Most B2B companies treat customer segmentation like a binary: you're either active or churned. That's like a doctor classifying patients as either "healthy" or "dead" — technically accurate, and completely useless for intervention.
RFM analysis — Recency, Frequency, Monetary — has been a retention staple in B2C e-commerce for decades. But the B2B SaaS application of it is different, and most implementations I've seen get it wrong. They use a 5-segment model (Champions, Loyal, At Risk, Hibernating, Lost) and treat it as a one-time exercise.
What we run at Artefact Ventures is an 11-segment, AI-native RFM model that produces different interventions for each segment and runs continuously against live CRM data. Here's how it works, why the extra segments matter, and what to actually do with the output.
What RFM Means in B2B Context
Before the mechanics: a quick note on how to interpret the three dimensions for B2B SaaS specifically, because they mean something different than in e-commerce.
Recency — When did this account last meaningfully engage? In B2C, this is "last purchase date." In B2B, it should be the most recent high-signal activity: a logged call, an expansion conversation, a support ticket that resolved successfully, a renewal. Not just login activity — that conflates usage with engagement.
Frequency — How often does this account transact, expand, or engage at a decision-making level? In B2B, a customer who buys once and stays for 3 years at the same contract value is very different from one who buys once and churns. Frequency should capture touchpoints that indicate active relationship investment.
Monetary — Total revenue contribution, including expansion. Critically, this should be weighted by margin contribution where possible — a large account on a heavily discounted legacy plan is worth less than a smaller account at full rate.
If your RFM model doesn't account for these B2B-specific interpretations, you'll misclassify accounts and send the wrong interventions.
The 11-Segment Model: Full Breakdown
Standard RFM uses a 1-5 score on each dimension, producing 125 possible combinations. We bucket these into 11 actionable segments. Here's the complete model:
Segment 1: Champions
Profile: High recency, high frequency, high monetary value. These accounts bought recently, buy often, and spend the most.
What they need: Recognition, early access, co-creation opportunities. Champions are your best reference customers and your best source of product intelligence. They should never receive a generic nurture email.
Intervention: Executive relationship check-in. Early access to new features. Referral program invitation. Case study or co-marketing proposal.
Warning signal: A Champion whose recency score drops suddenly is your highest-priority churn risk — because the fall from Champion to At Risk is faster and more expensive than any other transition.
Segment 2: Loyal Customers
Profile: High frequency, strong monetary value, moderate recency. They buy consistently but haven't engaged recently.
What they need: Re-engagement with new value, not a sales pitch. They already like you. Show them something they haven't seen yet.
Intervention: Product update briefing. Invite to a user community or event. Expansion conversation anchored in their specific use case.
Segment 3: Potential Loyalists
Profile: Recent customers with above-average frequency, not yet in Champion territory.
What they need: Velocity. They're on the right trajectory — the goal is to accelerate the pattern without over-engineering the relationship.
Intervention: Onboarding optimization check-in. Feature adoption nudges. Loyalty program introduction if you have one.
Segment 4: Recent Customers
Profile: Bought recently, low frequency, lower monetary value. New accounts in early lifecycle.
What they need: A successful first experience. Everything else is secondary. If they don't get value in the first 90 days, they will not become Potential Loyalists.
Intervention: Active onboarding support. Success milestone tracking. First-90-days check-in with a human, not an automated email.
Segment 5: Promising
Profile: Recent purchase, low frequency, low monetary value. Early-stage, low commitment.
What they need: A proof point. One clear win that makes the relationship feel worth continuing.
Intervention: Use-case-specific success story. Quick-win workflow or template. Low-friction expansion offer (not upsell — proof-of-value first).
Segment 6: Need Attention
Profile: Above-average scores across all three dimensions historically, but recency is declining. These accounts were strong and are starting to drift.
What they need: Proactive contact before they self-identify as disengaged. This is the intervention that most companies miss because the account still looks healthy in a revenue dashboard.
Intervention: Proactive QBR or success review. Direct outreach from account owner, not CSM automation. ROI recalculation to re-anchor value.
Segment 7: About to Sleep
Profile: Below-average recency and frequency, but not yet lost. They're fading.
What they need: A reason to stay that they haven't heard before. Generic renewal reminders will not work here.
Intervention: Personalized re-engagement campaign based on their specific product usage history. Limited-time expansion offer. Direct conversation about fit — sometimes it's better to right-size than to retain at the wrong tier.
Segment 8: At Risk
Profile: High historical monetary value but declining recency and frequency. High-value accounts showing churn signals.
What they need: Urgent, executive-level attention. Not a CSM — the account executive or a founder, depending on company size.
Intervention: Executive sponsor check-in within 5 business days. Competitive displacement assessment. If they're evaluating alternatives, you need to know now, not at renewal.
Segment 9: Cannot Lose Them
Profile: Made large purchases historically but recency is very low. These accounts were significant and have gone quiet.
What they need: A genuine reconnection, not a retention script. Something went wrong — find out what before making any offer.
Intervention: Honest conversation about the relationship. Service recovery if applicable. Re-scoping the engagement to match current needs.
Segment 10: Hibernating
Profile: Low recency, low frequency, low monetary. Minimal engagement across all dimensions but not technically churned.
