Churn cost at $100M+ ARR: what the public 10-Ks actually disclose

At $100M+ ARR your retention number is no longer a slide in a board pack. It is a footnote in a 10-K, a line item in analyst models, and the single largest input into your forward multiple. Snowflake's NRR going from 158 percent at IPO to 127 percent in FY2024 cost the company roughly five turns of multiple, on the order of $30B+ of market capitalisation across the cycle. The three-layer cost framework still applies; the consequences are now measured in nine and ten figures rather than seven.

Headline number. 0.5 percent monthly churn on $100M ARR equals approximately $42K of MRR lost monthly. Annualised three-layer cost: $12M to $23M, roughly 12 to 23 percent of ARR. Multiple-compression effect on enterprise value: typically $100M to $400M depending on NRR band.

What the top public SaaS 10-Ks actually disclose

Public SaaS retention disclosure follows a fairly standardised pattern. The 10-K typically discloses a "net dollar-based retention rate" or "net retention rate" in the MD&A section, defines the calculation methodology precisely, presents the trailing four to eight quarters of values, and references it as a key metric in the risk-factors and forward-looking sections.

Worked summary of the comps an aspiring public SaaS at $100M+ ARR should know:

Snowflake. Disclosed NRR of approximately 158 percent at IPO in September 2020. Peaked at 178 percent in early 2022. Declined to 142 percent in early 2023, 131 percent in early 2024, and 127 percent in their FY2024 10-K. The decline mirrors the broader compression in consumption-based SaaS as customers optimised cloud spend post-2022. Even at 127 percent, Snowflake's NRR remains comfortably above the public SaaS median.

Datadog. Reported NRR consistently above 130 percent from 2019 through 2022, declining to 115 percent in their 2024 10-K. Like Snowflake, Datadog's consumption-based model amplified both the upside in 2020-22 and the compression in 2023-24.

CrowdStrike. Reports NRR consistently in the 119 to 124 percent range across multiple years. Subscription model rather than consumption, so less volatile than Snowflake or Datadog. Demonstrates that high NRR is sustainable at scale with a seat-based model and a multi-product upsell motion.

ServiceNow. Does not disclose NRR specifically. Reports a renewal rate of 98 to 99 percent annually, which is GRR-equivalent. Multi-product cross-sell drives substantial expansion, but they choose not to publish a single NRR figure. The 10-K instead emphasises customer concentration metrics and Now Platform attach rates.

Salesforce. Discloses attrition rates of roughly 8 to 10 percent annually in their 10-K, which implies GRR in the 90 to 92 percent range. Salesforce does not publish NRR, but the implied figure from segment growth rates is around 105 to 110 percent at the consolidated level.

Three-layer cost on $8.3M MRR

A $100M ARR SaaS sits on $8.33M of MRR. Layer one at 0.5 percent monthly churn strips $42,000 of MRR each month, compounded to approximately $3.3M annually without expansion. The percentage rate is low (consistent with mid-market and enterprise medians at this cohort), but the absolute dollars are large.

Layer two at $100M ARR norms (CAC of $30K to $80K, ACV of $80K to $200K depending on segment mix) burns approximately $4M to $8M of acquisition cost annually on churned accounts. Layer three (forfeited expansion at 40 to 55 percent expansion share of MRR for healthy operators) destroys approximately $5M to $12M of annualised expansion value.

Total annual economic damage at 0.5 percent monthly churn: approximately $12M to $23M. At 1 percent monthly churn, the numbers approximately double to $25M to $45M. These are still small relative to the multiple-compression effect on enterprise value, which can move $100M to $400M of market cap per 5-point NRR slippage at this scale.

The arithmetic is what justifies the typical $10M to $30M annual investment in customer success at $100M+ ARR. A CS team that holds NRR one band higher than peers (say 120 percent vs 115 percent) is worth roughly $50M to $150M of incremental enterprise value at any given moment, which is 5x to 15x the CS budget on an annualised basis.

The public-comp multiple bands in 2026

Public SaaS multiples in early 2026, per Bessemer Cloud Index and Meritech Public SaaS data, anchored on NRR with remarkable consistency:

NRR above 130 percent: median EV / forward revenue of 12x to 15x. Cohort includes Snowflake, Cloudflare, CrowdStrike, MongoDB at various points.
NRR 115 to 130 percent: median EV / forward revenue of 7x to 9x. Cohort includes Datadog, ServiceNow, HubSpot, Zscaler.
NRR 100 to 115 percent: median EV / forward revenue of 4x to 6x. Cohort includes mature SaaS like Atlassian, Workday, Salesforce.
NRR below 100 percent: median EV / forward revenue of 2x to 3x. Cohort includes legacy SaaS in structural decline.

