Half the industry cut CSM headcount in 2025. The survivors aren’t the ones who defended the old model — they’re the ones who rebuilt it.
The era of abundant capital funded an industry-wide experiment in customer success. The results are now in — and they are uncomfortable reading for anyone who built their career managing a portfolio of accounts in a spreadsheet.
The bubble that built an industry
Customer Success grew to its current scale on the back of zero interest rates. With the Federal Reserve holding rates near zero, median SaaS EV/Revenue multiples ran from 4–5x in 2018–19 to nearly 19x by late 2021 (LongYield). Capital was abundant, growth was the only metric that mattered, and CS headcount became a visible signal of customer-centricity. CS organisations grew their expansion charter from roughly 10% of responsibilities in 2015 to 47% by 2020 (Customer Success Collective). It was, in retrospect, a function built as much for optics as for operational necessity.
Then the Federal Reserve raised rates by 525 basis points between March 2022 and July 2023. Companies began looking at their cost bases with a focus they perhaps should have applied all along. And they found the Customer Success organisation.
The numbers don’t lie
The data on the current state of Customer Success is stark. I want to note plainly: most research on CS effectiveness is published by companies that sell software to CS organisations. The figures below come from independent practitioner surveys with no financial stake in the conclusion.
In 2025, 44% of companies reported CSM layoffs — nearly half the industry simultaneously concluding the function is oversized — and 42% implemented hiring freezes (Customer Success Collective, State of Customer Success 2025). This is structural retrenchment, not a market correction. Meanwhile, 83% of CSMs still use Excel as their primary day-to-day tool, and 45% cite excess workload as a primary driver of job dissatisfaction (Custify, Customer Success Uncovered 2025). Burnout here is not cultural — it is structural.
The TSIA’s 2025 State of Customer Success report confirmed that tight budgets forced companies to fundamentally reassess their CS charters — and noted that many organisations are now quietly shifting back to sales tools rather than investing in dedicated CS platforms. They are reabsorbing Customer Success into existing functions rather than continuing to fund it as a standalone capability. They are not publishing press releases about this. They are just doing it.
The conclusion is inescapable: the CSM headcount boom was not driven by proven return on investment. It was driven by cheap capital, growth-at-all-costs investor mandates, and organisational momentum. When capital became expensive, the rationale evaporated.
The identity crisis at the heart of CS
The job cuts are a symptom. The underlying disease is a structural identity crisis that has never been resolved.
Customer Success occupies an uncomfortable middle ground — expected to deliver outcomes that require sales accountability, technical depth, and implementation discipline, yet lacking the structural characteristics that make those functions work. Practitioners who have studied CS model design closely argue that commission-driven CSMs reliably cluster their effort around renewal windows and abandon early lifecycle activities — onboarding, adoption, value realisation — when the financial incentive isn’t live. The result is neither a disciplined sales motion nor a genuine success model.
I’ve observed two organisational pathologies that recur almost universally.
The first is misplaced confidence in customer knowledge. CS relationships concentrate around contacts who respond to outreach — engaged users, cooperative champions, and accessible middle managers. These are not, as a general rule, the customers at risk of churning. More critically, CSMs tend to build relationships with accessible stakeholders rather than with the CFO, who controls the renewal, or with the executive sponsor quietly evaluating alternatives. When those people make a renewal decision against expectations, the CSM is genuinely surprised — not because they were negligent, but because they were systematically engaging the wrong people. This is a measurement problem masquerading as a relationship problem.
The second is the headcount reflex. When retention metrics disappoint, the near-universal response is to argue for more CSMs. But organisations that have added CS headcount and measured outcomes a year later frequently find that overall results have not materially changed, because the underlying model was not demonstrating causal impact. The correct question is not “how many CSMs do we need?” It is “what does CS activity actually cause, and how do we prove it?”
From relationship theater to analytical precision
The shift separating survivor organisations from those in decline is not a technology upgrade. It is a culture change — from human judgment as the primary instrument of customer intelligence to data-driven analytics as the operating foundation.
Traditional CS models assign CSMs only to accounts above a revenue threshold. The long tail — often the majority of the customer base by count — receives no proactive attention, not because companies have decided those customers don’t matter, but because the unit economics of human coverage make it impossible. The customers most likely to churn quietly are precisely the ones no CSM is watching.
AI-driven customer analytics changes this fundamentally. Login frequency, feature adoption patterns, support ticket velocity, and engagement trends can be monitored across 100% of the customer base simultaneously. Peer-reviewed research published in Scientific Reports in late 2025 demonstrated AI-driven churn prediction models achieving over 95% accuracy using behavioural and usage data alone (Abdelhady & Mohamed, Scientific Reports, December 2025). Separately, organisations embedding predictive signals in CS workflows have reported churn reductions of approximately 25% (G2, 2026) — yet 60% of organisations had not yet invested in AI for CS as of 2024 (TSIA, 2025).
When AI handles health scoring, usage monitoring, and risk flagging across the full base, the human CSM’s role becomes focused, strategic intervention with the accounts where human judgment genuinely changes the outcome. The ratio problem dissolves — not because you hired more people, but because the distribution of attention has become intelligent rather than uniform.
This shift requires CS leaders to dismantle narratives that have defined their function — the centrality of the human relationship, the primacy of advocacy, and the value of empathy as an operational strategy. These are not wrong values. They are insufficient operating principles when divorced from accountability and proof. The leaders who navigate this successfully will ask a different question: not “how do I make my CSMs more effective?” but “what does my customer data tell me, and what does it say I need to do?”
Four questions every CS leader must answer
- Can you demonstrate a causal — not correlational — relationship between CSM activity and retention outcomes? If NPS and renewal rates are influenced equally by product, support, and implementation, what is the isolated contribution of your team?
- What percentage of your CSM outreach is reaching economic decision-makers versus users and champions? Do you know which stakeholders actually control your renewal decisions?
- Does your team have visibility into behavioural signals for 100% of your customer base — or only the accounts above your CSM assignment threshold?
- At a fully-loaded cost of $140,000 to $200,000 per enterprise CSM, what is your cost per retained dollar of ARR? Does that ratio improve as you add headcount — or does it flatline?
The organisations that survive the current reckoning will not be those that defended the traditional CS model most persuasively. They will be those who rebuilt it most honestly — substituting analytical precision for relationship theater, and targeted human judgment for undifferentiated or absent coverage.
That is not the end of Customer Success. It is the beginning of a version of it that can actually earn its budget.