Product Information
How can Smart Predict help you to optimize your marketing initiatives?
Taking accurate business decisions and keeping control of costs can be a challenge for companies.
Thanks to Smart Predict, you can easily answer this common business question.
Smart Predict is an SAP Analytics Cloud feature that helps business analysts answer business questions about future trends. You are guided step-by-step to easily create a predictive model, using historical data to make trustworthy future predictions presented in simple views.
Let’s take the following example: Imagine that you are working for a sales department in a Bank. You’ve had a poor return on your recent offer to sell new banking products. Your manager has requested to send a new offer to your customers to promote a new type of deposit account, but you have a limited budget. So, you want to target only the customers that are the most likely to purchase this product.
Based on this example, you’ll see in 5 steps how you can build an accurate predictive model and control your costs to respect your budget constraints.
At the end of each step, you can find additional resources to explain the steps more in details.
Step 1: Collecting your data
The first thing you need to do is to collect your data. Behavior of customers can be predicted by looking at past trends in a similar situation. So, you might go to your customer data system and download information about your customers, including how they react on the past offers of similar products.
You will organize these data in a table format that will be your training dataset. The column that contains the information on how the customers react on the last offer will be your target variable.
Here is an example of what your dataset could be:
Once your dataset is ready to use, you need to upload it to SAP Analytics Cloud so that you can use it with Smart Predict.
Click to go further:
Creating a dataset from a local file
Step 2: Build the most reliable predictive model based on past data
Now that your historical data are ready, you can start creating your predictive model.
But what does it mean exactly?
Smart Predict will analyze the past data and identify which variables have influence on the customers’ behavior. As a result, it will generate a predictive model which is able to classify your customers into 2 categories: the ones interested in your offer that you want to target, and the ones who aren’t interested.
Once your predictive model is ready, you can evaluate the accuracy and robustness of your predictive model using a range of reports and performance indicators. You can also see which variables most impact your customers’ decision.
Tips:
You can conduct several experiments such as adding data to your dataset or changing the predictive model’s settings, until you get the best predictive model that best fits your business.
Click to go further:
Building the best classification predictive model
Analyzing the results of your classification predictive model
Step 3: Calculate the return on investment using your predictive model
Your predictive model is ready to identify customers who would most likely open a deposit account once they receive your offer. But now, you need to answer the second part of your business issue: you have a limited budget.
Even if you have created an accurate predictive model, it’s not 100% perfect and could have incorrectly identified a few customers as those “to contact with your offer”. So, you need to define a cut-off point (also called a threshold) that allows you to reduce the number of these incorrectly categorized customers. The confusion matrix can help you determine this cut-off point.
Your threshold is now determined, and you can evaluate the costs and profits using this predictive model. Thanks to the profit simulation tool in Smart Predict, you can assign a cost per contacted customer and estimate your profit per customer who opens a deposit account.
Tips:
Use the Maximize Profit option to see the threshold to get a maximum profit based on the profit parameters that you have set.
Click to go further:
Understanding the confusion matrix
Step 4: Use your predictive model and get your predictions
Now you are ready to use your predictive model on your actual customer data to get your predictions.
To do so, you need to prepare another dataset, called an application dataset, with the actual customers you intend to contact with your offer. This dataset must contain the same type of information and structure as the training datasets:
- Same number of variables (additional columns will be ignored)
- Same variable names as the training dataset
And of course, the column which answers the question “customers to target” is empty because it is what you are looking for.
Once again, you need to upload this dataset to the Files section of SAP Analytics Cloud so that you can select it in Smart Predict.
The last step is to define the predictive outputs you need for your business. In our case, what fits best for this scenario is the prediction probability.
Click to go further:
Applying a classification predictive model
Applying a classification or regression predictive model
Step 5: Present your results in a story
You might need to present the results to your management. Why not use other SAP Analytics Cloud features and make a beautiful story, highlighting what your predictive model has found?
Click to go further: