Question on New customers in RFM Model

rfm

#1

Hello Everyone,

I am working on RFM Model, Can anyone suggest me how to deal with New customer while performing RFM Analysis. New customers do not have good frequency count, even though these new customers have very good Recency and Monetary scores they are falling in under performing customer segements, how to deal with these new customers.


#2

Please add some data/charts to understand the problem better.

Even though you said that the new customers are having good recency and Monetary scores, the scores might be low when compared to other existing customers hence new customer might have been placed in ‘under performing’ group. However, this can be only tested by looking at the data.


#3

@mathan_leo : Hey this is just a transactional data that i am testing on, For example A company has 4 years (from 2013) of transactional data and based on this i have created five different segments like Active High value customers, Active Low Value , Warm , one time buyers and inactive customers. For example, Active High value customers have an average Monetary value of $1000 and has an average frequency of 10 and they return once in every 3 months to buy, considering this scenario we also have some new customers who have all these attributes like they purchased over $1000 and bought very recently, whom we consider as a new customer, so in this case these customers dont have very good frequency, they are falling under warm category, so my question here is that should we create any rules for these new customers ?


#4

Yes, may be you will have to adjust the weight for ‘Monetary’ as you are concerned about value. So that new customers with high value also be classified into a better group.


#5

Hey,

How can we adjust the weights for Monetary component alone? When I am
actually normalizing all the three components.


#6

Have a look at this… Weights can be given for each factor


#7

Hi @chaituchaitanya28 !
I guess you are trying to find dormant customers. Only consider those customers who were engaged for at least 6 months or so. This way your analysis will not be unfair and skewed. You can ignore the new customers who are engaged within the 6 months period.

I hope you find this useful.


#8

Hello akasmat,

This is what I was looking for, but don’t you think that time period (
whether it is 6 Months) or not should be determined by data, for example if
I start looking at my top segment customers they are engaged from first
year till last year. What could be the other best way to actually find out
that optimum time period where I can say that the customers who are not
atleast engaged for these many months are excluded from this analysis.


#9

For that very reason, many people prefer the RF score. This is discussed regularly and people seem to agree that the RF score is the strongest indicator. It kind of levels the playing field.


#10

Another point on the RFM scoring, it goes way back, predating computers. It was used to find the best of the 125 resulting segments to market to. It identified the segments with the highest propensity to respond to a direct mailing, usually a catalog. It wasn’t always what we would intuit. Your answer can best be found in the history of RFM.


#11

@chaituchaitanya28
6 months was merely an example. You’ve to find time period by considering your customer transaction trends and hence find the suitable time period for your data


#12

I’m not sure, but if you will use another derivative metric, such a “speed” of monetary score: money spent by time period, then you’ll get not only new buyers in high group, but find those, which spent much money for some period in past, and do not do so now (relocated or ran out of money).


#13

New customers should be removed from RFM segmentation and treated as a separate segment so they can be handled differently