Health scores built on CRM activity have a specific failure mode that's uncomfortable to name directly: they measure what gets logged, not what's actually happening. A CSM who is diligently logging their outreach, scheduling QBRs, and recording call outcomes is generating activity data that looks like a healthy account relationship — regardless of whether the customer is actually satisfied or whether they're planning to renew.
This isn't a criticism of CRM discipline. Logging activity is valuable for many reasons. The problem is using logged CSM activity as a proxy for customer health, while simultaneously ignoring the support channel where the customer is expressing their actual experience with the product.
The Structural Lag in CRM-Fed Health Scores
Most health scores in Gainsight, HubSpot, or Salesforce-based CS programs incorporate some combination of: last CSM outreach date, NPS score (if recent), executive sponsor contact recency, renewal date proximity, and product usage (when instrumented). This is a reasonable set of inputs. But notice what's not in that list: what the customer said to your support team in the last 30 days.
The support channel captures a fundamentally different stream of customer sentiment than any CRM activity log can. When a customer is frustrated with your product, they almost never tell their CSM directly in the first instance. They submit a support ticket. Then another. Then they escalate. By the time a customer is telling their CSM "we're reconsidering our renewal," that sentiment has already been expressed repeatedly — and increasingly urgently — in the support queue.
The CRM health score that's been getting updated through CSM activity may still look green during this entire window. The CSM had a QBR three months ago: green. They sent a check-in email last month that got a neutral reply: green. The executive sponsor responded to a Slack message two weeks ago with a thumbs up: green. Meanwhile, the customer's operations team has opened 6 tickets in 45 days, two of them escalated to Level 2, and their CSAT score on resolved tickets has dropped from 4.0 to 2.4.
The health score is green. The customer is leaving.
A Pattern We've Seen Play Out
Take a customer experience SaaS with about 380 B2B accounts — a profile like Orion CX in their CS infrastructure build-out. Their health scoring methodology: usage score (40%), CSM activity recency (30%), NPS score (20%), renewal proximity (10%). A mid-sized account — roughly $42K ARR — had been showing usage at 85% of baseline, CSM activity logged two weeks ago, and an NPS of 7 from 4 months back. Overall health score: 72 out of 100. Healthy.
What the health score didn't see: 3 escalated support tickets in the prior 6 weeks. The primary contact's tone in ticket text had shifted from politely frustrated to professionally pointed. One ticket had been open for 19 days without resolution. The account's support CSAT had dropped 1.8 points over the quarter.
The account didn't renew. When the CSM circled back to understand the decision, the primary contact cited the unresolved support issues as the deciding factor. "We didn't feel like a priority." The health score had been green until the renewal conversation started to go sideways.
Why Support Sentiment Is Structurally Separated from CS Health Scoring
There's a real organizational reason this gap exists: in most companies, the CS team owns the CRM and the health score, while a separate CX or support team owns Zendesk or Intercom. The data lives in different systems with different owners, and bridging them requires either engineering work, a shared data warehouse, or a platform that ingests both streams.
This separation is understandable. It's also a meaningful risk if your health score is supposed to predict renewal outcomes. The customer doesn't care which team owns which system. From their perspective, they've been telling your company for six weeks that they're having problems. The fact that those communications went into a separate system that your health score doesn't read isn't their problem — it's yours.
What a Dual-Signal Health Score Actually Sees
When support ticket data flows into account health scoring alongside usage telemetry, the picture changes materially. A few specific signals that are only visible from the support side:
- Ticket volume trend: accounts that open significantly more tickets than their historical baseline in a 30-day window, even if usage is stable
- Escalation rate: tickets that get escalated from L1 to L2, or that get CC'd to a manager — these represent frustration that has moved past the standard workflow
- Open thread aging: tickets that remain unresolved for 10+ days often correlate with accounts where the customer feels ignored, regardless of whether the issue is complex
- Sentiment text patterns: NLP scoring on ticket text can flag language patterns that indicate eroding patience — "still waiting," "as I mentioned previously," "I need to escalate this" — before those patterns show up in an NPS score or direct CSM feedback
None of these signals requires the customer to explicitly tell the CSM they're unhappy. They're behavioral signals in the support channel — the place where customers express dissatisfaction in real time rather than on a quarterly survey.
The NPS Problem in This Context
NPS makes the health score problem worse because it provides false confidence at the wrong frequency. A quarterly or semi-annual NPS survey gives you a lagging sentiment reading at a moment in time. An account with an NPS of 7 six months ago is not necessarily an account with NPS of 7 today — and if their support experience has degraded significantly in the intervening period, you won't find out until the next survey cycle, which may be after the renewal decision has already been made.
Support ticket sentiment is updated continuously. Every ticket is a data point. That's the fundamental advantage of the support channel as a signal source: it's not periodic, it's real-time, and it's unfiltered. Customers don't modulate their support communication through the same social filter they apply to NPS responses.
Not a Replacement — An Addition
We're not arguing that CRM-based health scores should be abandoned. The CSM activity data, renewal proximity weighting, and usage telemetry in those models are genuinely valuable inputs. We're arguing that a health score that doesn't incorporate support sentiment is measuring half the account relationship and systematically missing a category of at-risk accounts until the risk becomes visible in the wrong way — at the renewal conversation instead of 60 days before it.
The practical question for CS leaders: for your last ten non-renewals, what was the support ticket history in the 90 days before renewal? If the answer is "I don't know, we didn't look at that" — that's worth investigating before the next renewal cycle runs.