Machine Learning
Data Management's Machine Learning tools create predictive models from noisy data. This technology has many practical applications, especially in marketing, where such models are used for predicting future behavior of customers and prospects.
The sub-sections on this page illustrate some use cases.
Create a set of personalized product recommendations for customers
Append preference information (product IDs and attributes) to your customer list. This preference information might be previous purchase history, preferences obtained from questionnaires, surveys, website pages visited, or other data.
Append descriptive product information (keywords) to your list of product IDs.
Configure a Machine Learning Product Recommender tool with your customer list and product list as inputs. You can optionally assign weights to keywords and attributes to prioritize certain criteria (such as latest purchase, most often purchased product, highest ranking survey response).
Run the project to generate a set of product recommendations for each customer ID. Recommended products (for each customer record) are ranked by matching scores.
See the repository project Samples\Machine Learning\Machine Learning Product Recommender for an example of how to use the Product Recommender tool.
Score prospects using existing customer history
Summarize historical information on customers, and generate one or more scores for each customer on metrics such as the recentness of purchase, frequency of purchase, and total monetary value (known as RFM scoring. See Performing RFM customer analysis for an example.)
Match existing customers to a demographic universe and append attributes to customers (such as age, income, company size, or gender).
Use the Machine Learning Trainer tool to train a model in which the demographic attributes predict RFM scores.
Create a set of prospects by sampling a demographic universe and suppressing existing customers.
Use the Machine Learning Predictor tool to score these prospects using the model you created in step 3.
Use the scores to select the best prospects and drive marketing activity and decisions.
See the repository projects in Samples\Machine Learning for examples of how to use the Machine Learning Trainer and Predictor tools.
While traditional machine-learning or statistical-modeling approaches require a great deal of expertise and experience, Data Management's Machine Learning tools can be used by people who have little statistical training or machine-learning experience. You can get good results using the default tool settings.