Lifetime Value Part 2: How lifetime value is like a relationship: you live, you learn, you grow
Summary
We can use RFM scores to define segments within our customer base.
These segments can be used to define relevant actions that we can take with customers to increase their overall lifetime value.
Key segments include ‘Superstar’ customers with high RFM scores, for which you could develop a premium service; ‘Promising’ customers with mid range frequency and monetary value, for which you can run loyalty schemes to increase spend and frequency; ‘At Risk’ customers with low recency, for which you can send targeted messaging with possible discount to bring them back.
Run experiments to test these different ideas, and start by focusing on the largest audiences that will drive the greatest impact, with clear success metrics and next steps to scale the action.
How can I take action using RFM scores?
As you can see from the previous post on RFM segmentation, it is quite easy to get excited with your new RFM scores and to start generating elaborate experiments. However, the goal of this project should initially be to keep any action simple and easy to implement.
The main action from the RFM scoring is to create a few key segments. You want the segments to be large enough in scale so that you can actually see a difference when you run experiments using the segments. You also want to make sure that the segments are updated on a regular basis, and easy to export, so that you can easily integrate them into your marketing platforms.
What segments could you create? I often focus on the following three segments, as I have found that running experiments with these segments has driven significant revenue impact. Feel free to play around with the RFM ranges for these segments based on your goals.
Superstars: High value customers with RFM score 555 (or if you want to sum the individual RFM scores then you can segment those customers with a total score between 10-15). These customers are your ‘superstars’ and should be treated as such. This is the concierge, the airport lounge, the free delivery, the ‘free bottle of champagne on the birthday’ type of customer. These customers are already loyal so whilst they may not need a ‘buy one get one free’ discount, they do need to know that they are special. As an example, with one of our beauty client, we were discussing how they could send gift baskets to these top customers over Xmas. How might you treat your best customers if you want them to stay? What type of service might you provide?
Promising: Customers with high R, and mid range F and M (4-5)(2-3)(2-3). This is where building loyalty is important. Send these customers a voucher, an incentive to get them to spend more and spend more frequently. Start a loyalty scheme. In Australia, we love our coffee, and cafes tend to have a lot of customers in this 'Promising' segment, with many customers who come back occasionally. These cafes sometimes use loyalty (‘buy 10 get 1 free’ schemes) to develop these Promising customers into Superstars. For this segment, you want to push your brand to the top of their list of options so that it remains top of mind for your customers and they have an incentive to keep coming back.
At Risk: Customers that are likely to churn, with RFM score ranges of (1-2)(4-5)(4-5). Initially focus on the customers with the low recency score, and high frequency and monetary value scores, the ones that you most want to save. Why do you think these customers aren’t coming back? How can you test this hypothesis? Since this customer group may not be too big, try personalised messaging or calls to remind these customers that they are important. Or more simply, send them an e-mail to convince them to come back, potentially using a special offer to bring your brand to their top of their consideration set.
Whenever you are implementing any of these changes, do so with a controlled experiment. As an example, if you wanted to test a premium service offering for your high value customers, randomly segment them into two groups. Give one group the premium service, and do not give the other group the same offering. Set distinct control and test groups, and keep everything else the same. Then observe how spending varies over a fixed period of time.
Never just implement a change without an experiment. In the above example, if you were to run this premium service with all customers, and you saw frequency of purchase improve, how would you know that the positive change was a result of the service offering and not the result of some other factor, such as good weather?
But what about using this data to acquire new customers?
A lot can be said about finding ‘lookalike audiences’ to your high value audiences and importing them into your marketing platform to prioritise. These need to be tested just like anything else, with clear experiments with clear success metrics (are you looking for a greater number of clicks, a higher conversion rate, or increased spend?) I often like to think about the time vs reward of developing these experiments; it’s important that the audience size is large enough to drive significant change in core metrics to be worth the increased time and complexity in management of these lists.
--
I hope you now have a better idea about the different ways you can use the RFM scores to develop segments, and develop experiments to drive increases in overall lifetime value.
In the final post next week we will be discussing the “whys” and the “hows” of predictive lifetime value modelling (versus histoical lifetime value modelling with RFM). This model uses customer behaviour to predict how many times a customer will visit, what they will spend and the likelihood that they will churn to give us an estimate of lifetime value. You can also use this data to understand which customer attributes (age, gender etc.) are correlated to your high value customers.
How do you currently segment your customers? What actions do you take for the different segments?