What they need: A decision. Either re-engage with a compelling reason or let the relationship end cleanly. Maintaining hibernating accounts in your pipeline creates false coverage.
Intervention: Sunset campaign with a clear value proposition. If no response after two touchpoints, move to offboarding and clean the CRM.
Segment 11: Lost
Profile: Lowest scores across all three dimensions. Churned or effectively churned.
What they need: A clean exit, a post-mortem, and a future win-back path if appropriate.
Intervention: Exit survey (keep it short — 3 questions max). Flag for win-back sequence in 6-12 months if the churn reason was situational rather than product-fit. Feed insights into ICP refinement.
The Scoring Mechanics
Here's how we score accounts for placement:
Step 1 — Score each dimension 1–5:
- 5 = top 20% of your customer base on that dimension
- 4 = 60th–80th percentile
- 3 = 40th–60th percentile
- 2 = 20th–40th percentile
- 1 = bottom 20%
Step 2 — Combine into an RFM string: An account scoring R=4, F=3, M=5 is a "435."
Step 3 — Map to segment: Use a lookup table to assign each RFM combination to one of the 11 segments. Champions are typically 554, 544, 545, 455, 454. Lost accounts are 111, 112, 121.
Step 4 — Run interventions per segment: This is the step most companies skip. The scoring is not the output — the intervention is.
Running This with AI: How the Artefact MCP Server Does It
Manual RFM scoring is a spreadsheet exercise that gets done once a quarter and immediately goes stale. The Artefact MCP RFM Analysis Engine runs this continuously against live HubSpot data.
A natural language query like:
"Run RFM analysis on my customer base and show me which segments have the highest churn risk this month"
Returns a structured output with:
- Current segment distribution across all 11 buckets
- Accounts that have moved segments since the last analysis (the transitions are the most important signal)
- Pre-built intervention recommendations per segment
- Priority ranking by revenue at risk
Install:
pip install artefact-mcp
claude mcp add artefact-mcp
Pro tier adds live HubSpot integration with custom RFM thresholds — meaning you can adjust the recency, frequency, and monetary weightings to match your specific business model rather than using generic defaults.
The Transition Matrix: What Really Matters
The segment score at a single point in time is less important than the direction of movement. Here's the transition matrix you should monitor:
| From | To | Priority |
|---|---|---|
| Champion → At Risk | Any drop | 🔴 Immediate |
| Need Attention → About to Sleep | Declining recency | 🔴 Immediate |
| Loyal → Need Attention | Frequency drop | 🟠 High |
| Potential Loyalist → Need Attention | Stalled frequency | 🟠 High |
| Recent → Promising | Frequency increase | 🟢 Positive |
| Promising → Potential Loyalist | Sustained engagement | 🟢 Positive |
A company that knows its segment distribution but doesn't track transitions is reading a photograph instead of watching a film.
Common RFM Mistakes in B2B
Mistake 1: Using login data as a proxy for recency. Login frequency measures access, not engagement. An account with daily logins and zero expansion conversations for 6 months is not a Champion — they're a habitual user who may be actively evaluating alternatives.
Mistake 2: Treating all segments with the same communication cadence. Lost accounts do not need weekly newsletters. Champions do not need basic feature education emails. Segment-specific cadence is not optional.
Mistake 3: Running RFM on a too-short time window. B2B purchase cycles are long. A 30-day RFM window will misclassify most of your customer base. For most B2B SaaS companies, use a 12-month look-back for Frequency and Monetary, and a 90-day look-back for Recency.
Mistake 4: Ignoring segment transitions at renewal time. If a customer was a Champion 12 months ago and is now in the "Need Attention" segment, renewing them at the same price with the same pitch will fail. The conversation needs to be different.
FAQ
What is RFM segmentation?
RFM segmentation is a customer analysis framework that scores customers on three dimensions — Recency (how recently they engaged or purchased), Frequency (how often they engage or purchase), and Monetary value (how much revenue they generate). The scores are combined to classify customers into segments that each require different retention strategies.
How many RFM segments should a B2B company use?
Most implementations use 5 segments. A more effective B2B model uses 11 segments — Champions, Loyal Customers, Potential Loyalists, Recent Customers, Promising, Need Attention, About to Sleep, At Risk, Cannot Lose Them, Hibernating, and Lost — because each requires a materially different intervention and the 5-segment model collapses important distinctions.
How often should I run RFM analysis?
Monthly at minimum for active intervention. Weekly if you have an automated system. The transitions between segments — not the current snapshot — are the most actionable signal, and those transitions happen continuously.
Can RFM analysis be automated with AI?
Yes. The Artefact MCP Server's RFM Analysis Engine runs against live HubSpot data and returns segment distributions, transition alerts, and pre-built intervention recommendations through a natural language interface. Install with pip install artefact-mcp.
What's the difference between B2C and B2B RFM?
B2C RFM uses purchase date, purchase count, and purchase value — all cleanly transactional. B2B RFM requires a more nuanced interpretation: recency should capture high-signal engagement, not just last purchase; frequency should reflect decision-level interactions; monetary should account for expansion and margin, not just contract value.
Alex Boissonneault is the founder of Artefact Ventures, a Québec-based firm building AI-native GTM systems for SMBs. He has 15 years of experience across seven major enterprise revenue transformations.
