by Suraj Pabari, Partner in Customer Analytics at SingleView, a data consultancy in Australia.

Summary

  • Currently, many businesses value customers based on their first purchase. As a result, they may under-invest in customers that would be higher value over a period of time and over-invest in lower value customers.
  • However, we can value customers based on their lifetime value, which is the total spend of the customer over a given period of time.
  • One way to measure lifetime value is to segment based on recency, frequency and monetary value.
  • By calculating RFM scores for our customers, we can segment our customers and understand the highest value customers,as well as those with low frequency, low average spend and those who have not visited for a long while.
  • We can then take action based on these segments to increase long term business revenue.

What do we mean by lifetime value?

You acquire a new customer. Not only that, but the customer returns again and again, spending hundreds of dollars every time they return. This customer must really love you! However, traditionally, you value this customer based on the revenue from their first purchase and now, this customer does not look so good. In fact, you would pay the same for this superstar customer as you would pay for his discount-seeking neighbour, who only bought something because it had a heavy discount. But do not fret, as we are moving to a new world, a world in which the customer is valued based on their total spend over a defined time period… or their lifetime value. Now, you can more easily find the valuable superstars and give them that star treatment with the hope that they stay with your business, and help you discover more just like them, transforming your business into a business of superstars.

The problem with big shiny things

Let’s imagine an extremely technical person in your business spends six months on building a model to predict the customers that will churn in the immediate future. They tell you that they have found a number of relevant features that predict churn, such as delay in bill payment, location and even age. They even excitedly proclaim that they have refined the model to the point where the model can find almost 90% of the customers that will churn. You are excited! You decide to run an experiment by showing ads to customers that are likely to churn to try to prevent them from leaving. However, clicks are so low, that you see no difference in churn rate. You go home, dejected.

Can you relate? After nine years in the Marketing Analytics space, I certainly can. And I’m not the only one.

We often hear buzzwords about things we ‘should do’. ‘Optimise to lifetime value,’ they say. ‘The customer journey is complex, implement a custom attribution model,’ is a common recommendation. ‘Most of your sales are occurring offline, why not connect offline data?’ is often proposed as a solution. You spend time and money on these pet projects, thinking they will transform your business and you divert resources away from your acquisition marketing efforts. The results are often far from impressive.

Should you avoid these types of projects? Absolutely not. The intentions of these projects are generally sound. However, my learning has been that these projects only work when you think about ways that you can use these programs to drive value before you invest significant resources. Write these specific actions down. Get endorsement from other teams that need to be involved. Quantify the value (with some justifiable logic!)

One particular project that often can drive a high cost despite demonstrating a low return is ‘lifetime value’. However, hopefully by the end of this article you will realise that lifetime value can drive significant business growth, in ways that are neither time consuming nor expensive.

The most valuable customers are the ones who have spent the most, right?

To start with, let’s discuss why lifetime value is so important. The current state of measurement with many businesses can be demonstrated in the graphs below. Assume the bars represent the value of five individual customers and the horizontal line represents the cost to acquire a customer: naturally, you set that cost to be equal to the value that you will get from the customers to break even (or even lower if you want a higher margin). You can see that by optimising to this cost you end up over-investing in some ‘low value’ customers (shaded red) and under-investing in other ‘high value’ customers (shaded green).

Figure 1: The problem with optimising to the mean

In contrast, if you were to focus on acquiring more high value customers (the green bars), in many cases, the long term revenue would be higher, as can be seen below.

Figure 2: Maximise share of high value customers

So what do we mean by high value customers? The ones that spend the most money?

Not quite. Often using total spend over a period of time can be a good proxy, but it is important to take a more nuanced view. Take the example below. Which customer do you think has the greatest value to the business?

Figure 3: Looking at recency, frequency and monetary value

Were you able to guess correctly? The behaviour of the first customer suggests that they will be more loyal, as they have greater consistency and bought more recently. We saw this with one of our hospitality clients: they had customers spending a lot of money in their venues on expensive champagne, and then not spending anything for a long period; we had to think of them differently to those customers who spent the same amount, but over a more extended period. The former may have been enjoying themselves on a luxury holiday, and since they will not be coming back, it may not make sense to define them as a high value customer.

So how do we define a high value customer then?

We have a fairly simple way of measuring high value, which is based on three factors:

  • Recency = How recently the customer bought. Someone who bought a year ago is at risk of churning and may not be high value versus someone who bought last week. Note that, in some models, recency refers to the gap between the first and last purchases.
  • Frequency = Number of repeat purchases. More purchases demonstrates greater loyalty. Using frequency in this way allows us to distinguish between the tourist who came to the venue and spent a lot of money whilst she was on holiday, and the businessman who goes to the same venue every week.
  • Monetary value = Average order value. We sometimes exclude the first purchase in this average, particularly if the purchase has been driven by a voucher.

This RFM method gives us a score for each factor. The simplest way to come up with a score is to rank the value that each customer has for each variable between 1 and 5 (though the exact scoring might differ by business). As an example, for recency: bought within last week = 5, bought within last month = 4, bought within last year = 3 etc.; for frequency: 10+ purchases = 5; 8–10 purchases = 4 etc. The recommended approach to determine the actual boundaries is to ensure that 20% of the customers are in each bucket.

Now the fun begins!

With some simple maths (which can even be done in a spreadsheet) you now have three scores per customer. What does this mean?

Let’s say a customer has a RFM score of 555. Keep this customer close! They have recently bought a high value item, and will often return to buy high value items. But what about a 155? With a low recency, this customer hasn’t bought anything recently. Why haven’t they bought anything? How can we bring them back? This diagram makes it simpler to understand.

Figure 4: Developing segments using monetary value and frequency

..though note that in reality we are looking at a cube as a pose to a square!

Figure 5: RFM cube

You may already be thinking about some of the things you can do with this data. What could you do with a ‘High Spender’ to increase their frequency? Would you give the ‘Superstar’ a voucher or would you instead offer them a ‘concierge’ service? Now, rather than simply saying: “These are my high value customers and these are my low value customers,” you can answer some more complex questions, such as:

  • Who are my highest value customers?
  • Which customers are on the verge of churning?
  • Which customers have the potential to be transformed into higher value customers? How might we do that: by trying to upsell them, or getting them to return more often?
  • Who are the low value customers that you can ignore?
  • Which group of customers is most likely to respond to your current campaign?

I hope you now have a better idea about the power of LTV segmentation, and understand how you can segment your customers using RFM methodology to answer some important questions.

How do you currently segment your customers? How do you leverage the insights from your segmentation? Feel free to add some comments in the post!

Leave a Reply

Your email address will not be published. Required fields are marked *