Customer Shopping Basket
Analysis for Small-Scale
Grocery Stores and Vendors
February 18, 2022
-By Prachet Shah, A. Shraddha
Step 1: Prototype Selection
Abstract
Because of the Corona Pandemic and Lockdown, small-scale grocery shops and vendors
have been affected a lot since the number of customers coming to them have reduced
drastically and they have near to none profits.
So in my report, I want to apply the Association Rule Mining to predict and optimize the
shopping experience for customers and selling experience of vendors and storeowners
using Machine Learning and also provide a small-scale code implementation for it.
This technique is extensively used by large e-commerce stores and multi-chained
supermarkets to boost their income and retain customers.
1.Problem Statement
To predict similarities in shopping habits and most bought combinations of items in
grocery stores, so the vendor or store owner can know which items are sold the most and
which items are bought frequently so that they can manage their inventory and come up
with new schemes for increasing their sales and eventually profit.
2
It is an analyzing technique based on the idea that if we buy an item then we are
bound to buy or not-buy a group (or single) item. For example, if a customer is buying bread
then the chances of him/her buying jam is more.
2.Market/Customer/Business Need Assessment
There has been a huge downfall for Retail Shops and Vendors because of the unfortunate
pandemic which has led many to buy things online instead of shops. So, it is extremely
crucial for them to optimize their selling techniques by managing their inventory correctly
by purchasing items which are in more demand and providing schemes on items which are
generally grouped together to maximize their profits.
In addition, the epidemic has substantially altered client purchasing preferences. As a result
of this technique, we hope to supply small businesses with beneficial data insights and
revenue-generating opportunities.
3.Target Specifications and Characterization
1. Boost of Sales: Using this technique, vendors and shopkeepers can visualize which products
or items are more profitable for them, so they can focus on which items to buy more for their
inventory and which items to buy less according to customer needs so their management of
loadout and finance is improved.
2. Customer Retention: With usage of analysis created by models, shopkeepers and vendors
can group certain items together for easier shopping for customers and keep them happy
and satisfied as they don't have to look for things bought in combination differently. For eg:
keeping bread and butter closer to each other as they are bought in combination most of the
times. For vendors, analysis aims to suggest to them, frequency of bought vegetables and
fruits, so they can focus on that cultivation more to increase their sales and maximize their
profits.
3
3. Business Scheme: Various schemes for promotion of business can be created by
shopkeepers based on customer shopping habits so that customers can buy combinations
of things for cheaper rates which will help in increasing profits. Eg: Discount on milk when
customers buy cookies or cake powder as many buy them together.
4.External Search( Information and Data Analysis )
These are some of the sources I visited for more information and need for shopping pattern
analysis of customers.
1. Why Market Basket Analysis is Crucial to Gain a Winning Edge in the Retail Sector
2. Advantages of Market Basket Analysis in B2B Marketing
3. What is Market Basket analysis?
I am going to use this Dataset for my code implementation for this report.
Dataset Description:
The dataset used in these models contains customers bought items. Each row corresponds to the
item bought by one customer in one invoice. We have to find which items to be added in a buy one
get one deal.
First import the basic libraries for data preprocessing:
4
Let’s now see more info on our dataset:
5
Description and List of Transactions created for our Association Rule Mining Algorithm:
6
5.Benchmarking
Lot of ecommerce stores like Amazon, Snapdeal, and Flipkart use these techniques to improvise
their sales and to also create a smooth shopping experience for customers. Generally,
Benchmarking involves comparing project processes and performance metrics to either industry
best standards and practices or successful completed projects. For this there is a need to
continuously search for implementation of better techniques which lead to better results or outputs.
