Discount

As explained in Causal Inference for Data Science

A discount is a reduction from the standard price offered to a customer.

A discount reduces the price offered to a customer. Choosing an effective discount requires separating what predicts a purchase from what would change purchasing behavior. Historical purchase frequency or loyalty can affect both the discount received and the likelihood of buying, making it a confounder. section 3.3.2 section 4.1.4

Evaluate a discount strategy by checking:

• Whether large and small discounts reach both frequent and infrequent customers; otherwise, the positivity assumption may fail. section 3.3.2
• Whether the observed relationship between price and purchases reflects the discount effect or customers’ pre-existing loyalty. section 3.3.2
• Whether a predictive model still applies after changing who receives discounts. section 4.1.4

Changing the discount policy

Suppose an online retailer historically discounts mainly for loyal customers. A model using price alone may predict purchases well because price indirectly signals loyalty. If discounts are extended to occasional shoppers, that relationship changes, and the model may overestimate their purchase probability. Including loyalty and assigning both discount levels across customer groups supports a more reliable assessment. section 3.3.2 section 4.1.4

Figure 3.14 Graph representing pricing discounts based on historical purchasing frequency

Trace how historical purchasing frequency influences both discount assignment and purchasing, creating a confounding path. section 3.3.2

How Causal Inference for Data Science covers it

How deep it goes

Contrasts predictive and causal discount models, emphasizing confounding, positivity, random assignment, and policy shifts.

Examples

Uses digital-platform pricing, loyal versus occasional customers, purchase frequency, price, and purchase probability.

What you can do after

Evaluate discount policies by distinguishing predictive accuracy from causal effects and checking whether customer groups receive comparable treatments.

How it connects

Connects discounts with causal inference, confounding, the positivity assumption, randomized experiments, and supervised learning.

liveBookmanning.com

Read Causal Inference for Data Science on liveBook Get the book at manning.com

Related concepts

• supplierused by · A supplier is a person or organization that provides goods, products, services, or other resources to another party.

This explainer was generated by Manning's knowledge graph from the text of Causal Inference for Data Science. The section links point to the parts of the book it draws on.