Net dollar retention (NDR) measures how much revenue a group of existing customers generates today compared with what the same customers generated a year earlier. New customers are left out. Upgrades, extra seats, price rises and higher usage push the figure up; downgrades and cancellations pull it down. A result above 100% means the existing base grew on its own, even if the company signed no one new. The metric exists because subscription and usage-based businesses earn most of their money from customers they already have, and investors wanted a single number showing whether that installed base is expanding or leaking.
The same idea travels under several names: net revenue retention (NRR), dollar-based net retention, net expansion rate. Sean Fanning, writing in June 2020, counted 23 different names for it. None of them is defined by US generally accepted accounting principles (GAAP) or International Financial Reporting Standards (IFRS), and none is audited.
How the calculation works
The core formula divides the revenue of a fixed cohort at the end of a period by the revenue of the same cohort at the start. A hypothetical analytics vendor illustrates the mechanics. On January 1 its existing customers produce $10 million of annual recurring revenue (ARR). Over twelve months those customers add $2 million through upgrades and price increases, cut $500,000 through downgrades, and cancel contracts worth $700,000. The cohort is worth $10.8 million at year end, an NDR of 108%. Any revenue from customers signed during the year is excluded.
Gross revenue retention (GRR), the companion figure, applies the same arithmetic but ignores expansion. In the example, GRR is ($10 million - $0.5 million - $0.7 million) / $10 million, or 88%. GRR can never exceed 100%. The gap between the two figures shows how much of a company's retention depends on selling more to customers who stay.
Beyond that shared logic, companies make four choices that change the result.
- Measurement window. Some compare ARR at two dates a year apart. Others compare revenue earned in a quarter with the same quarter a year before. Snowflake uses a two-year window: it takes customers active in the first month of the first year and divides their product revenue in year two by their product revenue in year one.
- Cohort membership. Twilio's 2016 prospectus excluded "Variable Customer Accounts", large customers without contracted commitments. Anthropic's filing excludes Claude Pro and Max consumer subscribers. Many companies also report a separate figure for customers above $100,000 of ARR, which usually runs higher.
- Averaging. Some report a trailing twelve-month average of quarterly rates rather than a point-in-time figure.
- Currency. Global companies may use constant exchange rates to strip out currency moves. HubSpot changed its calculation from the first quarter of 2026 to refresh fixed exchange rates and include commissions owed to Solutions Partners.
Origin and evolution
Expansion revenue mattered to software vendors long before the metric had a standard name, but public disclosure of a dollar-based retention ratio spread with the software as a service (SaaS) listings of the 2010s. Twilio offers a clearly dated example. Its S-1, filed on October 7, 2016, reported a "Dollar-Based Net Expansion Rate" of 155% for 2015 and 167% for the first half of 2016, comparing each quarter's revenue from active accounts with the same quarter a year earlier.
Venture investors then turned the figure into a benchmark. Bessemer Venture Partners' State of the Cloud report, published on February 6, 2019, set expectations by customer size: net retention of 80% to 100% for small-business software, 90% to 120% for mid-market and above 100% for enterprise. Dave Kellogg, presenting at Gainsight's Pulse conference in June 2021, put median NDR at 104% across private SaaS companies of all sizes and 111% for public companies, and called it "the most powerful predictor" among common metrics of a company's revenue multiple.
The 2020 software boom produced the extreme readings. Snowflake's S-1, filed on August 24, 2020, reported net revenue retention of 158% as of July 31, 2020, having reached 223% a year earlier.
Regulators followed. On January 30, 2020, the Securities and Exchange Commission (SEC) issued guidance in Release No. 33-10751 requiring companies that present key performance indicators to define them and disclose changes in method. The staff applies it to NDR. In a letter dated February 23, 2024, SEC staff asked Model N whether customers lost since the prior-year period were counted in its SaaS net dollar retention. The company confirmed they were, reported that the measure had fallen to 115% for 2023 from 134% for 2022, and agreed to revise its disclosure.
Why it matters for the marketing industry
Most marketing and advertising technology vendors report NDR or a close variant, and the figure often explains share price moves that revenue alone does not. Monday.com reported net dollar retention of 111% overall and 117% for customers above $100,000 of ARR in the second quarter of 2025. Figma's IPO filing showed a 132% net dollar retention rate. Pattern, the Amazon marketplace agency, disclosed NRR of 116% for 2024 and 118% for the first half of 2025.
