Dynamic RFM Analysis for B2B

This is the second of three articles on Recency Frequency Monetary analysis for B2B marketers. In this one, we'll look first at the 3D RFM space in terms a of a single analysis; then move on to the advantages of regular analysis to view dynamic (time-series) data.

[If you're new to the subject, the link to the first article "A beginners guide to Recency Frequency Monetary analysis for B2B marketers" is here. This explains how to create a single RFM analysis.]

New Customers begin at R5,F1,M1

New Customers begin at R5, the most recent period. They have probably bought a single purchase of a small volume, so F=1, and M=1 as well. If they keep buying every period, they will maintain the R=5 score and hold the position on the front right face of the cube.

And if, over time, they steadily increase both order frequency and value, they may ultimately end up in the top right corner - where the RFM score of 5,5,5 identifies them as one of the small handful of customers in the prime position on the Pareto Power Curve.

The shaded plane represents the points where (Total Monetary / Number of Orders) = Average Order Value. So one way to become a Key Account is by progressing from 5,1,1 through segments 5,2,2 then 5,3,3 onwards to 5,4,4 to finally reach 5,5,5.

But there are two other paths to the 5,5,5 position. The lower dotted arrow shows that at each level of increasing monetary value, progress through the quintiles is made via a large volume of orders, each of below average value. By contrast the upper dashed arrow shows that progress through the quintiles is made via lower frequency of orders, each comprising above average value.

Is it important? It might be. It depends on your business model and the impact of administrative costs or overheads on margin.

Your most valuable customers are here

In terms of total monetary value, the 5,5,5 segment will always be worth more than the 5,4,5, which in turn is more valuable than the 5,3,5 ... down through the 5,2,5 segment to the 5,1,5 segment. However ... dividing the total revenue for the segment by the actual number of orders gives the average order value for the segment. And this is where the 5,1,5 segment may prove to be worth careful attention. These are the infrequent buyers who place outsize orders.

If your products are consumables, or cover multiple categories, then there is probably a very wide range between the minimum and the maximum number of orders placed within the time period covered by an analysis.

But the meaning of the frequency and recency pattern is different when you're selling durables or investment goods. In this situation, the re-purchase of the product may be influenced by tax write-off periods and measured in cycles of five or seven years. The impact is this: even with an annual RFM exercise, over a five-year time span, the 5,1,5 customer will appear to slip ever further backwards on the recency score.

Dropping from 5,1,5 (rank 21 of 225 segments) through 4,1,5 to 3,1,5 and 2,1,5 down to 1,1,5 (rank 121 of 125 segments) may appear to be a disastrous situation. But, once the re-purchase cycle comes around again, and the old machinery is replaced with the latest generation of technology, the customer will rocket forward to the front of the cube at position 5,1,5 again. And then the cycle will repeat.

Understanding these types of movement through three dimensions is a key part of using RFM analysis effectively in B2B marketing.

RFM reveals the Lifecycle stage

Some customers transition through the customer lifecycle faster than others. This speed of transition through the lifecycle stage is not necessarily the same as their Lifetime value.

Some examples will illustrate the difference. First, we'll sketch the path taken by an Average Customer (the dark blue dotted line):

• New customers begin the lifecycle with a Recency of 5; the first order is often a trail purchase, so the combined RFM values are 5,1,1 - the grey box at the front of the diagram.
• The typical repeat buyer will maintain their position on the front right face of the cube with a Recency score of 5; as the frequency of orders increases, so do the monetary values. The average customer progresses from 5,1,1 through 5,2,2 to a score of 5,3,3 (the green box).
• If the customer misses a regular order, the overall recency value declines: the customer will slip back from the front right face of the cube, to a middling value of say R=3 (the yellow box).
• Customers who cease ordering will gradually slip back through the rankings from a central position near 3,3,3 to a RFM score of 1,1,1 (red box).

The lifecycle for Key Accounts is shown in the light blue dashed line:

• After a trial order at 5,1,1 the account will maintain the R=5 score at the front right face of the cube. Over time it will typically grow in both frequency and monetary value until it reaches the 5,5,5 position.
• Key accounts in this position are at the Power Curve section of the Pareto distribution. Generally speaking they tend to stay here for a very long time.
• Within the Power Curve segment - the green area F4,M4 to F5,M5 - individual customers will rise and fall according to economic or industry pressures. The light blue loop indicates that active, ongoing nurturing is necessary to keep these customer relationships alive and healthy.
• Any customer moving outside this 'sweet spot' should be taken as a prompt for immediate analysis and remedial action. The long-term path is a decline in frequency and value of orders, an aging recency profile and a pre-exit score of 1,1,1.

Analysing a single RFM extract

A single RFM analysis is a snapshot in time: it's a static image. Analysing this data will identify cohorts within your customer database that exhibit similar sales behaviour. These patterns can in turn suggest the messages that are most likely to grab their attention. Examples include but are not limited to: identifying target groups for Cross-sell or Up-sell campaigns; identifying At-Risk customers prior to defection; identifing recently-lost customers for Win-back campaigns; and identifying candidates for loyalty rewards.

Some observations:

• The list of solid gold loyalists will always include some small organisations. Don’t ignore them just because they only make a minor contribution to total revenue.
• Some organisations rocket to major account status quite quickly. Be aware that they can fade away equally fast.
• Some customers are only ever of middling importance – and there’s nothing wrong with that. If they are loyal repeat buyers, they form the core of reliable revenues.

• Some very loyal long-term customers are irregular in their buying habits. It’s good to be able to identify these folks, because – while we may be tempted to think of them as “unreliable” - they think of us as their “preferred supplier”. They typically have a variable Recency (1 to 5) combined with consistently low Frequency, but high (4 or 5) Monetary.

The majority of customers are consistent in their habits and stay within the same cohort for a long time. Against that background, there is a relatively small minority of customers who move around the RFM landscape.

?

From static to dynamic RFM

While a single RFM analysis is essentially a snapshot taken at a specific time, comparing snapshots enables us to identify transitions over time. If you can observe when customers transition between cohorts then you've got a time trigger for talking to them. If you can understand the direction and velocity of travel between cohorts, you can figure out intent and focus the messaging to them with greater relevance.

Some general comments for clarification:

• R score: increasing = good | decreasing = warning
• F score: any change = compare against M to evaluate Average Order Value
• M score: increasing = good | decreasing = warning
• Score increases to 4 or 5 (any axis) = good (entry to Power Curve)
• Score drops below 4 (any axis) = warning (exit from Power Curve / slippage to Long Tail)

Shifting the focus from project to process

To identify those transitions between segments, we need to develop a way to take snapshots more frequently and with less effort. In short, we need to shift the focus of RFM from being an “occasional project” to a “regular process”. This increases the predictive value of RFM.

?

Regular RFM analysis brings added advantages

First, you get to see the data as a time series: comparing 'then' with 'now' shows patterns and directions of movement.

Second, you get the advantage of focus: many of the customers will fall into the same classification, time after time. But by attending to the critical few who transition from one category to another, you can design re-usable marketing and sales interactions for greater effectiveness.

Third, you get to turn a one-time project into a process. Regularity and consistency will make the exercise more time-efficient. Repetition will make the analysis more effective.

Add those together and you can design more effective campaigns (target group plus relevant message); and implement them at the critical moment they are needed.

Combine these factors - simplification plus repetition - with standardisation, and we have the foundations for taking marketing effectiveness to the next level via automation.

If you've enjoyed this / found it useful ... share it with a LIKE

You'd like to explore RFM Analysis in practice? You want a Guide & Mentor? CONTACT ME

?