Category Archives: Analytics Education

Using RFM Scores to Predict Profits

Jim answers questions from fellow Drillers
(More questions with answers here, Work Overview here, Index of concepts here)

Topic Overview

Hi again folks, Jim Novo here.

Subsidy costs. You’re just starting to hear people talk about these ideas in online marketing, but they’ve been around for years offline in direct marketing. The basic idea is this: sending a discount to someone who is very highly likely to make a purchase without the discount is a waste of margin dollars best spent elsewhere. And you can measure this effect quite easily using Control Groups, another concept starting to get some recognition with online marketers.

Discussing / implementing these topics can be a bit difficult, though the Finance people will get it immediately and love it if you go in this direction. A plus for fellow Drillers out there is you can start to see some of these ideas in action BEFORE you start going deep using the RFM & Lifecycle data we’ve been talking about and using for years.

Below is a great example using RFM data from a fellow Driller. You ready to go ?


Q:  Since our last conversation few months ago, we went ahead and tested 3 different promotions using the RFM model.  

The 1st promotion was the test for RFM method itself to see what patterns emerge for response rate, incremental sales, etc.  The next 2 promotions targeted the customers from RFM cells with the highest incremental lift from the 1st test promotion.  Here is what we saw.  Since the targeted audience were our loyalty card members, they transact and spend at a fairly high level (the data below is modified but the trend is maintained).  For the response rate, we saw a sawtooth pattern:

(Jim’s note: RFM is the 3 digit score, Rate is Response Rate.  More on RFM here and here.)

A: Yes…

Continue reading Using RFM Scores to Predict Profits

Using Multiple, Related Customer Models Across the LifeCycle

Jim answers questions from fellow Drillers
(More questions with answers here, Work Overview here, Index of concepts here)

Topic Overview

Hi again folks, Jim Novo here.

So you have all these simple but powerful customer models – Recency alone, Latency,  RFM, or LifeCycle Grids – how do you know which one (or ones) are best to use for your business? Guess what – it depends on the specific features of your business and also how you run the business. Now, while that might sound a bit scary, it’s really not that big a deal and in the end, the great news is you’ll end up with an approach customized to your business. So how do you accomplish this? Just segment and analyze your customers; they will tell you, my fellow Driller, which direction is the best to follow. You dig? Let’s go ahead and see what that looks like …


Q:  We recently purchased your book

A:  Thanks for that!

Q: and we are ready to start building some RFM analysis.  We are a search marketing business – we have a large customer /prospect base.  We have limited knowledge about them and we are keen to start on the journey.

A:  OK…let’s see what you’ve got.

Q:  We are hoping to extract database (approximately 25k names) of the last 6 months records and do some RFM analysis on key customer groups.   Specifically:

TEST GROUP A – people who initially purchased one of our trial products – we want to know what is their RFM score.

TEST GROUP B – subscribers to our “tool kit” product at $50 / month – we want to know what is their RFM score.

Q:  What kind of data are we talking about?  Is it web site visits, clicks on emails, transactional / subscription data, all of the above?

A:  Before setting up the model we have a couple of questions we hope you can shed some light on:

1.  How do we treat subscription – our business has a mix of one-off and subscription business – if someone “buys” every month with a subscription, is that included in Recency & Frequency?  Any insight you can provide us would be great – we found some info on this in the book but unsure given ours is a mix of subscription and one off.

Continue reading Using Multiple, Related Customer Models Across the LifeCycle

Segment to Best Determine LifeTime Value (LTV)

Jim answers questions from fellow Drillers
(More questions with answers here, Work Overview here, Index of concepts here)

Topic Overview

Hi again folks, Jim Novo here.

LTV has to be actionable.  If  you can’t take action on the information, it’s not relevant anyway.

There you go, the most universally true rule when attempting calculation of LTV.

And the best / easiest way to accomplish this is to identify similar customer behaviors and segment the customers by these behaviors – THEN figure out LTV by segment.

If you can’t actually take action on the information, then why spend countless $$ and hours fussing over all the reasons the number you come up with might be wrong and trying to solve unsolveable data or corporate issues? The best idea to implement when developing / using LTV is consistency – let’s get the team to agree on what LTV is and how to measure it, stick with those ideas for at least several years, test and take action on the results to uncover value, THEN (perhaps) discuss improvements!


Q:  I have just been reading your series on Comparing the Potential Value of Customer Groups. I am having trouble calculating the lifetime value of our customers.

A:  Yes, well, everybody does for some reason!  Often the problem is too much
focus on trying to look at the “average customer” as opposed to segmenting
customers.  By segmenting first, it’s both easier to get to LTV *and* more useful since it’s easier to take action on  a segment than the “average customer”.

Q:  Our company provide accounting software solutions to small to medium sized owner operated  businesses.  Because of what we sell and who we sell to, a lot of our customers are most likely to just buy one or two of our software products and unless they sign up for support (only around 15% do), we may never here from them again.  It is therefore very difficult to determine an average / standard lifetime that customers use our product.

A:  Sure.  First, the 15% segment that does sign up for support sound like good customers to me.  So that’s one segment.  How long do they typically stay signed up?  That’s the average life for this segment.

Then there are probably people who upgrade over time, right?  I can’t imagine an accounting product that people would not upgrade – perhaps not every cycle, but every 2nd or 3rd cycle.  That’s another segment.  Then there are probably some who both follow the upgrade cycle and pay for support.  These are probably the “best customers” and they are a unique segment as well.

And finally, you have the buyer who makes one purchase and you never see again.  These people are also a segment.

Q:  What should I base it on, how long our customers use our products (which would be almost impossible to determine), or how long they spend money with us?  So I measure on average the time between the first and last transaction of customers who have the highest Recency???

Continue reading Segment to Best Determine LifeTime Value (LTV)