A reading below 100% signals that the existing base is shrinking. Similarweb reported overall NRR of 98% in the first quarter of 2026, against 101% a year earlier, with the $100,000-plus cohort falling to 103% from 111%. Sell-side platform PubMatic, which applies the metric to publisher and buyer revenue rather than subscriptions, reported net dollar-based retention of 98% for the twelve months to September 30, 2025, against 112% a year earlier, and 98% again for the twelve months to June 30, 2026, against 102%.
Customer engagement vendors sit just above the line. Klaviyo's net revenue retention stood at 109% at its August 5, 2026 results, with management citing the prior year's profile enforcement. Braze reported dollar-based net retention of 110% for its first quarter of fiscal 2027. LiveRamp's subscription net retention fell to 103% from 104% in the quarter to June 30, 2026, which the company attributed to lower variable revenue.
Limitations and disputes
The first objection is comparability. A quarterly-revenue NDR, a trailing two-year product-revenue NDR and a point-in-time ARR NDR are three different measurements sharing one label. Exclusions change results further. Similarweb excludes LLM data evaluation contracts from net retention until they convert to recurring contracts, and Twilio removed its largest uncommitted accounts.
The second is that NDR says nothing about cost. Fanning, writing in March 2021, argued that "net dollar retention tells us nothing about a company's margin on expansion" and called it "non-GAAP and highly gameable". Upsells that require heavy discounting, extra customer success staff or new engineering can lift NDR while eroding profit.
Third, small bases inflate the ratio. When a company's revenue multiplies within a year, a cohort that was tiny twelve months earlier produces a spectacular percentage. Harrison Rolfes, a senior analyst at PitchBook, told Morningstar that Anthropic's figure would come down sharply as the comparison base caught up with its scale, according to Morningstar's October 28, 2026 report.
Fourth, usage-based pricing makes the metric volatile. Customers on consumption models can cut spending without cancelling, so NDR can swing with a single large account's workload or budget cycle.
Finally, benchmarks are drifting down. ChurnZero, citing a survey of more than 1,000 private SaaS companies in June 2026, put top-quartile NRR at around 110%, down from 120% to 125% in earlier years, with the median just above 100%. SaaS Capital reported a median NRR of 103% and GRR of 91% in 2026 for bootstrapped companies with $3 million to $20 million of ARR.
What it is not
Gross revenue retention. GRR caps at 100% and counts only losses. NDR adds expansion. NIQ reported net dollar retention of 105% and gross dollar retention of 99% for the second quarter of 2026, showing both side by side.
Customer or logo retention. This counts accounts kept rather than dollars. HubSpot's "customer dollar retention" in the high 80s is a gross measure distinct from its net figure.
Churn. Churn measures what leaves. Net revenue churn below zero, marketed as negative churn, is the same condition as NDR above 100%.
Annual recurring revenue. ARR is a revenue total that includes new customers. NDR is a ratio built from ARR or revenue movements within an existing cohort. The app-marketing sense of retention, such as day-30 user retention, counts people who return to an app rather than dollars.
Recent developments
The largest NDR figure yet disclosed came from an AI developer. Anthropic's S-1, filed on October 26, 2026, reported net revenue retention of about 900% as of September 30, 2026. The filing defines it as qualified customers' revenue in the three months to the measurement date divided by their revenue in the same three months a year earlier, excluding Claude Pro and Max subscribers. The figure was repeated in the amended filing with third-quarter results.
Software companies reporting in November 2026 showed a split. Cloudflare's dollar-based net retention rose to 122% from 120% in the third quarter. HubSpot reported net revenue retention of 101% and guided for a decline of about two points for 2026. NIQ said on November 12 that net dollar retention held at 105%. Snowflake, whose 2020 figure set the high-water mark for listed software, reported 126% for the quarter to July 31, 2026, according to its September 2, 2026 results.
