RFM Analysis — Who Your Customers Are … and Why You Need to Know

How can you ensure that your marketing campaign is targeted correctly? What kind of people are attracted to discounts? Knowing your customer is key to your business. RFM analysis can help!

If you are a shop assistant in a brick-and-mortar store, you can have a conversation with every customer; however, in an online shop, this is not realistic. But there is no need to know everyone personally. All you have to do is have information about the customer’s previous shopping behavior — when and for how much they purchased.

For us, this is enough to be able to create an RFM analysis. We divide customers who share common behavior into several segments. Thanks to that information we know:

  • who deserves to be treated with special services, e.g. loyalty program,
  • who has potential to spend more,
  • who shops regularly,
  • which important customer is about to leave,
  • who needs to be enticed to purchase using a properly targeted discount,
  • who is not worth targeted advertising because they wouldn’t shop anyway.

This helps by:

  • ensuring customer loyalty,
  • increasing conversion ratio, 
  • increasing income.

What is RFM analysis?

RFM analysis is a method of customer segmentation based on previous purchasing behavior. This technique is time-tested, first used in the 1990s, and is still considered the gold standard in data analytics. All we need are 3 basic metrics:

Recency, R – number of days since last purchase

Frequency, F – total number of purchases

Monetary, M – purchase value

For each customer, we can monitor other important metrics — average order value, first and last executed order – and at the same time it is possible to evaluate customers over the entire dataset according to the number of orders, total sales, etc.

Practical example

We always divide RFM metrics into several levels. Here we then define intervals at which we continue our analysis. Everything can be seen in an example of one of our clients.

 

In our case, Recency has five levels. Some intervals are in the range of two months; summer is joined into one period because there were fewer conversions. The period of the most important season for the client is a month and a half.

We process Frequency and Monetary metrics in the same manner. Now we check the spread of customers so that there aren’t too few of them here or too many of them there. Of course, we are still taking into account what the evaluated online shop sells. 

This breaks down the customers into individual intervals, which is nicely shown by the visualizations below. The colors reveal that the blue customers are inactive (they purchased once and for a small amount), while the red ones are the most active (they purchase regularly, with the highest spending, and not much time has passed since their last purchase).

 

Once the customers have been divided, we can move on to the segmentation. For each segment, we define which R, F or M interval is applied. Areas must not overlap, which would mean trouble in further evaluation and use. At the same time, we must be careful not to forget any customer; every customer should be included in some segment.

Using the example, we define the segments of active customers who have the highest and high spending. These are the ones who have purchased at least three times in the last six months. The average purchase value for the VIP segment is higher than CZK 3,000 and for the Loyal Segment it is between CZK 1,000 and 3,000. 

 

And so on. Which customers have only recently made their second purchase? Who are the savers who shop regularly but don’t spend much? Who should we focus on because they used to spend a lot, but haven’t made a purchase lately?

This results in several segments telling us exactly who these customers are and their benefits. We can also compare changes from the previous period

 

 

In the last step, we select the segments that are most important for the client’s business and we continue to work with them. For example, we can start emailing. We specifically target offers on selected segments or create a circle of users on Facebook based on email addresses for remarketing campaigns. 

As part of the acquisition, these segments can be high-quality resources for look-a-like audiences. We can target users similar to the most loyal customers. If we focus on customers who spend often and a lot, the investment is likely to return.

It is also good to choose those who have purchased recently and push them to purchase again, or to offer a discount or select customers for research. 

When Is It Appropriate to Process an RFM Analysis?

Are you wondering which period is best to perform RFM analysis? The basis should be one year; it is great to have data for year-on-year comparisons. To better control for the movement of customers between segments, it is beneficial to perform an RFM analysis for a half-year or, even better, each quarterly period.

This way you can be sure that you offer discounts to those customers who will respond to them the way you want. Monthly checks are best for optimal review of customers. 

RFM analysis is the basis for further predictions of future customer lifetime values, such as indicators of when a customer will purchase again or even gloomier forecasts of how many customers may leave in the future. These are the numbers that are crucial for planning a marketing strategy.

We Can Help You

We will be happy to process an RFM analysis for you or advise you on how to do it. What do we need in order to do this? Data from orders from the selected period for analysis and at the same time, if possible, data from the previous period for comparison:

  • customer ID, 
  • order date,
  • order value.

We then divide customers into segments, including their shares of sales, and compare the results with the previous period. We recommend how to work with the segments further, who is worth investing and who would be just a waste of money. Let us know if we can help :)


Anna Petříková

Starám se o analytiku a e‑mailing. Jsem přesvědčena, že informace na základě dat jsou to nejdůležitější pro správné rozhodování. A ta nejspolehlivější rozhodnutí lze přece učinit jen tehdy, když je člověk dobře najezený a vyspaný.