Summary Hi everyone, I would like if the next text is correct. Thank you! "After each class I will write a recap. I will try also to give an answer for all the questions; please, if you think that your question wasn't solved in the class time, please send me an email and I will work to give you a satisfactory answer. A. The class A1. The quiz There is some important concept which is necessary in this course. - The importance of Business Step - What are the most important variables? - What is the target in each situation? - The two samples (100 sample, 100K sample) - Usually, the algorithm not is so important that the variables. A2. Cyclical projects The projects usually have done in full subphases like business definition, uploading the data, auditing the information, creating variables, modeling, etc. Sometimes it isn't the best methodology because if you made a mistake in the first steps then the time of the project will become your worst enemy. That is way is better a cyclical methodology. The project can be done in partial subphases. Business definition (what are the most important variables? What is the target?), Uploading the data (only the most important data), Auditing the data (only the tables which has the information that you thought in your Business Definition), ... This is the first iteration. In the second iteration you can attach more variables and more tables. A3. Some supervariables - Previous cancelations - Previous purchases - Variation of the transactions - Financial Profile - Share of Wallet B. Your questions B1. The target in a card propensity model with Call Centre as channel. There are two ideas, the first one, it is necessary you have built the target with the inbound calls, because this set of calls has the information of the clients which has a propensity to purchase a card in a call. BUT, sometimes, the history of your data could be biased. So, you can imagine that the call centre always has called to certainly kind of customers. In this case, is also recommendable to take the data of the clients which has purchased cards in other channels. And to make two populations or two set of rules for each dataset. Thank you,