Many past webinars have focused on the technical details of Rapidminer products—but last week’s webinar put into context what it takes to put models into production to deliver business outcomes using RapidMiner Studio and RapidMiner Server.
In a recent webinar, Derek Wilson, president and CEO of CDO Advisors LLC, looks at three different use cases to help you use RapidMiner to solve business problems related to call center optimization, including agent churn, customer churn, cross-selling, and ultimately how to put predictive models into production to lower costs and increase customer satisfaction.
Agent & Customer Churn—RapidMiner Studio to RapidMiner Server
Building an accurate predictive model is important—but the real goal is making the model outcome actionable. The outcome of this model will help you discover which attributes in agents/customers lead to high attrition. That information can then be translated into agent training that weeds out candidates with attrition attributes and retains the best agents or changing processes that lead to customer attrition and switching services to a better fit for customers.
First, input agent performance, call center stats, HR data or customer CRM data and operations data. Output that model to your call center management team and into a decision tree so that the results are comprehensible to those who are not necessarily data scientists.
Next, you pull in RapidMiner Server so that you can operationalize and focus in on key attributes over periods of time without having to sit there and manually run your model in Studio over and over. Simply deploy your RapidMiner Studio model into RapidMiner Server, run a query that goes directly against the database and when you automate it you can choose to drop information out and build reports on top of it.
You can schedule the model, save the results and provide reports to your management team. Incorporating Server takes your models to the next level where your business questions get answered with action. Above, see the differences between the agent churn workflow basic versus the advanced workflow after Server comes into play.
Can you predict which customers are most likely to respond to a marketing campaign? All you need is information on the responsiveness of past customers to apply to an active campaign list. The input would be the same as for your customer churn model. If the output aims true, (prediction of which customers are most likely to accept a new product), you will be able to target customers for products and consequentially limit your budget to those customers with 100% accuracy.
Let’s build a financial cross-sell and output our files to the marketing and sales teams. RapidMiner Server saves the day again when it lets you create automations and build associations rules after publishing your RapidMiner Studio models to it. Above, we are looking at the lift association rule built in RapidMiner Server for a model that concludes that if a customer has an auto loan, they should have a home loan. Information like this can help you create campaigns that target people within different subsets or parameters which can help the marketing team to make their campaigns as specific and strategic as possible.
You can publish the output of your model in SQL, wrap it with any reporting tool you’d like and share your premise, conclusion, confidences and lift to various teams within your company by feeding your SQL Server table back into your CRM engine.
Using RapidMiner Server and RapidMiner Studio will help you to get more targeted results that benefit your whole company and build a foundation of intersectional trust between all teams.