[Infographic] The Six Steps of Churn Detection: How To Predict and Prevent Customer Loss
Contino takes a value-driven, iterative approach to customer loss and churn detection that’s focused on delivering tangible business benefits. The following six steps help you to engage stakeholders at key stages and map the process to your business’s unique needs.
1. Discovery Workshops and Business Questions
- Identify opportunities for analytics
- Determine key business questions to be answered such as:
- What are the main drivers of customer churn?
- Can revenue be increased by mitigating the drivers of churn?
- How does your organisation define churn?
2. Data Sources
Connect the data dots by identifying all the source data required to answer your business questions, and build data pipelines required to collect data into a cloud-based data lake.
3. Exploratory Data Analysis (EDA)
EDA provides a big picture view of the data and creates features that can be fed into machine learning algorithms.
- Verify the existence of hidden patterns and relationships in the data, setting the stage for formulating and validating formal analysis techniques
- Identify unanticipated structures in the data that may require further investigation or alterations to the analysis plan
- Engage stakeholders through the generation of data-driven insights
- Maximise the potential value of the data by providing deeper context to the business challenge
- Enable validation of the value of a product or technique before significant investment
- Fine tune, course-correct, or abandon ideas earlier rather than later
4. Model Development and Validation
Following a thorough data exploration process, we begin building models to explore different scenarios, mitigating weaknesses and limitations along the way, and ensuring the outputs are conceptually sound and reliable.
- Implement, train and validate machine learning models to predict churn events
- Based on the specific business questions, models can be built to predict, for example, whether a customer will churn or not in a given timeframe
5. Stakeholder Insights
Transform data into an actionable portrait of customer churn spanning every corner of your business.
- Use data visualisations and frequent communication to regularly showcase insights during the development phase
- Familiarise stakeholders with the features most predictive of churn risk
- Gather feedback on most useful insights for relevant business units including management, sales and marketing, as well as ensuring the views into the data are correct and exhaustive
6. Embed Within Business Processes
In order to realise continuous value, churn prediction models and their derived insights should be embedded within business processes, rather than exist as mere one-off reports. The churn prediction solution can be operationalised into a range of live systems, aligned to existing business processes and technology.
- Build an interactive customer profile dashboard that can visualise, for example, past purchase patterns or churn risk ratings for a customer or group of customers
- Integrate churn prediction insights within existing customer management systems (e.g. CRM)
- Implement an exception alerting solution that can trigger notifications to relevant business users when the churn risk rating of a customer exceeds preset thresholds
- Generate and send to relevant business users periodic reports with churn risk rating and related information for customers
This information can now be fed back into the machine learning models to enable prescriptive analysis, where the system can prescribe the best course of action for each customer identified to be at risk of churning.
For more information on customer churn, read our white paper Customer Churn: How to Use Data to Predict and Prevent Customer Loss.
Download the infographic here.
