The Problem with RFM

Every data scientist working in retail, e-commerce, or loyalty programmes has used RFM. Recency, Frequency, Monetary value — the holy trinity of customer segmentation. Clean. Intuitive. Widely adopted.

And yet, after decades of use across industries, RFM systematically misses an entire class of high-value customer. I call them Ghost Flyers.

They don't appear in your top-value segments. They don't trigger your retention campaigns. They don't show up in your "at-risk" alerts. But they represent some of the highest unrealised revenue sitting dormant in your customer base.

"The most dangerous number in analytics is the one that looks right but captures the wrong thing."

What Is a Ghost Flyer?

A Ghost Flyer is a customer who purchases infrequently but consistently, at high value, with long but predictable inter-purchase intervals.

Think of a traveller who books business class internationally twice a year. Every year. Like clockwork. Under RFM, they are "low frequency, moderate recency" — likely sitting in a mid-tier segment, receiving mid-tier treatment.

In reality, they are one of your most loyal and profitable customers. They just don't look like it through the lens of traditional RFM.

Why RFM Gets This Wrong

RFM scores are calculated at a point in time. They are inherently cross-sectional — a snapshot, not a film.

The frequency score penalises the Ghost Flyer because it compares them to high-frequency purchasers in absolute terms. A customer who buys once a month scores higher on Frequency than one who buys twice a year — regardless of spend or loyalty duration.

The Recency score further disadvantages them if they're in the gap between purchases. A Ghost Flyer who bought six months ago, and whose next purchase is six months away, looks identical to a customer who churned six months ago. RFM cannot tell the difference.

Introducing RFMC — The Fourth Dimension

The solution is to add a fourth dimension to the framework: Cadence (C).

Cadence measures the consistency and predictability of the inter-purchase interval. It answers a fundamentally different question than Frequency: not how often do they buy, but how reliably predictable is their buying rhythm?

A customer who buys every six months with a standard deviation of ±2 weeks has a very high Cadence score. A customer who buys sporadically — sometimes monthly, sometimes annually — has a low Cadence score regardless of their Frequency.

Calculating the Cadence Score

The Cadence score is derived from the coefficient of variation (CV) of the inter-purchase interval:

  • Compute the mean inter-purchase interval (μ) for the customer
  • Compute the standard deviation (σ) of those intervals
  • CV = σ / μ
  • Cadence Score = 1 / (1 + CV), normalised to a 0–1 scale

A Cadence score approaching 1.0 means perfectly predictable purchasing rhythm. A score approaching 0 means chaotic, unpredictable intervals.

"Cadence is the heartbeat of loyalty. RFM measures the volume. RFMC measures the rhythm."

Identifying Ghost Flyers

Ghost Flyers emerge at the intersection of four criteria:

  • Low Frequency (bottom 40% of your customer base)
  • High Monetary value (top 25%)
  • High Cadence score (top 30%)
  • Current Recency in the expected inter-purchase window (within μ ± 1.5σ)

When you apply these filters, you will typically find between 3–8% of your customer base qualifies as Ghost Flyers. This sounds small. It isn't. Because these customers have already demonstrated long-term, high-value, rhythmic purchasing behaviour — they represent the most stable and predictable revenue in your base.

What To Do With Them

Once identified, Ghost Flyers require a fundamentally different treatment strategy than your other segments.

Do not over-contact them. The most common mistake is to apply standard retention or re-engagement campaigns to Ghost Flyers. They are not at risk of churning. Bombarding them with promotions based on their apparent "low frequency" can actually disrupt their natural purchasing rhythm and create unnecessary price sensitivity.

Acknowledge their rhythm. The most effective outreach is to contact Ghost Flyers at the predicted moment of their next purchase — not randomly. If their mean inter-purchase interval is 180 days, reach out at day 160 with a premium, personalised message. Not a discount. A recognition.

Protect them from re-segmentation. In dynamic RFM models that update monthly, Ghost Flyers will frequently drop into low-priority segments during their natural "away" period. Build a rule to protect RFMC-flagged customers from aggressive re-segmentation.

A Real-World Example

At a major retail bank I worked with, we applied the RFMC framework to credit card customers. Traditional RFM had classified a segment of 42,000 customers as "dormant" and recommended de-prioritising them in marketing spend.

RFMC identified that 6,200 of these customers — 15% of the "dormant" group — had a Cadence score above 0.72 and a mean transaction value 3.4x higher than the overall active base.

These were Ghost Flyers. They weren't dormant. They were between cycles.

When we shifted our treatment strategy for this group — personalised premium outreach timed to their predicted re-activation window — transaction reactivation rate improved by 38% and average transaction value increased by 22% compared to the standard dormant re-engagement campaign.

The Broader Lesson

RFMC is not just a technical refinement. It represents a shift in how we think about customer loyalty.

Loyalty is not just frequency. Loyalty is reliability. A customer who returns every six months without fail is more loyal — in any meaningful sense of the word — than one who shops every month for six months and then disappears.

RFM was built for a world where high frequency was the primary signal of value. In a world where customers have infinite alternatives and make deliberate, considered purchasing decisions, the Ghost Flyer is not an anomaly. They are the future of high-value customer behaviour.

Build your models accordingly.

This article draws on work conducted across banking, retail, and e-commerce environments spanning multiple countries. The RFMC framework has been independently implemented across multiple organisations with consistent results.