6.Applicable Patents
1. Method and system for researching product dynamics in market baskets in conjunction
with aggregate market basket properties
2. Enhanced Market Basket Analysis
3. Method and apparatus for retail data mining using pair-wise co-occurrence consistency
7.Applicable Regulations(Government and Environmental)
1. Data collection and Privacy of Regulations of Customers.
2. Government norms for Small Businesses and Street Vendors
3. Rules against False Marketing
4. Employment Schemes and laws created by government
8.Applicable Constraints
1. Lack of initial data to perform algorithms.
2. Convincing Shopkeepers and vendors to use this technique of selling over traditional means.
3. Lack of technical knowledge of vendors.
7
4. Rarely bought items will not be detected by algorithm, so it won't be generated as an
output, so shopkeepers need to note which items are rarely bought and buy them in small
quantities.
5. Need to continuously update and manage the data and model.
9.Business Opportunity
This technique of using Association Rule Mining to group Product Combinations and
recommendations is extensively used by larger companies and it is still improving day by day. When
small shop owners and vendors start using these techniques, they will not only improve their sales
but they will also have an in-depth analysis of what things customers are buying and what they are
not buying. That will also help them with maintaining their budget and which will eventually help
them increase their reach and have growth in their business.
10.Concept Generation
This product requires the Machine Learning algorithm of Apriori to be written from scratch based on
our needs and requirements. The optimization of the model is generally found out through tweaking
in some hyperparameters. There are two methods to do this. If we want to have high speed but low
lift and confidence, we can use the Eclat algorithm which is a derived algorithm from Apriori which
only groups based on support parameters, but for overall high working efficiency we will be using
the Apriori Algorithm here.
Hyperparameters:
We have taken minimum support = 0.003 ,minimum confidence= 20%, minimum lift = 3,
minimum length = 2 and maximum length = 2
To have better output and we found the results satisfactory.
Then we have results generated for the hyperparameters according to them .
Then we come up with the result of hyper parameters in the dataset with confidence in it.
8
11.Concept Development
The required model can be launched by using the appropriate API (like Flask and Django) and its
deployment can be done using Heroku for Flask API. The cloud services have to be chosen
according to the need and budget of the customer.
Final Results:
9
10
12.Final Product Prototype/ Product Details
The final product provides service to operators about the most bought combinations of products for
them to analyze customer shopping patterns and helps them manage their inventory and also
create new strategies and schemes to increase their sales.The service implements the Market
Basket Analysis, i.e Association Rule Mining technique on the dataset of transactions collected
from the shopkeepers/vendors.
Some dynamics of the Apriori Algorithm used in this model and their meaning.
1. Support: It tells us about the combination of items bought together frequently. It gives the
part of transactions that contain both A and B.
11
2. Confidence: It tells us how frequently the items A and B are bought together, for the no. of
times A is bought.
3. Lift: It indicates the strength of a rule over the randomness of A and B being bought
together. It basically measures the strength of any association rule.
A) Feasibility
This project can be developed and deployed within a few years as SaaS( Software as a Service) for
anyone to use.
B) Viability
As the retail industry grows in India and the world, there will always be small businesses existing
which can use this service to improvise on their sales and data warehousing techniques. So, it is
viable to survive in the long-term future as well but improvements are necessary as new
technologies emerge.
C) Monetization
This service is directly monetizable as it can be directly released as a service on completion which
can be used by businesses.
12
Step 2: Prototype Development
Github Link: Customer Shopping Basket Analysis for Small-Scale Grocery Stores and Vendors
Step 3: Business Modeling
For this service, it is beneficial to use a Subscription Based Model, where initially some features will
be provided for free to engage customer retention and increase our customer count. Later it will be
charged a subscription fee to use the service further for their business.In the subscription business
model, customers pay a fixed amount of money on fixed time intervals to get access to the product
or service provided by the company. The major problem is user conversion; how to convert the users
into paid users.
Step 4: Financial Modeling
It can be directly launched into the retail market.
Let's consider our price of product = 250 for getting our graph
13
Financial Equation:
Y = X * (1 + r)t
Y = (X)* (3.2)t
Y = Profit over time, X = Price of our Product, r = growth rate, t = time interval
1+r = 1 + 3.2% = 1.032
Conclusion
Market basket analysis is being used by an increasing number of companies to acquire beneficial
insights about associations and hidden relationships. However, for small businesses, this extension
is a fantastic opportunity to boost sales and help them develop and grow their business.
