NDR Crisis

Net dollar retention is the clearest single measure of whether a SaaS business compounds or merely runs in place, and for the median private SaaS company it has fallen for three consecutive years, from 109% in 2021 to 101% in 2024. An eight-point decline sounds modest. It is not. At 109%, a company grows materially from its existing base before adding a single new customer. At 101%, that base barely holds, and growth must come almost entirely from new-logo acquisition, the most expensive revenue a SaaS company can buy.

One thing to square before the trajectory. The 2021 peak and the 2024 trough are not the same measurement. KeyBanc printed 109% in its 2022 edition on a median restricted to companies above $5MM ARR; the 101% is the 2025 edition’s unqualified median. Some unknown share of those eight points is that population change rather than the market. This report makes the comparison anyway, for a reason worth stating plainly: KBCM-2025 prints no 2021 column at all, so there is no single-edition way to reach back to the peak. The alternative to an imperfect comparison here is no comparison. The direction is not in question, because every edition since 2022 shows net retention falling on its own basis, but the magnitude carries a wider band than eight points suggests. Every chart on this page plots one edition only; where the 2021 anchor appears, it is labelled and kept off the line.

The Trajectory

The decline moved in three phases: a peak near 109% through 2021, a rapid inflection across 2022 and 2023 as budget rationalization and AI commoditization thinned the net-expansion uplift, and stabilization at 101% by 2024. The forecast recovery to 102–104% assumes a return to cross-functional alignment the current data does not yet show. The cohort also spread out. The 2024 range runs from 62% to 118%, which means the median increasingly describes fewer and fewer of the actual companies inside it.

Source: KBCM/Sapphire Private SaaS Survey 2022–2025 (longitudinal)Verified

The Pre-ZIRP Baseline

One caution on the starting line. The 109% of 2021 was not a plateau the median fell from; it was a peak. Before the cheap-capital surge, median NDR sat in the low 100s for three straight years: 102.7% in 2018, 102% in 2019, and 101.8% in 2020. Read against that baseline, 2024’s 101% is not a fall below a healthy norm. It is reversion toward the ~102% the metric held before 2021, with the 109% itself the anomaly, a ZIRP-era high that cheap capital and aggressive net expansion made briefly look normal.

Two caveats keep this honest. The 2019 and 2020 anchors come from the 2020 and 2021 editions, which straddle the COVID disruption, so read the three-year band as a settled pre-surge level rather than three precise points. And the anchors are shown here as reference points, not spliced into the trajectory above, because the report pins each figure to a single named edition and does not stitch vintages into one series.

This does not soften the finding; it relocates it. The problem was never the NDR level on its own. It is that the reversion happened with the cushion gone: gross churn is still frozen near 14%, and net expansion now covers churn only 1.07× over, against 1.6× at the peak and roughly 1.14–1.21× in the pre-ZIRP years (the coverage section below derives those anchors). NDR is back near its old baseline, but the machinery that used to hold it there, and could push it higher, no longer has slack.

Pre-ZIRP anchors: 2018 median 102.7% (KBCM-2019 p53, >$5M ARR); 2019 median ~102% (KBCM-2020 p11, $5M+ ARR cut); 2020 median 101.8% (KBCM-2021 p25, >$5M ARR). Each pinned to its own edition, not spliced.Cross-Survey

The 2022 Turn

If the pre-ZIRP baseline explains where 101% sits, 2022 explains when the descent began. The survey’s own edition titles bracket the shift. KBCM titled its 2021 report Acceleration of Growth; a year later the 2022 edition was Resilience Amidst Market Turmoil. Between those two headlines the cheap-capital era ended. The Federal Reserve began raising rates in early 2022 and, as Bessemer put it in its State of the Cloud 2023, the market moved out of a near-zero-interest environment at record-breaking speed. The frenzy that had funded aggressive net expansion drained with it.

That macro turn maps onto the trajectory. Bessemer describes 2022 as a paradigm shift from the age of growth-at-all-costs to the age of efficiency, one in which tighter customer budgets and longer sales cycles pressed directly on the growth fundamentals that feed net retention. For a metric built on existing customers spending more, tighter budgets are the mechanism: the expansion uplift that had carried NDR to 109% was the first component to compress when spending discipline returned. NDR did not collapse. It reverted, and it reverted for a reason the whole sector felt at once.

