Based on our experience at Kreato on applying ML techniques, we see that predictions can be utilized on CRM for the below use cases:
-
Predicting the profile-fit or the business match of incoming leads
-
Predicting the win likelihood of prospects
-
Predicting the closure time of progressing deals
-
Predicting the closure time of progressing deals
Data (Feature) Selection
Quality of the base data (feature selection) forms the core basis for prediction model performance. Low-quality features (profile attributes) have to be identified and removed. Also features with more missing values can be avoided.
Data Cleansing
Selected features missing values has to be cleaned using the median or any custom cleaning functions to further improve the performance of the prediction model.
Statistical Filtering
Not every feature that has been selected may add prediction value to the model. Hence any feature that doesn’t provide significant predictive strength has to be identified using statistical methods and have to be filtered out.
Once we are done with the above steps, based on the use case algorithms such as logistic regression, two class or multi-class linear regression can be employed on the pre-processed data for prediction.
