How Can Market Basket Analysis Help E-Commerce?
Nearly around 45% of the E-commerce customers purchase more than one product at a time and E-commerce users are growing overtime.
Therefore, if we can analyze the purchase pattern of customers and identify commonly purchased product combination together, revenue can be improved by around 35% without spending money unnecessarily towards Facebook or Google Ads Campaigns. By analyzing these patterns, we can also implement a more user-friendly website to the visitors to make purchases efficiently. One of the key methods to uncover the hidden strong relational between product combination is Market Basket Analysis.
Why Market Basket Analysis is so important for your business?
- Product Placement — Identifying relational products that have been purchased together and arranging the placement in the same product page to encourage customers to purchase both items.
- Customize Promotions — identifying the similar products that have been purchased together and give some bundle offers.
- Facebook and Google Campaigns Plan — Target the customers who purchased only one product from Association products and show them other related product.
- Personalize email campaigns — show the frequent products that customer viewed and products that customer did not buy from Association product.
- To improve user experience.
Market Basket Analysis provides two product sets.
- Frequent Product Set — Product items that have been mostly purchased in all transactions.
- Association Product Set — Strongly correlated commonly purchase products.
How to obtain Frequent and Association Product Sets
There are three measurements (Support, Confidence and Lift) in Market Basket Analysis that we should calculate to obtain the product sets. From below example, let’s get an understanding about the three measurements.
If we take above example, product Apple has been purchased within 5 transactions out of total 8 transactions. Therefore, support value for product apple is 62.5%
If the support value is high, it means that this particular product is purchased within most of the transactions made by customer. And those high support value products are defined as frequent products.
If a customer already purchased product apple there is a 75% of probability for them to purchase product beer.
When it comes to confidence value the range of two products can be 0% to 100%.
- If Confidence(X-> Y) = 0% indicates that this customer will never purchase product Y along with product X
- If Confidence(X-> Y) = 100% indicates that this customer will definitely buy product Y along with product X
As explained above there is a 75% chance to purchase beer if apple is already purchased. But we need to see if this relationship is occurred randomly or not. To clarify this, we need to calculate lift value.
When it comes to Lift value the range of two products can be 0 to infinity.
- If Lift(X->Y) < 1; implies that X can replace from Y and vice versa.
- If Lift(X->Y) = 1; implies that X and Y are two independent products
- If Lift(X->Y) > 1: implies that X and Y have a strong relationship and also both products can bundle together.
In our example, Lift value of Apple and Beer is 1.2 So, we can conclude that there is a strong relationship between them, and we can promote those two products by bundling together.
Since it is not practical to calculate mentioned measurement manually for each item and item sets, we can use Python or R programming Language to download the Google Analytics Data and ingest our Data Lake and make the transformation as required, insert into tstaging and staging table and calculate and identify the frequent and association product sets.
Results
We have used transactional data of one of our clients for Market Basket Analysis and we found interesting relationship between products that customers bought. I have shared some of finding results as below.
Case 1:
If we take above results, there is a higher chance to purchase dupatta if the customer has already purchased the Kurta and vice versa.
we identified that customers are more likely to buy exactly same products (dupatta) wore by the model along with Kurta.
Recommendation
- Showcase associate products together on the website. — If a customer comes to the Kurta page show them the relevant associate dupatta product and vice versa.
- Make sure to keep stock available for both associate products. — — If an associate product goes out of stock the probability of purchasing the other related product might be lower.
Case 2:
If we consider above results customers who have purchased accessories such as earrings and mask-chain, there is a higher chance to buy a similar product or the same product with a different color.
Recommendation
- Showcase similar products together on the website.
- Bundle the similar products and provide offers.