We define churn as the percentage of customers who leave over a given period, such as a month or year. The simple calculation is:
There are multiple natural reasons, such as moving home and price competition as well as technical reasons such as dissatisfaction with the current service received. Therefore, some churn is inevitable and there is little that can be done, however, much churn IS preventable with the right marketing, operational tools and focus.
What are Industry Typical Churn Rates?
Publicly traded companies usually publish their churn rates in annual reports, since it is a market significant KPI. A change in churn rate can indicate how well or badly a company is operating its business. Some recently published rates in 2023 from the UK are shown below:
What is the Cost of Churn?
The main reason companies try to reduce churn is the huge cost it incurs on the business. As anyone working in sales will tell you, new customer acquisition is both the hardest and most expensive to achieve. For consumer brands, there are marketing and advertising costs, out and onboarding processing costs, equipment and sometimes even engineer visits involved.
The cost of replacing a customer who has churned varies, but estimates of between 30x and 50x the cost of retaining a customer are generally accepted.
A Churn Cost Model
The cost figures below are examples only and need to be customised to each business, based on their specific costs. The model assumes that 100% of churned customers are replaced in the same period. Of course, if a churned customer is NOT replaced, then the costs (lost revenue) will be even higher.
Using the above figures, a modest reduction in churn from 1% a month to 0.9%, saves over £1M a year.
Today’s Churn Strategies – Based on Hope?
Churn is notoriously difficult to measure, as today’s strategies are mainly based on simple network KPIs such as throughput and packetloss and the hope that if they are ok, then the user experience will also be ok. However, just because throughput/latency/jitter looks good, this does not guarantee the end-to-end user experience will always be great. Often these KPIs are sampled over time and can miss short peaks. Micro-congestion in the core network, caches, transit links or home wi-fi issues can all cause problems.
Hope is not a good churn reduction strategy.
Without a way to accurately measure the user experience, reducing churn will always be a huge challenge.
Customer Experience and Churn Prediction
The holy grail of subscriber analytics is to be able to accurately predict if a subscriber is likely to churn. This article only focuses on churn that is due to receiving a poor user experience, since Sandvine’s core expertise is to be able to measure that customer experience.
Step one: identify the application/service in use. Sandvine can identify >95% of network traffic, even with encryption.
Step two: measure network KPIs very accurately, such as packetloss, latency, throughput and jitter.
Step three: match the KPIs to the service’s requirements. As an example, YouTube requires good throughput, whereas voice services require lower throughput, but very low packetloss.
Step three results in a “MOS” style user experience score for each application per user.
A Sandvine Churn Prediction Model
Having an accurate measure of the user experience is a great start, but this now needs to be translated into a churn model. Sandvine has been working with our customers for many years on this topic, notably with Telenet, who achieved an 84% churn prediction accuracy.
Churn modelling can be enhanced by optional customer data (such as the customer age), but the model structure is usually based on:
- Measuring the customer experience with Sandvine App QoE
- Focusing on the top 20 applications for each user
- Applying a weighted priority to the applications
- Applying minimum QoE scoring to the top 5 applications
- Calculating an overall customer churn ranking for every subscriber/household
To reduce churn rates, focus on the worst 1% of churn ranking customers:
- Network operations proactively search for common network root causes. Sandvine’s home wi-fi analytics can also be used.
- Customer retention teams automate proactively targeting customers with problem acknowledgement messages, action plans and if necessary, money back.
To improve the churn model quality, Sandvine recommends running it twice, once for top applications based on volume and one based on the number of connections, then combining both for a final user churn prediction ranking. The model may be further enhanced by excluding some “background apps” that might be creating a high volume or connections, but are often not user experience impacting. IoT devices and BitTorrent traffic can be good examples.
Predicting who is likely to churn due to a poor user experience IS possible and the Sandvine App QoE data provides a solid foundation to do so. Even small improvements to churn rates can have a significant impact on financial performance.
Sandvine has a proven track record of helping our customers successfully reduce churn and save money.
To learn more about Sandvine's App QoE solutions, meet with our executives and technical experts at MWC 2024, Feb 26-29. Book your meeting here.