Loyalty glossary · 7. Data and analytics (31)
Rfm Analysis
RFM analysis scores each member on recency, frequency, and monetary value, then groups them into segments for targeted action. It is the standard segmentation method in loyalty programmes, but its default version misreads accrual as value and ignores the difference between activity and engagement.
RFM analysis is the loyalty industry's standard way to break a member base into actionable segments, but its textbook form fails in programmes where money is stored as points. The three scores, recency, frequency, and monetary value, are meant to be independent axes. In practice, monetary value is often dominated by points accrual, which is a balance sheet liability rather than a sign of engagement. So the analysis can celebrate members who are actually a future cost.
Consider two members. The first spends 600 dollars in 3 months, an average of 200 dollars per month. The second spends 1,200 dollars in 12 months, an average of 100 dollars per month. RFM's monetary score ranks the second higher because the total is double, but the frequency and recency scores correctly show the first is more valuable right now. This is the arithmetic every programme manager should run before trusting a segment label.
Recency and frequency are not the same thing, and collapsing them into an active-member-rate hides this. A programme can report a healthy active member rate while most of those active members transacted only once in the last year. RFM analysis separates the recently active from the habitually active, and the two groups need completely different interventions. A recency campaign is a rescue operation; a frequency campaign is a growth one.
The monetary score's blind spot is redemption. A member accruing 50,000 points without redeeming any has a high monetary value, but that value is a liability on the programme's books. RFM analysis that treats accrual and redemption identically will rank hoarders above redeemers, which is backwards. Redemption frequency must be a separate input, or the segmentation will optimise for breakage without naming it.
Activity-based qualification makes the same error from the other direction. A tier rule that requires two stays per year will promote a member who made two cheap stays and ignore one who made five expensive stays but missed the activity rule. RFM analysis would rank the second member higher, which is usually correct. Qualification rules should follow RFM segments, not the other way around.
RFM segments are not static. A member in the champions segment today can be in the at-risk segment in 3 months if recency drops. This is why RFM analysis is a monitoring tool, not a one-off exercise. The scores must be recalculated at least monthly, because loyalty behaviour changes faster than tier years.