Retention Benchmarks: Provenance and How the Skill Uses Them

Every number in the skill carries a source and a vintage; retention benchmarks drift with the funding climate, so vintage matters. Recalibrate against current-year surveys where you have access.

Gross revenue retention (GRR)

• Median ~90% for private B2B SaaS; top quartile above 95% (SaaS Capital annual survey and independent 2025 aggregations, e.g. Statisfy 2025). GRR strips expansion out and answers one question: how much revenue did you keep? The skill scores the save motion against this line because expansion cannot be allowed to camouflage leakage.

Net revenue retention (NRR)

• Reported 2025 medians cluster around 100-106% blended, with a strong ACV gradient: enterprise (>$100K ACV) ~118%, mid-market ~108%, SMB ~97% (2025 industry aggregations of survey data). The skill references NRR only for context; the save motion is a GRR instrument. Segment any target against your ACV band, not the blended headline.

Retention vs acquisition economics

• The claim that retaining a customer costs a fraction of acquiring one traces to Bain & Company / Harvard Business Review work (2014 lineage) and is still reported across 2025-2026 industry guides at 5x to 25x depending on segment and method. The skill uses the conservative end (5x+) and only to justify staffing a save motion at all, never to price a specific concession.

Practice-based rules (no external study)

Flagged in the skill as practice-based, not benchmarked:

• Risk is visible in system behavior (usage decay, sponsor silence) well before it is spoken; discovering risk inside T-60 indicates a detection failure.
• Scope right-sizing retains more two-year revenue than first-lever discounting.

Both come from operator experience across B2B scale-up engagements. Treat them as defaults to test against your own cohort data, and replace them with your measured numbers as soon as two quarters of save-motion scorekeeping exist.