Services

Marketing Automation & Segmentation
Data Science & Machine Learning
Data Integration & Architecture
Data Warehouse Modernisation
Data Visualisation & Reporting
January 21, 2019 5:29 pm

Why Are My Customers Leaving?


7 steps for using customer analytics to identify and address why your subscribers churn.

Written by Tom Coppock

From gyms to TV packages, from software to e-commerce, more and more businesses are relying on subscriptions to monetise their products and services. Analysts McKinsey & Company have measured the growth of the online subscription business models to be over 100% per year for the last 5 years, while the 2018 State of the UK Fitness Industry Report puts gym membership in the UK at 1 in 7 people. Whichever way you look at it, selling subscriptions is big business.

But no matter what you are selling, if you are using a subscription-based business model at some point you’ll hit the same challenge: customer churn and how to avoid it. With many consumers preferring to sign up for “no ties” contracts, that allow them to change or cancel their subscription on a month by month basis, and the cost of acquiring new subscribers often not being recouped until many months into the membership, the importance of keeping your customers long enough to turn a profit has never been higher.

If you are facing a churn challenge, need to increase the lifetime value of a subscriber or are thinking about running retention marketing campaigns, where do you start? One of the first questions you should be asking your self is: “why are my customers churning in the first place?”

This is where customer analytics can help.

1. It all starts with data

The more sources of data you can bring together about your customers, the higher the chance of discovering an actionable insight. First off list out all the sources of data that you have available (bear in mind you need the right legal basis to process data about an individual – see our post on GDPR if you need more information). Typical sources of customer data include:

  • – Digital behavioural data, e.g. email opens, social media shares, website visits, mobile app downloads and usage and so on. This data is often available from your email or Marketing Automation tools.
  • – “Real world” behavioural data from loyalty card schemes, access or swipe cards, booking systems, consultation notes etc.
  • – Product usage data, for example viewing or playback data for media consumption (if you provide a video or music streaming service), or app usage data if you provide a digital service. This could also be data from devices in your physical locations, e.g. data from smart gym equipment.
  • – Purchase or subscription history from your e-commerce, billing or CRM system.
  • – Demographic data, often stored in a CRM system, such as age, gender, post code.
  • – Support call, customer service or help desk case data.
  • – Other sources of partner or third-party data (accessible with the correct licensing and user permissions).

 

2. Don’t be afraid to ask

One of the best sources of customer insight is, unsurprisingly, the customer themselves. Surveys, social media and user communities are great places to ask for feedback on your product or service, especially at key points in the user journey such as the start or end of the subscription. 

While using scores and “multi-select” options in your questionnaires can give structure to the responses to make them easier to analyse and compare, often the best insights come from “free text” questions such as the from the “why did you score us that way” question used in the NPS (Net Promoter Score) methodology.

Even if you end up with thousands or even millions of free text responses, today’s technology will let you quickly process them to find key insights. Services from companies like Amazon, Microsoft and Google use a technology call Natural Language Processing, that will turn these free text responses into scores for sentiment across different categories (cleanliness, staff, location, service experience etc.), intents (purchasing, using, servicing etc.) and product lines. This data can then be visualised as charts or phrase clouds to make it accessible and easy to understand.

3. Location, location, location

If you provide your services in different physical locations this can be one of the key dimensions to use when analysing your data. Wherever possible associate data with the location to which is relates. This will let you analyse to see if there are there recurring trends across all locations or do certain types of feedback correlate to specific locations?

If possible, also capture the time period for the feedback or behaviour so you can see which of your staff were on duty working at that time.

4. 1 + 1 = 3

To get the best insights, data needs to be joined together. Typically, this needs a common key such as an email address or phone numbers to identify about the same individual from different sources. Joining the data can be a quick task which you do in tools like Microsoft Excel or PowerBI or can be part of a large project such as a Data Warehouse, CRM or Single Customer View initiative. The best advice here is to start small and prove the value, before graduating to more automated, large scale solutions.

5. Group and segment your data

Often your marketing team will have persona or segments defined that can be used to cut and dice the data. If these are available, they are a great starting point. If you don’t already have segments, are there any natural groups in your data, such as age, gender, subscription level or total spend that could be used to create groupings?

When creating groupings, automation is your friend. “Machine Learning”, often associated with Artificial Intelligence uses mathematics and statistics (aka Data Science) to find the groupings and relationships between different aspects of your data. Often this allows you to find segments and associations that would not be obvious on less scientific analysis. Once in place, this analysis can be completed automatically, saving you and your team many hours of manual analysis. 

6. A picture is worth a thousand words

Visualise your data to see any quick insights. What are the top reasons for people to leave (from your exit surveys)? Are there any locations or times of year that see peaks in unsubscribes?

Next filter this data by your segment groupings. Here analytics tools like Power BI, Looker or Tableau come into their own as they make easy to filter a view based on selecting different attributes. If certain combinations of attributes have a higher level of churn, ask yourself why, then check to see if this can be validated by other sources of data or comments in the surveys.

7. Find the leading indicators

Widen the time frame to look for the leading indicators that a person is at risk of churning – do visits to the gym or engagement with marketing communications drop off in advance of the person leaving? If you have good historical data, see if the pattern is repeated in previous years.

If you have access to an analyst or data science team, machine learning can be used to find the groupings and attributes that are associated with churn and create predictive churn scores or probabilities that gauge the likelihood that an individual will churn.

What Next? Take positive action…

You now have the insights you need to start to address the reasons your customers are churning as well as insights into which customers are less likely to be at risk. In our next post we’ll explore some practical steps to take when conducting churn analysis, then in a third post we’ll look at how to use the results to improve retention and customer lifetime value with marketing automation.  

See Accelerator for Churn Analysis

Healthcare and Pharma
February 13, 2019

5 Testing Best Practices to Error-Proof Your Pharma Sales KPIs

How a solid KPI testing strategy underpins the success of your pharma sales team's digital transformation.

Fitness Industry
February 7, 2019

Getting from Insight to Action with Marketing Automation

Getting from Insight to Action with Marketing Automation: Applying the results of data science driven churn analysis to drive retention marketing efforts.

Customer Analytics
January 29, 2019

Applying Data Science to Predict Customer Churn

A practical guide for getting your data science churn analysis project up and running.