Information for: DEVELOPERS   PARTNERS   SUPPORT

Market basket analysis

Market basket analysis is a technique that looks for the relationships and associations between two objects derived from their purchasing patterns. The scope of market basket analysis is beyond the retail analytics from which its name is derived. It is the analysis of any collection of items to identify affinities that can be used in some manner.

With this feature, you can observe all combinations of product categories that are bought together. You can derive the combinations that frequently co-occur. For example, of the total transactions with bread, 10% of the time milk was also bought along with bread. Therefore, bread and milk are correlated product categories in a basket. Combined revenue from bread and milk is $xx. You can do a similar analysis at product category level. For example, dairy products were present in 30% of transactions that had cereal products.

Use cases

This feature is intended for marketers who want to:

  • Improve brand loyalty by:
    • Upselling or bundling of SKUs of the same category.
    • Cross-selling or bundling of SKUs of different categories.
  • Use product correlation data to improve inventory management. You can accelerate the sale of less-selling products with relevant combo offers.
  • Improve profitability and formulate pricing by identifying high-selling baskets with higher profitability.
  • Analyze the top trending baskets based on product categories, brands and other product meta data in addition to the existing product name dimension.
  • Quantify the occurrence of products in multiple baskets by no. of associations metric.
  • Analyze the combinations in baskets with multiple granularities like country, state, zip, and so on.
  • Understand the revenue generated by each basket with support, confidence and lift metrics

Prerequisites

To use this feature, you must have the transactionsummary, productcategoryxref, and productcategorysummary tables surfaced in IQ. If you do not have these tables surfaced in IQ, contact your CSM or AM to find out if you are eligible for such a configuration.

Using dimensions and measures

CDP offers multiple dimensions and measures that you can use to generate market basket analysis report. For more information on the standard market basket analysis dimensions and measures, see dimensions and measures respectively.

Standard dashboards

In addition to the customized market basket reports that you can create, CDP provides an out-of-the-box (OOTB) Market Basket Analysis dashboard.

To use the standard Market Basket Analysis dashboard:

  1. Sign in to your CDP user interface.
  2. Click Analytics > Metrics > Metrics Launchpad.
  3. Select D10 - Market Basket Analysis.

Feature description

The features of market basket analysis are:

  • Report Type: Filters the reports based on the following categories:

    • Transaction Level Analysis: This metric calculates product correlations for all the products that are bought together, and excludes single product transactions. It calculates the trending basket for one transaction at a time. You can select this if you are looking for insights around discount or pricing.
    • Customer Level Analysis: This metric calculates product correlations at a customer level. It calculates the top baskets across transactions grouped by the customer. You can select this if you are looking for insights around upsell or cross-sell campaigns or recommendations.
  • Regional Filters: You can apply city, country, state, store and zip code filters to all tiles. It helps you to understand the hyperlocal differences and preferences of customers. You can design discount strategy, product campaigns with a more hyperlocal context.

    The available options are:

    • Filter By City
    • Filter By Country
    • Filter By State
    • Filter By Store Type
    • Filter By Zip
  • Number of associations: Number of associations for a product or category is defined as the number of other products or categories that were bought along with this product. For example, if white bread was purchased along with milk, eggs, juice, and meat, then the number of associations for white bread is five. You can identify the product or category with a high number of associations and use it for combo offers with unsold products.

  • Support, Confidence and Lift metrics: This metrics helps to identify which combinations or baskets

    have statistical significance and use it for promotional campaigns and other use cases.

    • Support: It is the ratio of the transaction count of the corresponding basket
      to the total transaction count during the specified time interval. It quantifies the significance of the combination or basket to the total transaction volume. Higher the support, the more frequent the occurrence of the basket in the transactions.

    mba_support

    • Confidence: With market basket analysis, we define a basket as two products bought together. Confidence specifies the probability of the second product bought given the first product was bought. For example, consider that apples were bought in 100 transactions during a time period. Of these transactions, if bananas were bought together with apple in 20 transactions, then the confidence of bananas being purchased with apples is 20/100 = 20%.

    mba_confidence

    • Lift: Consider two products {X} and {Y}, lift quantifies the enhancement {X} causes to {Y} being in transaction count. In cases where {X} actually leads to {Y} on the cart, the value of lift will be greater than 1.

      For example, a user buys bread and milk together. Transaction count with bread as the only product in the basket is 14, the transaction count with bread and milk in the basket is 10, the transaction count with milk as the only product in basket is 80 and the total transaction count is 100.

      Now the Confidence of milk being bought when bread is bought is 71%. However, if you calculate the lift that bread provides to milk being in the basket is 71%/80% = 0.88. This value is less than 1, which means that bread being in cart with milk actually reduces the probability of milk being in the basket from 0.8 to 0.7. Thus, more the value of lift, greater are the chances of preference to buy {Y} if the customer has already bought {X}.

    mba_lift