Stop Treating Your SaaS Users Like a Single Customer

I spent three months last year watching a client's churn rate stubbornly sit at 18% while their retention team sent the same "we miss you" email to literally everyone. Same template. Same timing. Same message. I kept thinking: these people are engineers—they understand distribution, variance, edge cases. Why are they treating their entire user base like a single data point?

Then I realized they weren't alone. Most B2B SaaS teams I've worked with operate on a one-size-fits-all retention model because segmentation feels expensive and complicated. Building different playbooks for different customer types sounds good in theory but requires coordination between engineering, product, and support that most teams don't have. So they default to broadcast.

That's when I read about RFM segmentation adapted for SaaS, and it clicked. This isn't some academic framework—it's a practical lens for actually seeing the distribution you're working with.

RFM Exists, but It's Built for the Wrong Business Model

The original RFM framework (Recency, Frequency, Monetary) comes from retail. Buy often, buy recently, buy expensive = valuable customer. Simple. Clean. Completely useless for B2B SaaS.

Here's why: in e-commerce, a purchase is the whole event. In SaaS, purchase is just the start. A customer might have paid for five seats but only two people log in. Another customer logs in daily but hasn't upgraded in months. A third customer pays $2,000/month but is only using 5% of their capacity—which means there's massive expansion potential hidden under a "looks healthy" MRR number.

Classic RFM would miss all of this. It would also flag reactivation way too late. By the time login frequency drops enough to trigger an alert, you're often 60–90 days into a death spiral that costs more to save than the annual contract value.

Redefining the Metrics Actually Matters

The article's reframing of each dimension for SaaS is where things get real. Instead of "last login," you track last meaningful action—a core workflow, a new user added, an actual interaction with support or a QBR. That's the difference between "someone opened our app" and "someone is actually getting value."

Frequency gets split into breadth and depth: what percentage of licensed seats are active, and how many core features are they actually using? This catches accounts that look dead (nobody logging in) and accounts that look thin (everyone using one feature only). Both are retention risks, but you'd fix them differently.

Monetary is where I genuinely found something I'd been underweighting. Most SaaS teams look at MRR. A $200/month account at 10% of its capacity—with headroom to add more teams—is actually higher-value than a saturated $2,000/month account with no expansion path. The second one's a churn waiting to happen.

The 11-Segment Model Is Workable, Not a Silver Bullet

I like that the article doesn't pretend this is magic. Eleven segments mapped across three dimensions, each with a different playbook. Champions get referred to your beta program and referral network. High-value sleepers get immediate personal outreach because something broke recently. Accounts in your "needs activation" bucket get product training, not a sales call.

This is straightforward enough to actually build. You need your analytics SQL right, your metrics defined consistently, and a way to automate segment assignment. Then you can start varying your CSM approach, email cadence, and upsell timing by segment instead of blasting everyone identically.

Where I'd Push Back

Here's what bugs me: this framework still assumes you have good product analytics and clean usage data. A lot of B2B SaaS companies—especially ones that have been around for a few years—have fragmented event tracking, unclear definitions of "meaningful engagement," and historical data that's hard to trust. You can't accurately measure feature adoption if you've been inconsistent about what counts as a feature.

Also, the model weights recent engagement heavily, which makes sense for churn prediction but can undervalue accounts that are just slow-moving. Some enterprise customers work in 6-month sales cycles. Their recency score will be terrible, but they're not churning—they're just operating at a different tempo.

My Next Move

I'm implementing a version of this for a client next month. We're starting with four segments instead of eleven—champions, growth trajectory, at-risk, and dormant. Simpler to execute, easier to validate that the playbook differences actually move retention numbers.

What's your retention model look like? Are you segmenting at all, or still running the same play for everyone?

Source: This post was inspired by "RFM Segmentation for B2B SaaS: An 11-Segment Model to Cut Churn (with Python)" by Dev.to. Read the original article