Author: Travis Munn, Senior Application Developer
What is Market Basket Analysis?
Market Basket Analysis (MBA) analyzes Point of Sale data to identify products that are frequently purchased together or in sequence. This analysis creates “baskets” of the item combinations and calculates the strength of the item correlations in order to equip organizations to predict customer purchasing habits, analyze the effectiveness of strategic initiatives (such as product placement, advertisements, or discounts), and ultimately increase their revenue.
Why perform Market Basket Analysis?
Due to processing limitations, more traditional methods of POS analysis are difficult to scale and limited to analyzing only a subset of the data. Utilizing a modern data architecture instead, such as the Hortonworks platform, to perform Market Basket Analysis allows for the entirety of a dataset to be processed in a small fraction of the time, with less reliance on analysts or IT resources.
A few benefits include:
Utilization of Spark interpreters in Zeppelin (an Apache notebook tool) allows for analysis over the entire collection of data, rather than a subset
Improved efficiency of the entire process to calculate the market baskets; Results of large data sets containing hundreds of millions of rows of sales data can compute in less than 10 minutes
The flexibility of the design of the data ingestion processes allows for additional data points or data sources to be added in the future with minimal impact on the design
Self-service analytics; Zeppelin notebooks provide an interface that analysts can interact with directly to create custom reports and visualizations on the data while taking the burden off of IT resources. It could also integrate with one of your existing reporting tools
Incorporation of a variety of data formats; Combining multiple sources of data to the process can allow analysts to explore the data at a more granular level
What decisions can be made from the output?
After uncovering the relationships between product purchases, you can begin to gain further insight into customer purchasing patterns, compare results across all of the variables within your organization (suppliers, locations, employees, etc.), determine the effects of marketing initiatives or promotions, and so much more.
For instance, let’s assume you have a chain of hardware stores with multiple locations. Your Chicago location’s sales are lagging and you want to make some adjustments to try to get the revenue projections back on track.
You have another store in a cold market in northern Minnesota. You identify some products that are selling much better in Minnesota than in Chicago and then compare the Market Basket Analysis of those products across the two locations.
You notice from the output that the Northern Minnesota store sells a statistically significant amount of snow shovels and heavy-duty gloves together, but in the Chicago market, that correlation does not persist. The shovels are selling well, but rarely do the customers also purchase a pair of gloves, and your glove sales are well below the sales target.
How can you act on this information to meet your revenue goals in the Chicago market?
You pair this POS data with weather data to see that shovel sales consistently spike immediately after a snowstorm. You also compare the store layouts and realize that gloves are in much closer proximity to the shovels in Minnesota than they are in Chicago.
Armed with that information, you first create a glove display near the shovels.
You also decide that during the next snowstorm, you will send out an email promotion on shovels to your Chicago-area customer distribution list. The customer base that already has a higher propensity to purchase a shovel will now be enticed to choose your store over the competitors, and will hopefully pick up a pair of gloves while they’re at it.
Are you interested in a learning more about how Market Basket Analysis can benefit your company?
Contact us to see a demo and get more information about how our experts can improve your analysis today.