For years, Amedia has been a poster child for successful digital subscription acquisition in the local media sphere, with continuous high sales figures week after week. But while sales have remained steady, so has churn.
We’ve tackled churn as part of our normal day-to-day operations since starting down the digital subscription road in 2014. Our benchmarks show we’ve been performing on par, and in some cases better, than our peers. But for every two steps forward, we move back one. In late 2019, we decided to more systematically address the problem.
In September, we formed a cross-functional task force jokingly called “The Churnminators.” It drew on resources from data science and engineering; analysts, designers, and developers from our product development teams; and business analysts and project managers from the subscription management teams.
Crucial insight from data
Crucial for working systematically with churn is our vast data lake, which essentially flows from our single sign-on solution aID. A half-million subscribers log in every single day, giving us a rich level of detail in structured data on their engagement and loyalty, as well as a host of unstructured vectors for machine-learning and analysis.
This powerful data set and its business use is governed by a clear data strategy, encompassing everything from data capture and enrichment to privacy safeguards.
The process was kicked off with an intensive number-crunching and analysis phase on this data trove, processing 1.8 terabyte of data into sophisticated AI-models. We identifed no less than 147 indicators of churn — or user behaviour or events leading to subscribers canceling in the near future.
The teams were charged with full responsibility for both the solution and process, through a fairly standardised, agile set-up. This ensured full buy-in from developers and managers alike. Top management was present at all demonstrations throughout the process, offering a high degree of visibility for both the work and the successful iterations of value delivered.
The work took place in an intensive 10-week period with an ambitious goal to deliver a real, measurable reduction in churn by Christmas. From the long list of churn indicators, we drilled down to specific use cases, defining initiatives with the highest likelihood to reduce churn. Week by week, we tested, measured, and implemented activities in five different areas:
- Onboarding: How we meet the subscriber in the crucial first 100 days, particularly in the period immediately after subscribing.
- Personalisation: How we surface relevant content specific to reader needs on our front pages, article pages, and in newsletters.
- Payment: Optimising the recurring payment flow to mitigate non-voluntary cancellations.
- Cancellation flow: Ensuring that the benefits of the subscription are clearly understood before canceling and implementing a feedback loop to help us understand why subscribers cancel.
- Optimising the win-back process: Targeted messaging to the candidates most likely to return both after cancellation and in the final month before subscription time runs out.
During the process, we developed a live-user segment dashboard to monitor the “health condition” of users, grouped into micro segments. This dashboard tracks both our engagement index and loyalty index, which are calculated continuously per subscriber. Both indexes are trained to predict future value: the first solely through consumption (engagement) and the latter through a combination of non-editorial factors (such as demographics, tenure, and payment method) and engagement (loyalty).
We had high hopes going into the process, but the results were nonetheless remarkable. Seemingly small changes turned out to have significant impact. All tasks were performed with rigorous and statistically significant A/B tests and have since scaled to Amedia’s entire ecosystem.
Highlights from our tests and results including the following.
We demonstrated that establishing engagement immediately after subscribing is crucial to fend off churn in the first week.
With our first test, we optimised articles we show readers immediately after reading their first article using Amedia’s rich data sets to determine relevancy. Before optimisation, 25% of subscribers canceled during the first seven days. The optimised flow reduced cancelation to 18%. We fully expect this difference to drop when we scale the test to the entire population, but the results are still significant.
With the second test, we followed up with new subscribers through personalised e-mail newsletters and with a targeted SMS with a link to download the newspaper’s mobile app.
The result was a 10% jump in engagement for those exposed to the test compared with the control group as measured by our engagement index, which predicts future economic value.
Most subscribers cancel their subscriptions through the online cancellation flow. Each step in the flow offers an opportunity to present the benefits readers lose when cancelling.
With our first test, we designed a new cancellation flow optimised to highlight subscription benefits including showing which family members will lose access (we have individual logins for household members) and which great (personalised) stories they’ll no longer have access to. We also designed the flow to give us better data on why people are canceling.
We saw a 6.2% reduction in cancellations with the new flow compared with the control group. Even more promising was that most of those retained by the optimised cancelation flow carried over into the next renewal. The reduction into the next renewal was still a healthy 4.2% reduction compared with the control group, proving the long-term value of the changes.
With the second test, after subscribers have cancelled, we show an “undo cancellation” button embedded in articles, newsletters, and on administration pages as a reminder of the subscription term.
This initiative alone has an estimated impact of saving of approximately 4,000 subscribers a year. We are currently re-implementing it in more sophisticated ways to maximise the effect.
Subscribers paying with credit card are more likely to churn than those paying with direct debit. This could be because they do not have coverage at the point of payment or because the card has expired.
We tested different improvements to optimise payment and a more systematic approach in offering payment. We tested timing of withdrawals and frequency of attempts, incentives to change from credit card to direct debit, and, as a last measure if none of the previous efforts have an effect, we send a .pdf invoice by e-mail and, finally, a print invoice by post.
The results were simply tremendous. By optimising the timing and frequency of withdrawal attempts, we saw payments from 12% more customers than in the control group. A whopping 21% of customers that would have otherwise lost access to their vital local news source paid after receiving a .pdf invoice. And as much as 43% (!) of customers receiving the print invoice by mail paid to renew their monthly subscription.
Like most large companies, Amedia has a streamlined win-back call centre operation. The call centre obviously reaches out every single day to a large number of subscribers who have cancelled but given that it serves more than 80 newspapers, it cannot phone everyone. That’s why it’s important to reach the right customers.
We utilised data to determine which subscribers have the highest likelihood to continue their subscriptions when offered the right incentive. Each night, we generated optimised call lists, sorted by loyalty (i.e. those most likely to win back). So far we are experiencing a success rate of 12% of those who have cancelled and are now resuming their subscription.
The 10-week period of sustained effort, culminating in the launch and scaling up of all efforts before Christmas 2019 proved to be a great success. This is being carried into 2020 with dedicated team building on the results from the first wave of tests.
In total, the estimated number of subscribers “saved” by the initiatives coming from the assessment of the indicators identified in the analysis phase comprised of approximately 10% of our total subscription sales year-over-year. To state the obvious, this represents serious money in our efforts to continue funding a healthy local journalistic ecosystem.
Working on reducing churn is an ongoing process and continuing growth is as important as acquiring new customers. Our recent focus on systemising and reinforcing our commitment to churn has revealed major opportunities for improvement, even in areas where we were performing on par with the best performers in the industry.