"NDR is Slowly Being Euthanized" — CJ Gustafson
Why this is in the vault
CJ Gustafson's audit of how public software companies are quietly killing net-dollar-retention disclosure, and why the act of hiding a metric is itself a louder signal than the number.
⚠️ Sponsorship
This issue is sponsored by Abacum (FP&A platform, newly a "Built for NetSuite" SuiteApp). The sponsor block is clearly demarcated at the top of the email ("Mostly metrics is proudly powered by Abacum") and does not bleed into the editorial argument. The essay also carries an affiliate link to data partner Koyfin in the recurring metrics appendix. No detectable bias in the NDR analysis itself.
The core argument
After NDR (Net Dollar Retention) was the GOATED metric of the 2019-2022 IPO era (Snowflake 177%, GitLab 152%, Datadog "above 130%" for fourteen quarters), companies have quietly stopped disclosing it as the number fell. Gustafson reviewed SEC filings for 82 public software names and sorts the disappearance into four "flavors":
- Announce you're done — Fastly (Q1 2024 10-Q) declared the metric immaterial, conveniently at 115% after it had been "material" at 147%.
- Refuse the SEC — Workday told the SEC (Aug 2024) it would no longer disclose net revenue retention, the boldest move because the correspondence is on the public record. Workday is >$8B revenue.
- Move to a premonition — CrowdStrike swapped "above 120%" (every 10-Q, 2019-Apr 2023) for vague "in line with our expectations," which conveys no actual number.
- Change the cadence — Doximity and nCino moved to annual disclosure, hiding 40- and 31-point single-year drops that quarterly reporting would have telegraphed.
The load-bearing insight: "The disclosure is the disclosure." The choice to stop sharing a metric is a separate piece of information from the metric itself, and usually tells you more. Nobody changes the cadence or fuzzes the language of a metric that is going up.
A second sharp takeaway for operators benchmarking themselves: the median is not the median. Public comp sets are self-selecting (winners keep disclosing, losers go quiet), so the NDR distribution you benchmark against has survivorship bias baked in and the real distribution is worse than the charts suggest. An honor roll (Cloudflare, Snowflake, GitLab, Klaviyo, Braze, et al.) kept disclosing consistently through the down-cycle. Salesforce, ServiceNow, Adobe, Oracle and others never reported true NDR at all. A quantitative follow-up issue is promised next week.
Mapping against Ray Data Co
Mapping strength: medium.
- Measure-through-the-downside discipline. The piece's core ethic (track a retention metric consistently, good or bad, rather than going quiet when it dips) is directly transferable to RDCO's subscription surfaces. [[Squarely]] (iOS puzzle subscription) and the [[Sanity Check]] newsletter both have retention/monetization dimensions where the temptation, once instrumentation exists, will be to surface the metric only when it flatters. The honor-roll vs. four-flavors framing is a useful self-check for how RDCO reports its own numbers to itself.
- Survivorship bias in benchmarks. When RDCO eventually benchmarks Squarely retention against published mobile-subscription cohorts, or Sanity Check churn against newsletter-industry medians, the "the median is not the median" caveat applies: public/quoted comps over-represent winners. Discount external benchmark medians accordingly.
- Founder's FP&A muscle. Ben does Progress + Nutrafol reporting at Mammoth Growth; this is squarely in his professional wheelhouse and reinforces metric-disclosure-integrity instincts he already exercises for clients. Reads more as confirmation than new input on that axis.
- Investing angle (secondary). For RDCO's investing thesis work, the "disclosure choice as signal" heuristic is a screening tip: a software name that quietly drops or re-cadences NDR is flagging deterioration. Tangential to the active chip-fab/memory capital-cycle thesis, but a reusable diligence reflex.
Honest read: genuinely useful as a metrics-integrity reminder, but RDCO's surfaces are pre-scale and B2C-subscription / info-product rather than B2B-SaaS, so the NDR mechanics themselves are more analogy than direct playbook. Hence medium, not strong.
Related
- [[Squarely]]
- [[Sanity Check]]
- [[2026-05-21-mostlymetrics-spacex-ipo-s1-breakdown]]
- [[project_investing_markov_capital_cycle]]