One detail keeps this honest, and it strengthens the report’s central claim rather than softening it. Even as KBCM named the turmoil in its 2022 title, the same release told readers that private SaaS companies had “not been materially affected by the current market landscape.” The turn was already underway, and the contemporaneous read minimized it. That is the reckoning in miniature: the shift is invisible inside any single edition and visible only when the editions are held to their own years and lined up. Dating the turn to 2022 is interpretation, corroborated by outside observers; the retention figures themselves are certified and unchanged.

KBCM edition titles: Acceleration of Growth (12th annual, 2021); Resilience Amidst Market Turmoil (13th annual, 2022), which reported SaaS “not materially affected.” Industry framing of the 2022 rate shock and the growth-to-efficiency pivot: Bessemer Venture Partners, State of the Cloud 2023. Dating is interpretation; retention figures certified.Inference

The Narrowing Gap

Gross dollar retention held near 86% throughout, but the gap between GDR and NDR narrowed from 20 points in 2022 to 15 points in 2024. The 2021 ZIRP peak was wider still at 23 points. That reading comes from KBCM-2022 on a “Median (≥$5MM)” basis, a different population from the series plotted here, so it is not drawn on the same line. Because NDR equals GDR plus the net-expansion uplift, a shrinking gap on flat GDR is not churn improving. It is net retention’s lead over gross thinning. The business is not losing fewer customers; it is expanding the ones it keeps by less.

Plotted series: GDR and NDR from KBCM-2025 p4, 2022–2026E; gap derived (NDR − GDR). The 2021 anchor quoted above is not plotted and is not from this edition: NDR 109% is KBCM-2022 p24, Median (≥$5MM); its GDR of 86% is derived, because KBCM-2022 prints no gross dollar retention figure at all, only gross dollar churn of 14% (p22), giving 100 − 14. The 23-point gap follows from those two.Calculated

Expansion-to-Churn Coverage

One ratio captures how much cushion net expansion gave the business, and it collapsed. Take the points already on this page. In 2021 gross churn ran 14 points (GDR 86%) and net expansion added 23 points on top (NDR 109%), so expansion covered churn 23 ÷ 14, or 1.6× over. By 2024 gross churn was still 14 points but expansion added only 15 (NDR 101%), so coverage fell to 15 ÷ 14, just 1.07×. Expansion once outran churn comfortably; now it barely clears it, which is precisely why NDR sits a hair above 100%.

This is expansion-to-churn coverage, not expansion’s share of new ARR, which rose over the same period (see The Expansion Myth). The two are different measurements and move in opposite directions. The middle years are interpolated between the endpoints and shown as estimates.

Derived: (NDR − GDR) / (100 − GDR), net-expansion points per point of gross churn; endpoints from published GDR/NDR (2021, 2024), 2022/2023 interpolatedCalculated

The same reversion discipline the report applies to the NDR headline has to be applied to this ratio, or the ratio inherits the same distortion it was built to correct. The 1.6× endpoint is the ZIRP peak, and measuring the fall from a peak overstates it. The pre-ZIRP editions carry both inputs on a matching median basis, so coverage can be derived for each year inside its own edition: in 2018, 12.7% gross churn against 102.7% NDR gives 15.4 expansion points over 12.7 of churn, or 1.21×; in 2019, 14.5 over 12.5, or 1.16×; in 2020, 14.4 over 12.6, or 1.14×. Before the surge, expansion covered churn roughly 1.14–1.21 times over. The honest statement of the decline is therefore not that a 1.6× cushion collapsed; 1.6× existed for one anomalous year. It is that 2024’s 1.07× sits below even the pre-surge band, in a metric where the distance to 1.0× is the entire margin for error.

