Data Integrity

Every claim in this analysis traces to a single primary source — the KBCM/Sapphire Private SaaS Survey — with an explicit per-data-point verification tier and a public corrections log. This report went through a documented revision (2026-06-13) that fixed six distinct data errors from an earlier draft; those corrections are disclosed below rather than silently absorbed.

Methodology

Every data point in this analysis is tagged with one of four verification tiers: verified (sourced directly from a published survey edition, cross-checked against the original release), cross-survey (consistent across two or more survey editions, range documented), calculated (derived from verified inputs using a documented formula — e.g., the Net Magic Number, the NDR/GDR expansion multiplier), or composite (aggregated across multiple survey years, each component independently verified). A small number of directional claims — noted inline — carry no tier above inference and are labeled [Speculative — uncertified] rather than presented as fact.

Data Source

This analysis draws from a single primary source: the KBCM/Sapphire Private SaaS Survey, annual editions 2021 through 2025 (the 12th through 16th annual editions). Using one continuous, methodologically consistent survey — rather than blending several distinct surveys with different sample compositions and definitions — is a deliberate choice: it avoids the cross-survey definitional drift (what counts as "expansion," how churn is measured, which company sizes are sampled) that makes multi-source SaaS benchmarking notoriously unreliable.

The full press-release text and source URLs for each survey edition are catalogued on the Resources page.

Corrections Log

This report underwent a data-integrity review on 2026-06-13 that identified and corrected six errors present in an earlier draft. Each is disclosed here rather than corrected silently:

• B1 — Expansion share splice: An earlier draft spliced a median-regime series (2018–2021, peaked 46%) and a pooled-regime series (2022–2024) onto one continuous line, and separately transposed the expansion/new-logo columns for the pooled years — producing a false "59% → 48% decline" narrative. Corrected: the two series are distinct statistics, shown separately; the pooled series actually rose 42% → 52%. See The Expansion Myth.
• B2 — Downsell segment rebuild: The downsell-by-segment cut mixed ARR-size bands and growth-rate bands inconsistently. Corrected: rebuilt as a clean growth-rate cut (downsell is 59% of loss for slow-growers, 32% for fast-growers). See Downsell.
• B3 — OpEx/EBITDA year-label fix: Figures reported as 2024-actual (75% total OpEx, +1% EBITDA) were actually 2026E projections. Corrected: 2024-actual is 88% OpEx / -12% EBITDA; 75% / +1% are labeled as projections. See Operating Expenses.
• B4 — S&M / R&D transposition: The two cost-line series were swapped in an earlier OpEx breakdown. Corrected in the current opexData series.
• B5 — NDR 2022 reconciliation: Different survey editions reported 2022 NDR between 102% and 109% depending on which edition's retrospective figure was used. Corrected: the 16th-edition (2025 survey) retrospective figure of 106% is used as authoritative, with the range disclosed in the Metric Explorer's cross-survey toggle.
• B6 — Churn-by-contract-length correction: An earlier draft reported 17/12/8/5% churn across contract-length bands from an unmatched table. Corrected and verified against the 2024 survey (p.12): 14/10/6/3%. See Churn.

Known Limitations

Survivorship bias: The survey, by definition, measures companies that survived the measurement period. Failed companies are underrepresented, which likely makes retention metrics appear more stable than the true population-level picture.

Sample composition: The survey skews toward funded, private SaaS companies. Bootstrapped and publicly traded companies are underrepresented, and the findings in this report are strongest for the private, VC-backed segment that makes up the bulk of the sample.

Self-reported data: The survey relies on respondent accuracy. Retention metrics (NDR, GDR, churn) are particularly vulnerable to this, since companies sometimes define them differently or estimate rather than calculate precisely. Single-source design means this analysis cannot cross-validate against an independent survey the way a multi-source study could — the tradeoff made deliberately in exchange for definitional consistency (see Methodology).

Uncertified directional signals: A small number of claims in this report — most notably the AI-driven seat-compression trend referenced under The Expansion Myth and AI & Pricing — are included as directional signal but have not cleared the verified/cross-survey/ calculated tier. These are explicitly labeled [Speculative — uncertified] wherever they appear and should not be treated with the same confidence as the tiered claims above.

Macro confounds: 2022–2025 saw significant macro shifts — interest rate increases, AI acceleration, cloud infrastructure cost inflation. This analysis isolates SaaS-specific metrics where possible, but some findings may reflect macro factors as much as SaaS-specific structural change.

The corrections disclosed above, and the single-source design that made them traceable, are the basis for the confidence levels applied throughout this report. Where a claim's tier is anything other than verified or cross-survey, that is stated explicitly at the point of use.