Targeting & List scoringBackground
Orange Apron is a subscription-based meal delivery service, considering a customer acquisition campaign and plans to run a field experiment to determine appropriate target customers.
To implement the campaign, Orange Apron has rented a list containing information on 500 households. The first variable is a binary indicator of whether children are present in the household (1=yes, 0=no).
The remaining variables are three “hotline” buying indices. Similar to a credit rating, these indices are variables computed by the list owner and represent different index variables that generally indicate positive or negative purchase interest (for different product categories, some indices are positively correlated with purchase interest while other indices are negatively correlated with purchase interest).
In consultation with the list owner, Orange Apron has selected three hotline indices, h1, h2, and h3. Orange Apron’s hypothesis is that h1 is positively correlated with interest in a meal delivery service while h2 and h3 are negatively correlated with interest in a meal delivery service. In addition, Orange Apron feels the presence of children in the household may increase interest in the service.
Orange Apron has sent an invitation to all 500 names on the list to join the service. The invitation offer includes a deep discount on three weeks of service. We observe whether or not each of the 500 consumers accepted the invitation: the value of y is 1 if the person joined the service and the value is 0 otherwise.
I used a random sample of 244 persons as the estimation sample to estimate the scoring model on this data.
The second list of 256 is used to test list scoring and evaluate how successful the target selection was. The 244-person list will henceforth be referred to as the estimation-list, and the 256-person list will be referred to as the holdout-list.
Logit Model
Because the response variable is binary (either buys or does not buy), I built a logit model using a logistic regression on Excel.
Then, I was able to interpret the effects of the coefficients.
Calculating Score, Probability, Lift, and Marginal Effects
Using the coefficients from the logit model above, I calculated the scores for each holdout ID.
Then, the probability was found using the following formula.
Predicting Sales
By adding the probabilities, sales can be predicted. I first sorted the customers by most likely to make a sale to least likely. Then, graphed this against actual sales.
Targeting
Orange Apron estimates the average customer lifetime value (CLV) to be $13.50 and a solicitation cost of $3. Based on the marginal cost rule, I determined the cut-off probability of 0.2222. The cutoff is calculated using average CLV and solicitation cost to get the breakeven rate.
69% of the hold-out list would be targeted as 177 households have marginal probabilities above the cutoff threshold of 0.2222.
Based on the analysis, I found that targeting leads to a profit while soliciting all 256 households leads to a loss.
