SAP Analytics Classification algorithm: Predict the potential profits of marketing campaigns

Profiling the client’s attributes that influences the positive response to sales

Classification is one of the Machine Learning algorithms of SAP Analytics cloud to find what are the variables that have more influence to get a positive result.  Based on this information we can calculate what is the maximum profit of a marketing campaign. The algorithm takes past information and evaluates the results to give how good is the data to do predictions.

We are going to train the Classification process with the data of a past marketing campaign to know the possible effectiveness percentage of a campaign based on:

1) the most influential attributes of the clients that buy our products

2) the optimal number of contacts based on the cost of reaching them

3) optimizing the return of investment by giving the profit we expect for each sale.

The Classification process algorithm evaluates itself to give information of prediction effectiveness with new datasets.

The test input dataset has these clients’ attributes in figure 1.

Figure 1: Client’s characteristics and the target variable is positively Accepted

From these variables like housing, loan, age in figure 1, the Classification will process will determine what are the most influential to get the positive sale result indicated by the variable ACCEPTED.  With this information, the algorithm evaluates the dataset to measure its effectiveness to do predictions of new data and calculate the optimal profits with the minimum of contact costs.  The Classification shows that we don’t get the optimal return of investment by doing more contacts as demonstrated in figure 2.  Look that the optimal number of contacts is 744 before the profits start to decline.

Figure 2: Profit simulator to find the optimal number of contacts for the highest profit

From the data of figure 1, the Classification process says that it has the following forecast capability in figure 3:

1) Predictive Power: the model auto evaluation using its data.

2) Prediction confidence: the model trust with new data.

Figure 3: Confidence of the data model to do predictions

From figure 3 we see that the model auto-evaluation after using the data for training has a percentage of confidence of 67.88%. The prediction confidence indicates the reliability of the model to do predictions using new unknown data.  Meaning that we can load new data campaigns to the model, and we will get a 93.04% of assertion.

The most influential characteristics of the clients to get a positive sales response are in Figure 4:

Figure 4: The variables with more influence to get a positive response in Accepted the sale

You can see from figure 4 that the clients with the high-profit outcomes and house owners’ characteristics have more probability to accept a sales proposal from us.

The potential profits of marketing campaigns  using the Sap Prediction Classification Algorithm Based on:

- Cost of each positive contact of 15 dollars.

- Expected revenue for each sale of 25 dollars.

are 5350 dollars contacting a number of 35 potential customers?  If we contact 851 potential new customers, our profits do not grow, instead, they decrease to 5275. This scenario is in figure 2.