The bands are not tight bands in the statistical sense (there is wide dispersion within each, driven by growth rate and margin), but the median separations are large enough that NRR cohort effectively sets a multiple corridor. A $100M ARR business going from 116 percent NRR to 109 percent NRR moves from one cohort to the next and typically loses 2 to 3 turns of multiple, which on $100M of revenue is $200M to $300M of market cap.

Operational implications at public-SaaS scale

At $100M+ ARR the retention organisation is typically 50 to 150 people across customer success, renewal operations, professional services, and support. CS spend frequently exceeds 15 percent of ARR for enterprise-led operators, with separate budget lines for the CS platform (Gainsight or Catalyst, typically $1M to $5M annually), the renewal forecasting infrastructure, and the customer marketing function.

The renewal forecasting motion is what most distinguishes this stage. Renewal forecasts are produced at the account level, rolled up to segment, and presented to the CFO as part of the quarterly close. Renewal risk is reported alongside new-logo pipeline in board materials. CS contribution to revenue retention is treated as a financial KPI rather than a CS-internal metric.

The data infrastructure that supports this is non-trivial. Retention metrics calculated in the data warehouse (Snowflake or BigQuery), validated against the billing system, exposed to the CS platform via reverse ETL, and reported in finance dashboards. The infrastructure cost alone is typically $500K to $2M annually at this scale, just for the data plumbing.

What this means in practice: retention work at $100M+ ARR is a cross-functional operating system rather than a CS team initiative. CS, Finance, Product, Sales Operations, and Data Engineering all have specific responsibilities in the renewal motion. Companies that treat it as a CS-only function rarely sustain NRR above 110 percent at scale. Companies that institutionalise the operating model across functions are the ones holding 120 percent+ NRR through cycles.

Frequently asked questions

What NRR do top public SaaS companies report?+

Snowflake reported NRR of approximately 127 percent in their FY2024 10-K, down from 158 percent at IPO in 2020 and a peak of 178 percent in 2022. Datadog reported 115 percent for fiscal 2024, down from 130+ in 2020-22. CrowdStrike consistently reports around 120 percent. ServiceNow does not disclose a specific NRR but their renewal rate is consistently above 98 percent, with strong implicit expansion. These are useful comps for any SaaS approaching $100M ARR.

What does 110 percent NRR actually mean in a 10-K?+

A 110 percent NRR means a cohort of customers from twelve months ago is now paying you 110 percent of what they were paying twelve months ago, net of any churn or contraction. It includes expansion (more seats, more usage, upsells to higher tiers) offset by churn (entire customers leaving) and contraction (existing customers downgrading or reducing seats). It does not include any net new customer acquisition.

How much does 0.5 percent monthly churn cost a $100M ARR SaaS?+

On $100M ARR (approximately $8.3M MRR), 0.5 percent monthly churn loses approximately $42K of MRR each month. Compounded across 12 months without expansion: approximately $3.3M of direct revenue. Add wasted CAC of $4M to $8M and forfeited expansion of $5M to $12M. Total annual economic damage: $12M to $23M. At the multiple compression effect typical of public SaaS, the enterprise value impact is $100M to $250M.

Why does NRR compression matter for public SaaS multiples?+

Public SaaS multiples trade in tight bands by NRR cohort. Bessemer Cloud Index data shows that companies with NRR above 130 percent traded at median multiples of 12x to 15x forward revenue in early 2026, while companies with NRR between 110 and 125 percent traded at 7x to 9x. A 5-point NRR slippage that moves you from one cohort to the next typically costs 2 to 4 turns of multiple, which on a $100M revenue base equals $200M to $400M of market capitalisation.

What is the difference between gross retention and net retention at public scale?+

Gross retention is the percentage of revenue you retain before any expansion (i.e., subtracting only churn and contraction). Net retention adds expansion back in. Public companies disclose net retention prominently because it is the more flattering number, but underwriters and sophisticated investors triangulate gross retention from churn rates, customer concentration, and average revenue per customer growth.

What is the typical CS investment at $100M+ ARR?+

Industry benchmarks suggest CS spend of 8 to 15 percent of ARR for SMB-led operators and 15 to 25 percent of ARR for enterprise-led operators at this scale. At $100M ARR, that is $8M to $25M annually. The CS organisation is typically segmented across enterprise account management, mid-market success management, and SMB tech-touch, with separate platform spend on Gainsight, Catalyst, or equivalent.