RFM Analysis: Your (Probably) Unknown Profit Hack – Now Unlocked by AI

In a recent workshop I asked ~50 finance leaders if they're familiar with RFM analysis – not a single hand went up.

That didn't really surprise me.

Even back in my data scientist days, hardly any business leader knew about this powerful technique. Typically, because it felt too "distant": too technical, too complex, too much hassle.

Thanks to AI, that has changed. Now, anyone who can filter an Excel spreadsheet can use RFM to find profit pockets for their business.

Let's find out how this works.

What is RFM Analysis?

RFM stands for Recency, Frequency, and Monetary value – three simple metrics that tell you a lot about your customer relationships.

• Recency: When was their last purchase?

• Frequency: How often do they buy?

• Monetary: How much do they spend?

By combining these three factors, you can quickly group your customers into different segments, like:

• Best Customers – Recent, frequent buyers with high spending.

• Loyal Customers – Buy often but may not be your biggest spenders.

• Potential Loyalists – New customers with multiple purchases

• At-Risk Customers – Previously active, but didn't shown up in a while.

• Lost Customers – Haven't purchased in a long time.

Each group requires a different approach – or informs your planning.

Now, here's the best part:

In many cases, you don't need to go through a complex data integration project to get these insights. You probably already have the data you need!

The Data You Already Have

To perform an RFM analysis you really just need three simple variables:

• Customer ID – A unique identifier for each customer.

• Dates – Transaction timestamps to calculate recency and frequency.

• Amounts – The value of each purchase to measure monetary value.

Ideally, you want to have this in the most granular form (one line = 1 transaction, see example below). And you want lots of them! For robust insights, more data is generally better. But even a few thousand transactions can reveal patterns.

Where to Find This Data

The nice thing is that you'll likely find these data points in one single system.

For example, your business likely collects this kind of information already in:

• E-commerce platforms

• CRM systems

• Accounting software

• POS systems

• Spreadsheets

In the past, the challenge wasn't having the data – it was extracting, cleaning, and analyzing it. Traditionally, this required SQL queries, Python scripts, or data teams.

But now, AI makes it possible for anyone to do this without writing a single line of code.

Here's how to get it done in 10 minutes.

10-Minute Guide to AI-Powered RFM

Step 1: Export Your Data

Pull a CSV of your last 6-12 months of transaction history from one of the systems above. Your data should be structured like this:

Get rid of all other columns. You won't need them for now.

Want to try this out? Here's a sample dataset you can use for this exercise.

Step 2: Use ChatGPT to Process Your Data

Upload your data to a tool like ChatGPT (with code capabilities) and run a prompt like this:

Step 3: Segment Your Customers

Once you have RFM scores, AI can automatically label customers into segments like Best Customers, Loyal Customers, At-Risk, and Churners based on their behavior patterns.

Here's how the result could look like:

Step 4: Visualize and Interpret the Results

Now that you have the RFM scores and customer categories, you can dive deeper with the help of AI.

Use visualizations to understand the data better and get insights more intuitively.

Examples:

• Show me a heat map of Recency vs. Frequency Scores
  ➜ This heat map below shows you that (besides two common clusters of low/high recency/frequency customers) there is a relevant segment of at-risk frequent shoppers (lower recency but higher frequency) who could be prime candidates for targeted re-engagement campaigns.

Have a hard time understanding these plots?

Well, just ask AI and it will explain it to you: (Good use case for reasoning models like o1/o3 btw!)

Here's another example:

• Show me revenue of At-Risk customers over time:
  ➜ This plot shows that the absolute revenue from At-Risk customers spiked recently, indicating we could lose more money than usual if they churn. Worth investigating further!

Avoiding Common Pitfalls

RFM is cool and it's easy to get excited (and sometimes carried away) by it.

So here are some typical pitfalls to avoid:

• Assuming RFM scores are universal
  A "4-4-4" might be a VIP in one business but just "meh" in another. Context matters!

• Not cleaning your data
  Garbage in, garbage out. Take time to understand your data and its intricacies (e.g., is each ID really a different customer, or could one customer have many IDs? Do you have negative transaction values and what does that mean? Etc.)

• Overthinking it 
  Keep it simple! You don't need 20 customer segments. 4-7 is plenty.

• Doing nothing with it
  Seriously, don't just admire the numbers. Use them. Data is useless unless it drives action!

Which brings us to the last point…

Turning RFM Insights into Profit (Worth $10K+)

Okay, so how EXACTLY can you turn those admirable numbers into something you'd feel in your P&L?

It all comes down to your role in the organization.

While RFM is traditionally seen as a marketing tool, you can actually apply it through multiple lenses and drive recurring profit improvements of $10K or more.

Here are a few practical examples:

1. Marketing

The classic. RFM helps personalize campaigns—reengage inactive customers before they churn or introduce special offers to your most loyal clients. But take it a step further: Combine RFM with an attribution analysis to find out which marketing channels bring in high-value customers. Stop throwing money at channels that attract one-time, low-value buyers.

💰 Profit: $10K+ from higher ROAS and prevented customer churn

2. Finance: Improve Liquidity

While RFM is pretty much unknown in finance, it can be an incredibly useful tool – especially for financial planning and liquidity management. If you can predict revenue cycles better, you can allocate cash reserves more effectively, offer early-payment incentives to high-value customers, and reduce reliance on costly credit lines.

💰 Profit: $10K+ in interest savings or faster cash collection

3. Operations: Better Supply Chain

Most companies stock inventory based on gut feeling or outdated sales averages. But if you know who your high-value, frequent buyers are, you can stock up on what they actually purchase while cutting orders for low-RFM customers.

💰 Profit: $10K+ in reduced inventory holding costs and fewer lost sales due to stockouts

4. Pricing: Charge More Where It Matters

Not all customers are price-sensitive. Instead of blanket discounting, introduce tiered loyalty pricing or premium offerings for your top RFM segments while reducing unnecessary discounts to customers who would have bought anyway.

💰 Profit Impact: $10K+ by optimizing pricing and discount strategies

Conclusion

RFM has been around since the 1970s. But for decades, only data nerds and analysts had the skills to use it. Today, AI has removed a lot of the barriers and you don't need SQL, Python, or a data science degree to do this – as long as you have a good data (domain) understanding and you're able to filter a spreadsheet, you can do this!

Pull 12 months of transaction data from any transactional system, do sanity checks and run a quick RFM analysis with the help of AI. Take action on just ONE segment. 

Maybe it's cutting discounts for people who don't need them. Maybe it's sending a "we miss you" email to at-risk customers.

Either way, stop guessing.

Use the data.

Make more profit.

See you next Friday! (Or tomorrow)
Tobias