, where people are continuing the conversation by highlighting and responding to this story.

To improve Customer Retention and Reduce Churn

I’ve read a lot of articles and blogs on cohort analysis, churn and customer retention, most of them are very technical & hard to understand. So, I decided to choose a different path and try to make it simple as well as actionable.

If you want to know more about cohort analysis, and how to apply it to improve your business, this blog is perfectly made for you.

Let’s jump right into it!

WHAT IS COHORT ANALYSIS?

A Cohort is a group of people who share a common characteristic over a certain period of time, such as first purchase date, signup date, the location they belong to, etc.,

Example: Class — 2020 is a cohort, and all the students who graduated in that year are parts of that same cohort.

In cohort analysis, we track these group of people over a certain period of time, and try to draw insights to get a better understanding of your product/service / Marketing campaigns, etc…,

Cohort analysis is a widely used method to reduce the churn rate and improve customer retention. Before we dive into the cohort analysis, first try to understand what is churn rate and customer retention.

CHURN RATE:

Percentage of customers who stop using or buying from your website/store/e-commerce site etc, within a given period of time. Churn rate is sometimes also called a ‘Rate of Attrition’.

CUSTOMER RETENTION:

It is the ability of your business to retain its customers over a specific period of time. High retention means, your customers are buying and engaging with your product or service continuously.

According to me:

HOW TO READ A COHORT ANALYSIS CHART:

At first, glance, reading a cohort chart is a little confusing and hard to interpret. I also felt the same, when I first saw a cohort chart, but trust me it is indeed very easy to interpret.

Confusing right? Let’s break it and just look at one cohort…

1. 2010–12: Cohort of customer signed up or purchase something from you.
2. 885: Total number of customers belongs to this cohort or 885 customers signed up or brought something from your business in 2010–12.
3. 37%: 37% (of 885) customers visited your store next month. So we lost around 73% of customers.

HOW TO USE COHORT ANALYSIS TO REDUCE CHURN:

If you look at the first cohort (highlighted in the box), too many customers are churning out. In the first three months, we are losing around 70 -80% of customers.

In the last month (highlighted with dark red), only 11–15% customers are decided to be connected with you, that means you are losing 90% of your customer which is a very bad sign.

The major reason behind this could be, you recently started something new (maybe launched a new product or a marketing campaign, etc..,) and your existing customers are not happy with your initiative.

THE TOP REASONS BEHIND THE HIGH CHURNING RATES:

1. There is a mismatch between the products or services and to customers’ expectations.

2. Poor customer service or after-sales service.

3. Quality or trust issues.

4. Checkout process.

3-STEP GUIDE TO IMPROVE CUSTOMER RETENTION:

STEP 1: Identify the drop-off trend in the cohort analysis chart. Drop-off: from which month most of the cohorts are falling down. Let say, after 6 months most of the cohorts are falling down.

STEP 2: Communicate with your customer before the drop-off trend, start sending check-in emails from the 4th month, and see how things are going, find out whether your customers are happy or not, make sure to note all the issues they are facing with your product or service.

STEP 3: Follow — up with your customer until their issue resolved, and after that don’t forget to take the feedback. In this step don’t pressurize them to use your product or service, and try to focus more on understanding the customer problem and solve it as soon as possible.

NOTE:

1. Collect all the reasons from the customers, who stopped buying or engaging with your product or service. It will help you to understand why they are leaving?

2. Send a survey to those customers who stayed or be connected with you. Try to find what they like and dislike about the product or service.

3. Compare 1 and 2, try to identify the exact problem in your product or service, as well as find out what you can do to keep your customers happy.

Cohort implementation in python click here

CONCLUSION:

The best part of cohort analysis is, it enables us to find which customers are leaving, when, and why they are leaving. Which indeed very helpful to improve customer retention and reduce churn.

What do you think about this blog? Comment!

Thanks for reading!

Follow: Shaik Subhani @ LinkedIn

Cohort Analysis : Simplified, Actionable was originally published in Analytics Vidhya on Medium, where people are continuing the conversation by highlighting and responding to this story.