Automate ABC Analysis & Product Segmentation with Streamlit π
A statistical methodology to segment your products based on turnover and demand variability using an automated solution with a web application designed with the framework Streamlit
streamlit Application UI
Product segmentation refers to the activity of grouping products that have similar characteristics and serve a similar market. It is usually related to marketing (Sales Categories) or manufacturing (Production Processes). However as a Supply Chaine Engineer your focus is not on the product itself but more on the complexity of managing its flow.
Your want to understand the sales volumes distribution (fast/slow movers) and demand variability to optimize your production, storage and delivery operations to ensure the best service level by considering:
- The highest contribution to your total volume: ABC Analysis
- The most unstable demand: Demand Variability
I have designed this Streamlit App to provide a tool to Supply Chain Engineers for Product Segmentation, with a focus on retail products, of their portofolio considering the complexity of the demand and the volumes contribution of each item.
Understand the theory behind π
In this Article, you can find details about the theory used to build this tool.
Access the application π₯οΈ
Access it here: Product Segmentation for Retail
Step 0: Why should you use it?
This Streamlit Web Application has been designed for Supply Chain Engineers to support them in their Inventory Management. It will help you to automate product segmentation using statistics.
Step 1: What do you want to do?
You have two ways to use this application:
- π₯οΈ Look at the results computed by the model using the pre-loaded dataset: in that case you just need to scroll to see the visuals and the analyses OR
- πΎ Upload your dataset of sales records that includes columns related to:
- Item master data For example: SKU ID, Category, Sub-Category, Store ID
- Date of the sales: For example: Day, Week, Month, Year
- Quantity or value: this measure will be used for the ABC analysis For example: units, cartons, pallets or euros/dollars/your local currency
Step 2: Prepare the analysis
1. πΎ Upload your dataset of sales records
Step 1: upload your dataset of sales records
π‘ Please make sure that you dataset format is csv with a file size lower than 200MB. If you want to increase the size, you'd better copy this repository and deploy the app locally following the instructions below.
2. π [Parameters] select the columns for the date (day, week, year) and the values (quantity, $)
Step 2: select the columns for the date (day, week, year) and the values (quantity, $)
π‘ If you have several columns for the date (day, week, month) and for the values (quantity, amount) you can use only one column per category for each run of calculation.
3. π [Parameters] select all the columns you want to keep in the analysis
Step 3: select the columns for the date (day, week, year)
π‘ This step will basically help you to remove the columns that you do not need for your analysis to increase the speed of computation and reduce the usage of ressources.
4. π¬ [Parameters] select all the related to product master data (SKU ID, FAMILIY, CATEGORY, STORE LOCATION)
Step 4: select all the related to product master data (SKU ID, FAMILIY, CATEGORY, STORE LOCATION)
π‘ In this step you will show at what granularity you want to do your analysis. For example it can be at:
- Item, Store level: that means the same item in two stores will represent two SKU
- Item ID level: that means you group the sales of your item in all stores
5. ποΈ [Parameters] select one feature you want to use for analysis by family
Step 5: select one feature you want to use for analysis by family
π‘ This feature will be used to plot the repartition of (A, B, C) product by family
6. π±οΈ Click on Start Calculation? to launch the analysis
Step 6: Start Calculation
π‘ This feature will be used to plot the repartition of (A, B, C) product by family
Get insights about your sales records π‘
Pareto Analysis
Concept Pareto Analysis
INSIGHTS:
- How many SKU represent 80% of your total sales?
- How much sales represent 20% of your SKUs?
For more information about the theory behind the pareto law and its application in Supply Chain Management: Pareto Principle for Warehouse Layout Optimization
ABC Analysis with Demand Variability
Streamlit App Screenshot: ABC Analysis plot
QUESTIONS: WHAT IS THE PROPORTION OF?
- LOW IMPORTANCE SKUS: C references
- STABLE DEMAND SKUS: A and B SKUs with a coefficient of variation below 1
- HIGH IMPORTANCE SKUS: A and B SKUS with a high coefficient of variation
Your inventory management strategies will be impacted by this split:
- A minimum effort should be put in LOW IMPORTANCE SKUS
- Automated rules with a moderate attention for STABLE SKUS
- Complex replenishment rules and careful attention for HIGH IMPORTANCE SKUS
For more information: Article
Streamlit App Screenshot: ABC SKU split for each family/category
QUESTIONS:
- What is the split of SKUS by FAMILY?
- What is the split of SKUS by ABC class in each FAMILY?
Normality Test
Streamlit App Screenshot: Normality test
QUESTION:
- Which SKUs have a sales distribution that follows a normal distribution?
Many inventory rules and safety stock formula can be used only if the sales distribution of your item is following a normal distribution. Thefore, it's better to know the % of your portofolio that can be managed easily.
For more information: Inventory Management for Retail β Stochastic Demand
Build the application locally ποΈ
Build a python local environment (recommanded)
Then install virtualenv using pip3
sudo pip3 install virtualenv
Now create a virtual environment
virtualenv venv
Active your virtual environment
source venv/bin/activate
Launch Streamlit π
Install all dependencies needed using requirements.txt
pip install -r requirements.txt
Run the application
streamlit run segmentation.py
Click on the Network URL in the shell
-> Enjoy!
About me π€
Senior Supply Chain Engineer with an international experience working on Logistics and Transportation operations.
Have a look at my portfolio: Data Science for Supply Chain Portfolio
Data Science for Warehousingπ¦, Transportation π and Demand Forecasting π