Machine Learning and Deep Learning for Retail Business

Retail businesses sell items or services to customers for their consumption, use, or pleasure. They typically sell items and services in-store but some items may be sold online or over the phone and then shipped to the customer. Examples of retail businesses include clothing, drug, grocery, and convenience stores.

Different types of Retailers and their offerings:

Retail Value Proposition:

Retail Innovations in Machine Learning:

Machine Learning is used to analyze the previous as well as current data in order to make decisions regarding business. It is used for predicting supply and demand.

Purpose:

1. Increase revenue

2. Reduce Cost through automation

3. Impact both cost and automation

Applications in Online Retail:

Digital catalog browsing gives ability to search visually.

- Gives quick product info

Recommendation engines for speed up the searches — e.g. Similar products

Pricing Strategy: Competition pricing, similar products, collecting information from users buying history

Automated price system will adjust to external events such as supply demand, season ability etc.

Social media: helps brand to expand and reach to customer and customer monitoring.

• analytics department will extract information from social media campaign about user demand and choices, target audience from social media posts of users as well as campaigns run by company.

Chatbots: they help customers to get the solution or they provide way to reach till solution. They work as smart assistant

Contact center: If anything beyond chatbot or any issue arises then user can contact center for human assistant.

Applications in Brick and Mortar:

1.Predicting Inventory needs:

Profitability of these business mostly depend on logistics. Any issue with supply demand leads to losses. Season-ability, Environment issues such as disaster, epidemic, lets consider latest case of Corona Virus will also change customer behavior towards businesses.

Any supply or demand issue will affect Logistics which are related to Warehouses, transport etc. Delivering Products from Warehouse situated far away from customer, when nearest warehouse doesn’t have sufficient goods leads to loss.

Such Scenario needs predictive, prescriptive analysis and machine learning or deep learning solutions to guide business about supply demand and make predictions.

Computer vision can be used in shops. With the help of CCTV as well as sensors users can be tracked. One can track their behavior towards product and get idea of products in demand. Video analytics is very helpful here.

One can analyze navigations routes in malls. It also helps in theft prevention.

Virtual mirros:

Current state of Analytics in Retail Sector:

1. Descriptive Analytics: Product wise sales report

2. Diagnostic Analytics: Find Root cause of Problem

3. Predictive Analytics: Outcome in future, Projecting budget (data analysis)

4. Prescriptive Analytics: Report from data analysis processed through Machine learning algorithm for further pinpoint analysis.

Machine Learning in Retail can:

1. Find Retail consumers buying patterns.

2. Help managing consumers.

3. Supply chain management improvement.

4. Predicting future scenarios.

5. Helps business to take decisions.

Challenges in Machine Learning in Supply Chain Management:

Need of lot of Data:

Suppliers take time to calculate the optimum inventory

Data sources may be different and need single system to gather data and organize the data , even we need historical data

Data organization includes combining data from multiple formats, cleaning them, correcting data if needed

Quality and accuracy of data is questionable as it came from multiple sources e.g. procurement system, transport management system moreover traffic , reasons such as Corona virus, Socio-political stations affect logistics and then data.

If supply chain system is not there, then we need to build it from scratch.

Implementation of machine Learning:

1. Unsupervised learning for Data mining and learn buying pattern

2. Creating prediction models based on discovered dimensions

3. Use of Supervised learning on these dimensions

Terms to be looked after:

Inventory:

A) Cost :

1. Prices

2.Prodcuts

3.Unit Pipeline Cost

4.Donations

5.Unit Penalty Cost — this occurs if goods not delivered on time

B) Time:

1. Average Transportation time

2.Average time goods stored in the warehouses

C) Sales:

1.Projected sales unit for a particular item in a stock

2.Stock review intervals

D) Predictions:

1.Projected goal for arrival time interval

2.Recommendations on the unit of inventory that needs to be ordered based on transportation time

3.Inventory pipeline (in time delivery)

Faulty Machine Learning techniques:

1.Event based automation

2.Blind automation without any analysis

3.No prechecks before automation

4.Opting for automation when its needed less than 50%

5.Bad implementation

Correct Machine learning:

1.Rule Based Automation

2.Analytics based automation

3.Lots of pre-checks

4.Opting for automation when need is more than 50%

5.Good implementation (when machine learns quickly than human)

6.Use of Internet of technology

7.Connected Stores and warehouses with technology

Case For Deep Learning:

Machine learning is in demand in retail sector but it needs domain expertise as well. However, Deep learning can be implemented for image processing (CNN )as well as Time series prediction can also be done through deep learning (LSTM networks).

One can implement Genetic Algorithms in Deep learning for making decisions.

Deep learning doesn’t necessarily require domain expertise. It is also better than machine learning in many fields now. It takes care of Math and results/predictions are more accurate than machine learning.