The math is uncomfortable but unavoidable: if you're a CSM managing 250 accounts, and your average account engagement call takes 30 minutes including prep, you cannot have monthly contact with your full portfolio. The hours don't exist. Something has to give — and in most CS organizations, what gives is the informal distribution of attention: the loudest accounts get heard, the quietest accounts get checked in on when there's time, and the middle-tier accounts get whatever's left over.
The problem with informal attention distribution is that it's not correlated with risk. The quietest accounts aren't necessarily healthy — they might be disengaged to the point of not bothering to complain. The loudest accounts might be demanding because they're power users, not because they're at risk. Risk segmentation replaces informal attention distribution with a systematic one: CSM energy goes to the accounts where intervention is most likely to prevent ARR loss.
The Four-Tier Risk Segmentation Model
Most risk segmentation frameworks use three to five tiers. At Vendarix, we've found four tiers to be the right granularity for most CS teams managing 200-500 accounts — specific enough to drive meaningfully different playbooks, coarse enough to not overwhelm CSM decision-making.
Tier 1 — Critical: Health score below 40, or health score below 55 with a declining trend velocity greater than 5 points per week, or renewal in under 30 days with health below 60. These accounts are actively at risk of churn within the current or next renewal cycle. CSM intervention is immediate and escalated — this is the list you review daily.
Tier 2 — At-Risk: Health score 40-64, or health score above 64 with a consistent declining trend, or renewal in 60-90 days with health below 70. These accounts are showing behavioral signals that suggest churn risk is elevated but not yet critical. Proactive CSM contact within the week, with a specific agenda tied to the signals driving the score down.
Tier 3 — Watch: Health score 65-79, or health score above 79 with a recently declining trend. These accounts are in range but trending toward risk. Automated playbooks handle most of the touchpoints here — a check-in email, a feature adoption prompt, a renewal timeline reminder — with the CSM reviewing the playbook activity weekly rather than actively managing every account. CSM attention is reserved for accounts in this tier that have high ARR or strategic importance.
Tier 4 — Healthy: Health score above 79 with a stable or improving trend. These accounts are on track. Automated renewal timeline playbooks keep them engaged as they approach renewal. CSMs review this tier monthly, not weekly. The attention budget here is expansion identification, not retention defense.
Why Tier Assignment Alone Isn't Enough
The tier framework is necessary but not sufficient. Two issues emerge in practice that require additional structure:
ARR weighting: A Tier 2 at-risk account at $2k ARR deserves different CSM attention than a Tier 2 account at $40k ARR. Pure tier-based routing without ARR weighting will have your CSMs spending equal time on accounts with very different business impact. The practical fix is a two-axis prioritization within each tier: risk tier on one axis, ARR tier on the other, with the high-risk/high-ARR quadrant getting the first hour of every CSM's day.
Trend velocity: A static health score misses the directionality of risk. An account at 62 (At-Risk tier) that was at 55 last month is trending in the right direction — it may need monitoring but is less urgent than an account at 62 that was at 74 last month. Trend velocity should be a primary sort key within each tier, not just the score itself.
Building the Portfolio View That Makes Segmentation Actionable
Tier classification is only useful if it drives a daily CSM workflow. The most common failure mode we see is teams that run excellent risk segmentation analysis once a quarter and then return to informal attention distribution the other 89 days. Segmentation needs to be a live operational view, not a periodic reporting exercise.
The portfolio view that makes segmentation actionable has a few key properties:
It's sorted by priority action, not by account name or CSM assignment. The default sort should surface accounts that need action today: Critical tier accounts with high ARR, At-Risk accounts with renewal dates approaching, and Watch tier accounts where a playbook just fired and needs CSM review.
It shows trend direction alongside current tier. A single-row display for each account should include: account name, ARR, current health score, tier, 30-day trend direction (up/flat/down), days to renewal, and active playbook status. That information set is sufficient to prioritize the day without opening individual account records for most accounts.
It updates daily, not weekly. Behavioral signals that move accounts between tiers can develop quickly — a login frequency drop and support ticket cluster can push an account from Watch to At-Risk in 72 hours. Weekly portfolio reviews create a lag that means you're acting on a 7-day-old picture of a situation that moved 5 days ago.
A Scenario: How Risk Tiers Change Over a 30-Day Window
Let's walk through what a 30-day snapshot looks like for a hypothetical CS team managing 310 accounts across two CSMs:
At the start of the month, the portfolio breaks down roughly as: 14 Critical, 38 At-Risk, 84 Watch, 174 Healthy. That's realistic — about 17% of accounts in the risk zone at any given time for a mid-growth SaaS business.
Over the next 30 days, the tier distribution shifts. Eight accounts move from Watch to At-Risk because their usage trends declined further. Three accounts move from At-Risk to Critical — one of them because a support ticket cluster triggered the threshold. Two accounts move from Critical to At-Risk because a CSM engagement stabilized their health score trend. Four accounts that were At-Risk at month start churned at renewal — none of them were in the highest ARR tier, but their combined ARR was meaningful.
The insight from this kind of dynamic tracking isn't just about individual accounts. It's about the portfolio-level flow rate: how fast are accounts moving through the risk tiers? If 20% of your Watch-tier accounts transition to At-Risk in a 30-day window, either your tier thresholds are miscalibrated or you have a systemic product or support issue driving a broad cohort toward risk simultaneously. That's a different problem than isolated account churn and requires a different response.
The ARR Protection Calculation
Risk segmentation's value is ultimately an ARR protection calculation: if you intervene in 70% of your At-Risk accounts and save half of those interventions, what's the ARR impact versus a scenario where you intervene in 30% because CSM attention is spread too thin?
The math is specific to your account economics — your average ACV, your base churn rate, your intervention success rate — but the directional answer is consistent: focused attention on the highest-risk, highest-value accounts produces better GRR outcomes than uniform attention distributed evenly. That's not a surprising finding. What's surprising is how few CS teams operate with the data infrastructure to actually implement it.
We're not saying that low-ARR accounts in your portfolio don't matter. An account at $3k ARR is still a customer who chose you and can refer others, write reviews, and expand if their business grows. The segmentation model doesn't say "ignore Tier 3 and 4 low-ARR accounts" — it says "let automated playbooks handle most of their engagement and reserve your CSMs' direct attention for situations where human judgment makes a material difference." That's a different claim, and it's one that holds up when you look at where CSM time actually moves the needle on retention outcomes.
Starting Small: Segmentation Without a Full Platform
You don't need Vendarix to run basic risk segmentation. If you're early in operationalizing CS, a manually-maintained spreadsheet with health scores updated weekly, tiered into four buckets, sorted by ARR within each tier, and shared in a daily CSM standup is a meaningful step up from informal attention distribution.
The limitations show up quickly: manual scoring is slow and inconsistent, you miss the behavioral signals you can't see in a spreadsheet (support sentiment, login frequency trends, billing page visits), and tier classifications go stale between weekly updates. That's where a signal-integrated health scoring system earns its cost — not by inventing a better concept than risk segmentation, but by making it run continuously, automatically, and with richer inputs than a CS team can reasonably maintain by hand.