Targeting & Recommendations
9 important questions on Targeting & Recommendations
What is a house file (characteristics)?
- Contains transactions and customer data
- Already contains the "best" customers for the firm
- Far more responsive than prospective list of customers
What determines the profitability of a mailing or recommendation?
- Cost of contact
- Revenue if customer buys
- Opportunity cost
- Response probability
How can we estimate the response probability and the expected revenue/margin?
- Higher grades + faster learning
- Never study anything twice
- 100% sure, 100% understanding
Three steps in the targeting process:
Step 2: Validation sample (score, sort and decide which customers to target, validate on customer characteristics and choice behavior)
Step 3: Test sample (score customers and send out mailing/offerings, based on customer characteristics but not on choice behavior)
Similar approach to recommending techniques:
- Predict likelihood of purchase for each customer, and expected profits
- Pick most profitable products to recommend
Commonality Index can be used for several data:
- Behavioral data
- Geographical location
- Past purchases
Intuitive ways to determine how similar products are:
- Analyse the description (or reviews) of each product and see similarties
- Look at the ratings given to products
- Look at classification fort the product (e.g. for movies/genre/actors)
Popular methods for recommendation engines:
- Random forests
- Logistic regression
- Neural network
How do you obtain insights into causal links with recommendations?
The question on the page originate from the summary of the following study material:
- A unique study and practice tool
- Never study anything twice again
- Get the grades you hope for
- 100% sure, 100% understanding