Most CX measurement programs cover only 10-15% of your customers. With Predictive NPS you can predict the satisfaction of silent customers and scale the impact of your CX program to each and every customer.
Predictive NPS accurately predicts satisfaction scores for the whole of your customer base at any moment of time, not just after a customer completes a satisfaction survey. In the model we combine operational data and customer satisfaction data – NPS survey responses or other satisfaction metrics. The model then identifies which factors in customer behaviour have the highest impact on what the relational NPS of a specific customer would and assigns predicted satisfaction scores to silent customers who have never replied to surveys.
Predictive NPS integrates seamlessly with leading CX platforms. Predicted scores are displayed in your existing setup, side by side with survey responses.
The model can predict any type of satisfaction score – transactional or relationship NPS, CSAT, CES or other, as long as you are already running a survey-based program.
Our general rule of thumb is the more data, the better. Before the start of each project, we typically have a data discovery session where we evaluate the available data the client can provide. Based on that, we make a final recommendation whether it is feasible to proceed with Predictive NPS.
1. NPS/CSAT data – at least 300 unique cases, customers, accounts that goes back at least 1-2 years depending on survey frequency
2. Additional Data points – at least 10 CRM/operational variables and 2 time-specific behavioural variables (e.g. help call, cases, issue, usage, etc.) with the needed
If you can’t acquire operations and business data, we typically run Predictive NPS Light as a proof-of-concept. pNPS Light runs on existing CX data only and is quicker to implement.
One thing to be mindful of is that with predictive analytics it is never an "all or nothing" situation. It is always a trade-off. If you aim for perfection, you may never get the insights you need. You need to weigh what your organization needs, what it can handle, and what will get the job done. Predictions will almost never be 100% correct but you still get so much value by being able to save some customers from leaving you, that the investment pays off.
The accuracy of Predictive NPS light is usually around 60% compared to less than 25% chance if assigning super detractor / detractor / passive / promoter categories at chance. The accuracy of Predictive NPS full rollouts is higher (usually above 80%) since it is based on higher quality and quantity of data as inputs for the model.
Predictive NPS works well for any industry or business model. We have seen great results in telecoms, financial services, insurance, b2b manufacturing, medtech, retail and others. Check out our Success Stories page for more information.
One-time fixed cost for a pilot project and afterwards a monthly subscription that varies depends on number of markets & touchpoints. Contact us for more details or, if you are a client of Medallia, Forsta or InMoment, get in touch with your account manager.