Cross-Product Association

Author: Robin Jia，Tommy Lin

Last Update：2024/06/14

 

1. Model Introduction

The Cross-Product Association model focuses on analyzing the associated purchase behavior and trends between different products. By examining the situation where users purchase other products at the same time when purchasing specific products, product combinations suitable for bundled sales can be identified, thereby optimizing product bundling recommendations, cross-marketing, and other strategies. This model supports custom parameters such as analysis time range, product scope, etc., flexibly adjusting the analysis perspective in combination with actual business needs.

Through this model, the following business questions can be analyzed:

• Which products are more likely to generate cross-purchases with other products?
• The purchase of specific products usually drives the sales of which other products? How strong is the association between different products?
• After browsing which products do users usually choose to purchase other products of the brand?

 

2. Model Interpretation

This model mainly consists of three parts. The insight part at the top shows the two ASINs that generated the most associations within the selected time range under the filtering conditions. The associated sales part in the lower left shows the ranking of ASINs that generated the most associations with other ASINs within the selected time range under the filtering conditions. The relationship graph in the lower right shows a visualization chart of other ASINs that have an associated relationship with the selected ASIN.

In the model's top menu, the predefined model is for fixed Amazon users, all Amazon ad types, and all ASINs under the store. You can select a time range by month. The association granularity option allows switching each association item to ASIN, parent ASIN, or custom classification. Switching this option can change the default association between ASINs at the bottom to the association between parent ASINs or the association between any custom ASIN combinations (brand/product line/model, etc.).

The association type option supports switching between three different association methods, allowing multiple selections, with the default being all selected.

• Simultaneous: The user purchases two different products in the same order
• Subsequent: The user purchases two different products in two separate orders
• Alternative Purchase: The user views the detail page of one product first and then purchases another product.

To use the association granularity function, you need to first upload a relationship table between ASIN/parent ASIN/category. Click the upload button on the right side of the association granularity option to open the popup window. The popup window includes an Excel template for uploading.

The insight part shows the two ASINs/parent ASINs/categories with the most associations under the current time range. Here you can filter the association type to see the results under different association types.

The Top 20 Associating Selling part shows which ASINs/parent ASINs/categories have the most associations with other ASINs/parent ASINs/categories under the current conditions. This can also be used as a reference when filtering.

The relationship graph part uses a more intuitive visualization method to depict the association structure between products. You can see the associations between the selected ASIN/parent ASIN/category and all other ASINs/parent ASINs/categories. The size of each bubble represents the number of associations. Hovering the mouse over each bubble shows the represented ASIN/parent ASIN/category and the number of associations. Here it is recommended to first filter the association method using the association type filter option at the top before viewing. This will also include the percentage of the number of associations out of the total number. The download button in the upper right corner can download the data table of the model.

The downloaded data sheet will only include data for the ASIN/parent ASIN/category selected in the graph. If you need to download more data, please use the Edit - Filter - ASIN/Parent ASIN/Category options to filter the data before downloading.

Click the edit option to expand the sidebar. Here you can perform the following operations:

• Select the metric corresponding to the size of the association node
  • Number of Associations
  • Associated Sales
• Filter based on conditions
  • Number of Associations and Associated Sales: Show data within the filtered range
  • ASIN and Associated ASIN: Filter out the desired ASIN or associated ASIN

The Cross-Product Association model needs to select an ASIN/parent ASIN/category as the root node to view the associated nodes. The root node by default selects the ASIN/parent ASIN/category with the most associations under the Instance. The filtering function can be used to select the association data of a specific ASIN/parent ASIN/category.

 

3. Model Customization

This model supports Amazon Purchase Data Insights and supports customizing associated ASINs. After setting, only the conversion data of the set ASINs will be counted.

 

4. Model Data

The core data of the Cross-Product Association Analysis model includes: the order quantity and sales amount of different product associations. This involves the statistical caliber of associated purchases and the interpretation of analysis data, mainly including:

Data lookback time period:

• Predefined model data is the purchase data generated each month. For the association types of Subsequent and Alternative Purchase, the first browsing/purchasing behavior will be traced back to the previous three months at most.
• Custom model data is the purchase data within the user-selected time range period. For the association types of Subsequent and Alternative Purchase, the first browsing/purchasing behavior is only traced back within the user-selected time range.

Differences in association metrics:

• Number of Associations: Reflects the popularity of product combinations, but is easily affected by the sales volume of the products themselves.
• Associated Sales: Reflects the income level brought by product combinations.
• Association percentage: Reflects the proportion of the current association among all associations of the same type.

Comprehensively considering various types of indicators can provide a more comprehensive and objective understanding of the association effect of products.

Precautions for data interpretation:

• Association strength and association scale are not always consistent: Some niche product combinations have a high association rate but few purchasers. When formulating bundling strategies, it is necessary to weigh the pros and cons.
• Some products may generate associated purchases due to specific marketing activities. When analyzing trends, it is necessary to consider the impact of marketing factors.
• Cold start problem of new products: There is less association data for new products when they are first launched, and it is necessary to give a certain observation period during analysis.

When counting the two scenarios of Bought After Viewing and View After Buying, there are certain differences in the criteria for defining two associated ASINs.

• In the Alternative Purchase scenario, assuming the user behavior chain is View A->View B->Buy C, the model will count the Alternative Purchase association of AC and BC.
• In the Subsequent scenario, assuming the user behavior chain is Buy A->Buy B->Buy C, the model will count the Subsequent association of AB and BC. AC will not be counted as Subsequent.

In summary, to fully utilize the insights from cross-analysis, it is necessary to flexibly interpret the connotation of the data and comprehensively consider the impact of external factors. By fully understanding the caliber and characteristics of the data and eliminating interfering factors, the true association rules between products can be accurately mined.

Glossary

Type	Term	Description
Dimension	ASIN	Refers to the first ASIN in a relationship, which could be one of three types: the initial ASIN in a consecutive purchase sequence, the viewed ASIN before a purchase, or one of the ASINs in a multiple-ASIN order.
	Associated ASIN	Refers to the subsequent ASIN in the relationship, such as the second ASIN in a consecutive purchase, the ASIN purchased after a view, or another ASIN in the same order.
	Association Type	The type of association includes purchasing one after another, purchasing after DPV, or purchasing multiple ASINs in one order.
	Association Granularity	Determine the association between ASIN/Parent ASIN/Product line.
Metrics	Number of associations	Number of associations.
	Associated sales	Sales summation of both associated ASINs.