Timeline
- October 7, 2016: Twilio files its S-1, reporting a dollar-based net expansion rate of 155% for 2015
- February 6, 2019: Bessemer's State of the Cloud sets net retention benchmarks by customer segment
- January 30, 2020: SEC issues KPI disclosure guidance in Release No. 33-10751
- June 2020: Sean Fanning counts 23 names for net dollar retention
- August 24, 2020: Snowflake files its S-1 with net revenue retention of 158%
- March 19, 2021: Fanning's "Down with NDR" criticises the metric as non-GAAP and gameable
- June 10, 2021: Dave Kellogg puts median NDR at 104% for private SaaS and 111% for public companies
- February 23, 2024: SEC staff question Model N on whether lost customers are counted in its NDR
- July 2025: Figma's IPO filing shows 132% net dollar retention
- August 22, 2025: Pattern files its S-1 with NRR of 116% for 2024
- Q1 2026: HubSpot revises its NRR calculation
- May 13, 2026: Similarweb reports overall NRR of 98% for Q1 2026
- September 2, 2026: Snowflake reports 126% net revenue retention
- October 26, 2026: Anthropic's S-1 reports net revenue retention of about 900%
- November 4, 2026: HubSpot reports NRR of 101%
- November 2026: Cloudflare reports dollar-based net retention of 122%; NIQ reports 105%
Related PPC Land coverage
- Explaining annual recurring revenue - The revenue total from which most NDR calculations are derived.
- Explaining churn - How revenue losses are counted, and why negative churn equals NDR above 100%.
- Anthropic files for Nasdaq IPO with $16.3bn in first-half revenue - The S-1 disclosure of about 900% net revenue retention and its quarterly definition.
- Anthropic's IPO filing shows $16.3bn of revenue and no plan to sell ads - Notes the exclusion of Claude Pro and Max subscribers from the metric.
- Anthropic swings to $567m Q3 operating loss on $5bn of short-term compute - The amended S-1 repeating the retention figure.
- Cloudflare revenue gains 39% to $782.5m, fastest growth in nearly four years - Dollar-based net retention rising to 122% in Q3 2026.
- HubSpot forecasts Q4 revenue below Q3 as growth slows to 12% from 17% - NRR of 101% and a revised calculation method.
- NIQ swings to $35.8m profit, its first since the July 2025 IPO - Net dollar retention of 105% alongside gross retention of 99%.
- NIQ AI-native revenue gains 34% as agentic commerce product nears launch - Q2 2026 net and gross dollar retention.
- Similarweb hits $73.9M in Q1 as CEO succession and Manus deal reshape its AI bet - NRR falling to 98% overall and 103% for large customers.
- Similarweb posts eighth straight positive cash flow quarter in Q3 - Exclusion of LLM data contracts from net retention.
- PubMatic beats Q3 2025 expectations with CTV growth, AI strategy - Net dollar-based retention dropping to 98% from 112%.
- PubMatic revenue gains 11% after three quarters of decline - Retention of 98% for the twelve months to June 30, 2026.
- Monday.com reports Google AI search changes impacting traffic acquisition - Net dollar retention of 111% overall and 117% for large customers.
- Figma files for $1.5 billion IPO on NYSE after Adobe acquisition collapse - A 132% net dollar retention rate in the IPO filing.
- Pattern Group files S-1 for IPO with 94% Amazon revenue dependency - NRR of 116% and 118% for an Amazon marketplace agency.
- Explaining Klaviyo - Net revenue retention of 109% after profile enforcement.
- Explaining Braze - Dollar-based net retention of 110% in fiscal Q1 2027.
- LiveRamp faces August 17 vote on Publicis buyout at $38.50 a share - Subscription net retention slipping to 103%.
Summary
Who: Subscription and usage-based companies report it, including Twilio, Snowflake, Cloudflare, HubSpot, Klaviyo, Braze, Similarweb, PubMatic, NIQ and Anthropic. Investors such as Bessemer use it as a benchmark, and the SEC reviews how companies define it.
What: A non-GAAP ratio comparing a fixed customer cohort's revenue today with the same cohort's revenue a year earlier, counting expansion, contraction and churn but excluding new customers. Figures above 100% mean the existing base grew.
When: Twilio reported a dollar-based net expansion rate in its October 2016 S-1. Bessemer published benchmarks in 2019, the SEC issued KPI guidance in January 2020, Snowflake reported 158% in August 2020, and Anthropic disclosed about 900% in October 2026.
Where: Earnings releases, IPO prospectuses, investor presentations and venture benchmarks, mainly in SaaS, martech, ad tech and AI.
Why: Most revenue at subscription businesses comes from existing customers, so NDR shows whether the installed base compounds or erodes. Because it is unaudited and each company sets its own cohort, window and exclusions, figures are hard to compare and open to flattering construction.
Discussion