The Expansion Myth

The SaaS land-and-expand promise, that most new revenue comes from expanding existing customers, is real at the top of the distribution but mis-stated as the universal default. Across every year KeyBanc published a company median for expansion’s share of new ARR (2018–2021), the typical SaaS company sourced more of its new ARR from new logos than from expansion: the median never exceeded 46%. Expansion’s share crossed 50% only in the 2024 data year, and only on a pooled, dollar-weighted basis (52%) driven by companies above $25M ARR. Below that threshold, new-logo acquisition still dominates. Expansion-majority growth is a large-company, top-of-distribution achievement, not the trajectory the median company actually walks, and the promise’s most-repeated headline (“~80% of future revenue comes from existing customers”) is an industry folk-belief no one can trace to a primary source.

The Median Reality

The textbook says expansion should overtake new-logo acquisition as a company scales: crossing 50% of new ARR in the growth stage and trending toward 60–70% at maturity, on the back of 120%+ net dollar retention. The benchmark says otherwise, and immediately. Across every year KeyBanc published a company median for expansion’s share of new ARR, 2018 through 2021, the figure never exceeded 46%. By the definition of a median, the typical SaaS company sourced more of its new ARR from new logos than from expansion in every measurable year. The land-and-expand majority is not where the middle of the market lives.

Source: KBCM SaaS Survey 2019 p15 / 2020 p12 / 2021 p16 / 2022 p15: median expansion share (2018–2021 data years)Verified

Two Regimes, Not One Trend

Reading the full series honestly requires care: KeyBanc changed the statistic mid-stream. For data years 2018–2021 it published a company median; from 2022 it switched to a pooled, dollar-weighted “Overall” column. These are different measurements, the middle company versus the dollar-weighted aggregate, and no data year appears under both, so they cannot be stitched into one line. Within each regime the direction is the same: up, not down. The one adjacency that looks like a fall, 46% (2021) to 42% (2022), is a statistic change, not a decline.

Source: KBCM SaaS Survey 2023 p10 / 2024 p9 / 2025 p18: pooled “Overall” expansion share (2022–2024 data years). Median (2018–2021) and pooled (2022–2024) are different statistics, not one trend.Verified

A Privilege of Scale

The crossing in 2024 is a scale story, not a “the average company now expands more than it hunts” story. The pooled 52% is pulled up by the largest firms. Split by ARR segment (2024 data), expansion’s share rises monotonically with size. Only companies above $25M ARR are expansion-majority; below that, new-logo acquisition still supplies most of new ARR. The smaller the company, the more it depends on hunting new logos, the exact inverse of the claim the earlier version of this report made, which had the largest companies least expansion-driven.

Source: KBCM SaaS Survey 2025 p18: distribution of new ARR by ARR segment (2024 data; 59 respondents; visually re-confirmed)Verified

There is also no certifiable count of how many companies are expansion-majority. KeyBanc publishes no “% of companies whose expansion share exceeds 50%” statistic, so this report makes no such claim: only the median (≤46%, i.e. fewer than half by definition) and the by-size cuts are defensible.

For an operator the implication is directional. A company below $25M in ARR that treats land-and-expand as its primary growth engine is planning against the pattern the typical company in its band actually exhibits. Expansion is earned by scale, not assumed at it. The reliable path below $25M runs through new-logo efficiency and the retention that makes later expansion possible; expansion-led growth is better treated as a consequence of reaching scale than as the route to it.

What the Evidence Supports

The economic logic behind land-and-expand is sound, and where expansion is real it is genuinely the higher-quality dollar. What the benchmark data supports, though, is narrower than the doctrine claims. Median net dollar retention for private B2B SaaS sits around 102% (top quartile 111%, bottom quartile 97%), barely above the line that separates net expansion from net contraction, and the retention-economics origin holds, with a five-point reduction in customer defection worth 25–85% in profit. The familiar headlines that outrun this evidence, that roughly 80% of future revenue comes from existing customers, or that expansion reaches 60–70% at maturity, trace to no primary source and are used nowhere in this report as fact. The defensible conclusion is the one the data draws on its own: expansion as a primary growth engine is a privilege of scale. It arrives late, appears only in the dollar-weighted aggregate, and never at the median.

SaaS Capital 2025 (private B2B SaaS, ACV $25–50K; primary, methodology stated); Reichheld & Sasser, Harvard Business Review, 1990 (origin figure 25–85%). The ~80% and 60–70% figures are named as untraceable industry folk-beliefs.Verified

Why Expansion Stays Hard

If expansion is the higher-quality dollar, why does the median company still get most of its new ARR from new logos? Three forces, offered as editorial interpretation rather than survey findings.

1. AI commoditization. AI-native alternatives erode the differentiation upsells depend on, narrowing the expansion path.
2. Budget rationalization, the best-attested of the three. Post-2022 vendor consolidation and tighter software budgets shrink the headroom for existing customers to spend more.
3. A pricing-model shift toward usage-, consumption-, and outcome-based pricing and away from flat per-seat, which in principle should aid expansion.

That the pricing shift hasn’t lifted the median is itself the tell: the tailwind exists, yet expansion-majority remains a large-company phenomenon, suggesting the binding constraint is demand-side, not pricing mechanics.

Editorial interpretation; thematic support KBCM SaaS Survey 2025 p18, p27 (Sapphire on agentic AI). Not survey findings.Inference

The Engine Nobody Forecasts

There is a harder version of the myth, and it is not about how expansion is counted but about whether anyone plans it. New-logo acquisition arrives fully instrumented: it has a pipeline, a quota, a productivity benchmark the survey reports every year, and a forecast. Expansion, now the majority of new ARR above $25M, gets none of the four. It is measured after the year closes and managed, when it is managed at all, as a hope.

The survey makes the omission visible. Its forward estimates are respondent self-forecasts: companies report their own next-year ARR, growth, and retention, and those answers become the 2025E and 2026E columns. No one is asked to forecast the expansion-versus-new-logo split of next year’s new ARR, which is why the composition rows on the certified data page run 42, 45, 52 through 2024 and then stop. The blank forecast cell is not a data limitation; it is the industry’s reactive posture toward expansion, printed as an empty column.

The objection writes itself: you cannot forecast the split. It does not survive contact with how the same companies forecast everything else. A forecast is prior achievement plus a target plus pipeline judgment, which is exactly how every company builds its new-logo number, and no one calls that guesswork. Calling a new-logo forecast a plan and an expansion forecast a guess describes the operator, not the metric. The operators who can say what expansion revenue will be next year, and where it will come from, are running expansion as a pipeline: named accounts, staged opportunities, coverage, a committed number. The ones who cannot are treating the majority engine as weather.

This is the forward-looking half of the measurement asymmetry the report documents in the rearview. Customer Success carries no expansion-per-head benchmark (see Customer Success Coverage), and the entire net-retention deterioration sits in the expansion layer (see The NDR Crisis). Add the planning gap and the picture is whole: the industry pipelines and forecasts the minority engine and takes whatever the majority engine delivers. The most expensive form of the myth is not that expansion-majority is oversold; it is that the one engine now carrying most new ARR at scale is the one no one runs a pipeline for.

Editorial interpretation. Grounded in the report’s own cuts: the survey benchmarks AE productivity annually but publishes no expansion-per-CSM series (KBCM SaaS Survey 2025 p30, p31), and the expansion-composition rows carry no forecast (2025 Survey p18). Not a survey finding.Inference