Every loyalty or CRM program has a churn problem. The question is, how big is the problem?
Setting Expectations of Churn
The definition of churn can vary enormously by industry and transaction frequency. Once you define your churn, before you set targets and compare to industry benchmarks, reflect on your core business KPIs and the maturity of your program.
If your program is only 1 to 3 years old, your business objective likely is to attract new customers in to the program, and therefore, grow your base. As new customers are more likely to churn versus existing customers (as shown in Table 1) your churn rate will naturally be higher. In high frequency industries like banking/credit cards, we see that there’s an average churn rate* of 25% for new customers** versus 16% for existing customers. Thus, your new customers are almost 1.5 times as likely to churn versus your customers with longer tenure. For lower frequency industries like health & beauty or retail, the churn rate for new and existing customers is 79% and 52% respectively, again showing that your existing customers are more likely to stick around. This also illustrates the need to determine your churn definition based on your industry and likely frequency of transaction. A slower moving industry may not expect to see customers returning more than once per year if their replenishment cycle is, as an example, every 12 to 18 months.
Early Signs of Churn
In most industries, there are typical “warning signs” of churn; 61.8% of customers show signs of decline before dropping off completely.
If you don’t have an analytics team or the funds to justify outsourcing a churn model exercise, you can still build trigger-based campaigns that work off business rules rather than machine learning algorithms. Reactive, trigger-based campaigns will not be as effective as predictive based campaigns but will still yield positive results and ROIs.
The best option, however, is to predict and prevent churn. Funnily enough, this is the easy bit. Any data scientist can build an algorithmic model that will predict which customers are most likely to churn in the upcoming period, but what do you do next? I recommend you take two more steps.
Understand Your Customer
Build a value segmentation. This will allow you to differentiate your offer generosity within the campaign to ensure the best customers (who are likely to churn) are getting the highest possible incentive to stick around and continue spending.
You could go on to build a behavioral segmentation, market basket analysis or next best offer model to understand what product to offer in the prevention campaign, however, as you are simply trying to encourage your customers to continue to purchase, it’s in your best interest to keep the offer broad rather than restrictive.
Build a Safety Net
The next thing to do is to build a safety net. Even an exceptionally accurate churn model will not predict every customer who is likely to Churn, so to give yourself every opportunity to keep your customers engaged in your brand. You can build a fairly simplistic secondary campaign that targets those customers who are inactive, for a defined period of time, despite being or not being targeted via the churn model campaign(s). The defined period can be determined to ensure you are not leaving it too long and hence losing your customer, but also not giving an offer too soon to avoid cannibalization due to natural reactivation.
It’s Then, All About the Campaign
Now that we know which customers to target and which will receive a better, more generous offer to encourage further engagement, it’s then all about how you execute the campaign. It is critical to keep in mind: the model is only as strong as the campaign that follows. You need to get your messaging correct, your creative memorable, your timing spot-on and your offer attractive. Just as importantly, you need to test and learn to determine what works and what doesn’t and how to leverage the results for ongoing success.
Seeing Success in Churn Reduction
Here at Aimia, numerous clients have improved their churn and retention rates when they utilise our SmartJourney® methodology. Here are just two examples:
A leading global CPG/FMCG brand was facing a problem with customer retention due to increasing competition in the market and the short customer lifespan with the product; they were looking to proactively retain their customers who demonstrated early signs of churn. Aimia designed data-driven test & learn campaigns off the back of a churn prevention model. Two test & learn campaigns executed over two independent periods resulted in incremental sales of more than USD $350k, with response rates over 30% and an ROI in excess 1,000%; meaning for every dollar spent more than 10 dollars was returned.
Based on the success of the test and learn exercise, Aimia executes monthly churn prevention campaigns for this CPG brand, which continue to drive incremental value and keep the churn rate low. 1
A global retailer achieved similar success across multiple markets in Asia. After building a churn model and going through a test & learn exercise, the monthly ongoing churn prevention campaigns, including the second-touch safety net discussed above, generated triple digit ROI and ongoing churn reduction of 30%. 2
Assess your churn problem by determining your churn definition based on your industry and maturity of program. By analysing your new and existing customers behaviour you may find that your issue is primarily with your new customers and hence better on-boarding and cultivation activities are required.
Churn prevention campaigns, done well, will always generate the highest ROIs amongst all your BTL activities. This is simply because, “all spend is incremental spend.”
Personalization, as with any communication, is key, however don’t restrict your chance of a response by limiting or narrowing the offer to certain products or categories when your primary objectives is any transactional spend. Be as broad as possible whilst still showing your customer that you know them and you know what they like.
1 Based on internal Aimia data from 2017 – 2018
2 Based on Aimia data from 2018
3 Based on Aimia data from 2017