The decomposition sharpens the finding rather than softening it. Compare 2024 with the pre-ZIRP years component by component: the expansion side is intact, 15 net-expansion points now against 14.4–15.4 then. What deteriorated is the denominator. Gross churn runs 14 points now against 12.5–12.7 then. The thinning cushion, measured against history rather than against the bubble, is not an expansion collapse; it is a churn floor that settled more than a point higher than its pre-surge level and has not moved since. That reading converges with the report’s structural claim from the churn side: the binding constraint is the frozen floor, and the coverage ratio is thinner today mostly because the floor is higher, not because expansion forgot how to grow. The usual caveats travel with the anchors: pre-ZIRP figures are a $5M-plus median basis from redrawn panels, the 2019 and 2020 anchors straddle COVID, and each ratio is computed entirely within one edition, never across two. Calculated

Derived within single editions: 2018 churn 12.7% (KBCM-2019 p51) with NDR 102.7% (KBCM-2019 p53) → 1.21×; 2019 churn 12.5% with NDR 102% (KBCM-2020 p11) → 1.16×; 2020 churn 12.6% (KBCM-2021 p23) with NDR 101.8% (KBCM-2021 p25) → 1.14×. All median >$5M ARR basis, verified at source; formula (NDR − (100 − churn)) / churn.Calculated

What 101% Hides

The headline figure conceals as much as it reveals. Two companies can both report 101% NDR and face opposite risk. One reaches it with 5% gross churn and 6% expansion: a stable base, lightly grown. The other reaches it with 14% gross churn and 15% expansion: a leaky base, aggressively patched. The reported number is identical; the durability is not.

The certified median sits at the second, riskier composition. Roughly 14 points of gross churn (GDR 86%) offset by 15 points of expansion (NDR 101%). That is why the coverage ratio matters more than the headline: the median company is holding 101% by running expansion hard against a high churn rate, not by keeping its customers well. Remove a point or two of expansion, which is exactly what a soft quarter does, and the number falls below 100%. Calculated

The comparison that makes this legible: a blood-pressure reading that spikes and then returns to its old number looks recovered on the chart. But if the years in between thinned the arteries, the reading is identical and the margin before something gives is not. NDR at 101% today reads the same as NDR at ~102% in 2019. What changed in the interim is the cushion, not the number on the chart.

Cross-Survey Measurement Variation

For 2022, reported NDR ranges from 102% to 108% depending on which survey edition measured it, a six-point spread that reflects different respondent pools rather than a real movement. The directional trend is reliable even where absolute values differ, which is why the report pins each figure to a single named edition and never splices across them. Cross-Survey

Why It Matters

The cost is quantifiable. At the median 14% gross churn, a $30M company destroys roughly $4.2M of ARR each year before it books a dollar of new business. Moving to top-quartile retention, near 7% churn, would preserve about half of that, some $2.1M annually, on the same revenue base. That preservation is not an operational nicety. It is a capital-allocation return, and it compounds.

Illustrative: $30M ARR × 14% median gross churn = $4.2M; at 7% (top quartile) = $2.1M; scale hypothetical, churn rates certifiedCalculated

Below 100% NDR, several effects arrive together. CAC payback extends as the net-expansion uplift that used to subsidize acquisition thins. Holding revenue flat requires several times the acquisition volume it once did. The revenue mix shifts from expansion-led to acquisition-led, and sales and marketing spend grows steadily less efficient. Retention, in other words, is not a downstream operational outcome. It is the input that sets the efficiency of everything upstream of it, and it is largely determined by decisions, contract structure, onboarding, and the functions a company chooses to fund, made twelve to twenty-four months earlier. Inference

Where the line goes from here is the survey’s estimate, not a measurement: the 2025E and 2026E columns in the trajectory above, drawn dashed, project 102% and then 104%. The last time the survey’s estimates could be checked against actuals, the actual came in below the estimate in 10 of 12 metrics, so the report reads those columns as a ceiling; the KPI Scorecard carries the full calibration. Inference

One asymmetry in that forecast is worth naming. The survey projects the NDR outcome forward, yet never the expansion that produces it: companies are asked to forecast retention, but never the expansion-versus-new-logo split of next year’s new ARR. The layer where the entire net-retention deterioration lives is the one layer no one is asked to plan. That forward blind spot is developed in The Expansion Myth. Inference