Proactive, data-driven approach reduces churn for news media
Ideas Blog | 23 June 2022
News media companies hoping to mitigate churn need to become more proactive, Eva Ascarza, a Jakurski Family associate professor of business administration at Harvard Business School, recently told INMA members.
“I don’t want to wait until customers or subscribers disappear,” Ascarza said. “I’m going to anticipate and try to prevent it.”
News publishers should look at their subscribers and determine which customers are at a higher or lower risk of churning, Ascarza said. This requires being strategic, efficient, and smart about which customers are targeted: “I don’t want to waste resources on the people who are going to churn in the first place.”
Even though there has been great progress in collecting data on customers thanks to machine learning and Artificial Intelligence, many companies are unhappy with their retention rates.
“The idea is we have a lot of methodology, we have a lot of data, and we have the ability to personalise much better than before,” Ascarza said.
Companies aren’t happy with retention rates for one reason: They are targeting the wrong customers, she said. A company can be more successful with retention by taking the same group of customers and practising what Ascarza calls prescriptive churn management.
“The idea here is not trying to figure out which customers are likely to churn or not,” Ascarza said. “If the goal is who should I target, the right question to ask is not who is about churn, but who is going to be more receptive or less receptive to my intervention.”
Sorting customers by low or high receptivity makes it much clearer who the company should target. Media ompanies can actually find the answers with testing, she said: “If you have the capabilities [for] experimentation, you have the capabilities to understand who is going to be more receptive to your company.”
Ascarza works with companies that want to know if it’s worth it to target this way. Without testing, it’s an empirical question, she said.
“Your typical types of intervention could be more receptive with these people but it could actually not be the case,” Ascarza said.
Changing the focus from risk to receptivity
Ascarza ran experiments with companies in different industries to see if they could affect retention rates based on the types of customers they were targeting with communications, discounts, and other incentives.
Ascarza said in both companies A and B, each was running a retention campaign. Before any intervention, there was no clear connection between people who churned and those who didn’t. Targeted customers all received an incentive to stay with the company and they chose targeted customers randomly. After the experiment, they analysed churn.
“In both cases, what I really wanted to answer was: Are the people who have a higher risk of churn those who have a higher sensitivity to intervention,” Ascarza said.
When targeting the top 10% of customers at risk of churn, Company A found only 16% of those customers were receptive.
“Eighty-four percent of them were not being receptive to the company,” Ascarza said. “That’s money wasted.”
At Company B, of the top 10% at risk of churn, only 6% were highly receptive to intervention and 94% were not.
“If you just go to the top 10%, what happens is these people are about to leave anyhow, so it’s kind of this idea that these people are already gone,” Ascarza said.
To see if they could retain more customers who weren’t in the top 10% at-risk, Ascarza also looked at the top 30% of at-risk customers:
- At Company A, 37% were receptive and at Company B, only 16% were receptive.
- When Company A targeted the top 10% highest risk customers, its retention rate increased by 2%. Changing who it targeted by looking at receptivity instead of risk increased the retention rate by 8%.
- At Company B, communications to the top 10% of at-risk customers actually proved to turn people off and its retention rate decreased by 2%.
- By targeting the same number of people but basing it on lift, retention increased by 4%.
Predictors vs. drivers
In many cases, companies confuse predictors with drivers, Ascarza said. It is becoming too easy to let technology do the targeting instead of stopping to think: “We don’t do it because we keep asking computers to do many things. Sometimes I think we just don’t stop and ask the right question.”
If machines are going to continue to give answers, people need to be sure they’re asking better questions, she said.
“Instead of prediction, what we need is prescription. It’s not useful for the doctor to know which patient is more likely to die. What is useful for the doctor is to know how the patient is going to respond to the treatment.”