Protecting Revenue with Predictive Analytics

From common sense baby steps to sophisticated data science: how to protect revenue by avoiding customer churn.

Data means revenue

Two-thirds of “very successful” companies use marketing automation (MA) systems extensively, according to a recent study. That is perhaps to be expected. But here’s a surprise: More than a third (37%) of companies achieved best-in-class status with just limited use of MA.

Marketing automation systems

But despite MA being an obvious tool, and despite research to show that a higher percentage of companies in high revenue tiers use MA, less than half of companies are actually using it.

Sometimes using data is difficult. Companies have massive amounts of data from a variety of sources and platforms. Also, the bigger the company the greater the likelihood it may have inherited IT systems from mergers. The good news is that, despite difficulties, baby steps will pay off.

Lowering customer churn with predictive analytics

The two main barriers, according to research, to getting started are lack of an effective strategy and system complexity. But one doesn’t need a massive amount of resources and planning to begin to pick the low-hanging fruit using predictive analytics (PA) in subscription-based services.

Customer churn is an area in many businesses where PA can be applied for quick benefit. Take telecommunications companies, for example.

Telcos collect a lot of data related to customer usage of services and their networks. They also track customer preferences and buying habits. They store and manage vast amounts of data on a daily basis. This data can be used to analyze customer behavior, uncover drivers of churn, and identify potential churners before they leave.

Predictive models

Predictive models can be built, too, which estimate probabilities for any customer leaving. Factors can be identified (4G, data usage, mobile plans, incoming messages) that affect churn. And it can also be predicted to which new network a customer will go.

Ad hoc methods can be disposed of and actions can be automated to allow targeting based on objective customer characteristics. For example, specialized offers can be made at a point before churn becomes a danger.

CRM efficacy strategies can be measured via data on response rates, designing experiments, and analyzing the treatment effects of various methods. Know if a personal email is sufficient to reduce churn, or whether a gift is required. In the case of a gift, know for sure whether the monetary gain from customer retention will outweigh the cost of the gift.

Common PA practices in the telco industry

In the experience of Proekspert employees, we have noticed these particular patterns in the telecommunications business:

If a single customer departs for another company, odds are high his friends will carefully consider doing so, as well. Data can be used to identify friends and make offers if necessary.
Family changes are easy to monitor and make appropriate offers. While some life changes are beyond control (perhaps someone marries and switches to the spouse’s company), but others, like a school-age child are easy to handle, with offers of an additional number at a favorable price.
Regional peculiarities are also identified by using PA. Some regions will churn at much higher rates than others, enabling a much more effective spend of advertising budgets.
Behavior monitoring is also possible to a certain extent. A change in television viewing patters often indicate that service quality has deteriorated. This information enables proactive change.
Cultural churn can also be predicted and monitored. In countries where many cultures are present, churn characteristics and rates can be distinctly different for different language groups.
Text analysis is another available tool. For example, words like “temporary” used by the customer in email or chat channels often signal that the odds of churn have increased.

Beyond common sense

But common sense, despite it being not so common (as Jefferson apparently remarked), can sometimes be misleading. Common-sense metrics like age, marriage status, sex, etc., are often non-behavioral and therefore not optimal for segmenting customers. But a fully-reactive digital overview of a customer base, combined with AI-assisted suggestions of actions to take, can generate some amazing results like these.

Life-time values of individual customers can be estimated by aggregating average revenue per user (ARPU) data. This allows you to know how much a customer is worth at the beginning of a contract, or a few months or few years into a contract. With accurate cost-benefit calculations, you know which customers are worth more to keep and you can spend accordingly.

By clustering customers into data-driven segments, easy-to-use dashboards can be created for optimal CRM. Query the entire 360-degree customer profile (e.g. most recent call center interaction information) and summary statistics, going well beyond just churn indicators.

How data scientists can help your company

The importance of PA and MA cannot be underestimated. In 2017, CMOs are on track to spend more on technology than even CIOs. The big budgets of marketing must now stretch to important investments in marketing technology, including infrastructure, talent, and paid media. According to a Gartner CMO spending survey, marketing chiefs are spending more than a quarter of their budgets on technology and the marketing department’s tech budget generally rivals that of the CIO’s entire IT budget.

MA is also connected directly to customer experience, since frustrated customers lead directly to churn. Gartner further predicts that by 2018 more than half of organizations will redirect investment toward improved customer experience. Research shows that among unhappy customers, 13 percent will tell 15 or more people that they are unhappy. Interestingly, customers generally do not direct their complaints to the company who could help them — 67 percent cite bad experiences as a reason for churn, yet only one in 26 (ca 3,9%) actually lodge a complaint before leaving.

Proekspert data scientists can help you take PA to the next level, with complex algorithms specifically designed to account for individual aspects in every customer segment. Accurate churn predictions enable planned customer engagement before the risk of churn becomes critical — and a healthier bottom line thanks to stable revenue.

If you are interested in what data scientists can do for your company, contact: Proekspert Data Science Business Development Robert Filic robert.filic@proekspert.ee