Propensity modeling is used to predict the causes and implications of behaviors. PSM tries to avoid bias by creating a balance between the two groups and in our case, our current client and non-client.
The balance is created when we add other independent variables (covariates) such as age, gender or product use, geographic, ethnicity.
We think that restaurant owners with a similar propensity score resemble each other. In our business, propensity scoring can identify potential new clients. If an owner has a propensity score similar to another owner who is already a client, that they too will become a client. In this way, we can identify potential clients in non-client populations who are likely to make a purchase given the correct stimulus.
Propensity Score
The propensity score is the probability that a subject performs an action. The action to make a purchase and become a client.
PSi = P(A=1|Xi)
Where A=1 is the probability that a person ends up in a group of buyers given a set of covariates (independent variables) X.
Suppose that the restaurant age is the only X variable and that older restaurant age is more likely to buy your product. Then, the propensity score is larger for old restaurants. If a restaurant has a PS of .30, this means that the restaurant has 30% chance of being in a group of buyers given a set of covariates and not a good candidate.
Reference
Propensity modeling for business & Targeted Maximal Likelihood