Existing-Customer Expansion
Do current customers spend more over time by buying more seats, using more, or adding products?
More customers, less growth
Twilio sells the plumbing behind the texts and calls that apps send, such as the message that says your ride is arriving or the code you type in to log in to your bank. Customers pay by usage: the more messages they send, the more they pay. There is no big yearly contract to cancel. A customer that wants to spend less simply sends less.
For years that model looked like a perpetual-motion machine. Twilio's "dollar-based net expansion rate", its measure of how much more its existing customers spent than a year earlier, was 137% for 2020.[^1] In 2021 revenue rose 61% to $2.84 billion, and the rate was 131%. As a group, the customers Twilio already had were spending about 31% more than they had a year before.[^2] Many of those customers were young, fast-growing apps. In its annual report Twilio listed among its customers businesses in social media, cryptocurrencies, online retail and direct-to-consumer brands.[^3]
Then the engine stalled. The rate fell to 121% for 2022 and to 103% for 2023. Revenue growth slowed to 35% and then to 9%.[^4] In February 2023 Twilio cut about 17% of its workforce.[^5] By August 2023 the stock was more than 85% below its all-time high.[^6] In January 2024 Jeff Lawson, Twilio's co-founder, resigned as chief executive.[^7][^3]
Yet Twilio never stopped winning customers. Active customer accounts rose from 256,000 at the end of 2021 to more than 305,000 at the end of 2023.[^2][^4] Nor did a single giant client walk away: Twilio's ten largest accounts made up only 10% to 12% of revenue in each of those years.[^3] What disappeared was the extra spending from the many customers it already had. The same industries that had boomed were now cutting costs, and because Twilio charges by usage, they could save money by sending fewer messages without ever cancelling.[^3][^6] Investors who treated 131% as permanent were betting that existing customers would keep spending almost a third more every year, forever.
The story has a second act. Expansion crept back to 104% in 2024 and 108% in 2025, and revenue growth recovered to 14% in 2025, on more than 402,000 active accounts.[^8] The customers had not gone anywhere. What changed was the answer to one question: how fast will the customers we already have keep growing? That question is the subject of this chapter.
What it is
Existing-customer expansion is revenue growth that comes from customers a company already has, rather than from new ones. It can happen in several ways:
- More seats. A company that pays per user adds more users.
- More usage. A customer sends more messages, stores more data, or runs more transactions.
- Upsell. A customer upgrades to a pricier plan or tier.
- Cross-sell. A customer buys a second or third product from the same company.
- Price increases. The same customer pays more for the same thing.
The opposites are contraction (or downsell), when a customer stays but spends less, and churn, when a customer leaves. Investors care about the net result.
A related idea is share of wallet: of everything a customer spends on a type of product, how much goes to you. A company can expand by growing its share of wallet even when the customer's total budget is flat, for example when a store moves more of its card payments to one processor.
Think of a pizza shop that caters lunches for local offices. One office starts ordering twice a week. Another adds salads to every order. A third cuts back. Whether catering is expanding depends on how those pluses and minuses add up.
How to measure it
The main measure is net revenue retention (NRR), also called net dollar retention. Twilio calls its version the dollar-based net expansion rate. Take the customers you had a year ago, and compare what they spend now with what they spent then.
Net revenue retention = Revenue this year from last year's customers ÷ Revenue from those same customers last year
The key detail: new customers are left out. Only the starting group counts, and those who left count as zero.
Suppose the pizza shop's ten office clients spent $100,000 last year. This year, two offices ordered more (+$30,000), one cut back (−$5,000), and one left entirely (−$10,000). Those same ten clients now spend $115,000:
NRR = $115,000 ÷ $100,000 = 115%
Above 100%, existing customers as a group spend more; below 100%, the group is shrinking, even if a few grew. At 115%, catering revenue would grow 15% even with no new clients.
Now compare a stricter measure. Gross revenue retention (GRR) counts only the losses and ignores expansion: ($100,000 − $5,000 − $10,000) ÷ $100,000 = 85%. GRR can never exceed 100%. It answers a different question, "How much do we keep?", which is the subject of Customer Retention. The gap between the two, here 30 points, is the expansion itself.
NRR also splits total growth. If the shop also signs new offices worth $10,000, revenue goes from $100,000 to $125,000: 25% growth, fifteen points from existing clients and ten from new ones. The new-customer side is the subject of Customer Base Growth.
NRR also compounds. If every year's group of customers keeps spending 15% more than the year before, $100,000 of first-year spending becomes about $175,000 after four more years (1.15 × 1.15 × 1.15 × 1.15 ≈ 1.75). At 120% it roughly doubles in four years. That is why a few points of NRR make such a difference to long-term forecasts, and why a fall from 131% to 103% is so painful.
One warning: companies define NRR differently. Some use annual recurring revenue (ARR), the yearly value of current subscriptions; others use revenue actually earned. Some compare one full year with the year before. Twilio compares each quarter with the same quarter a year earlier and then averages the four quarterly rates to get a yearly figure.[^3] Datadog compares ARR from the customers it had twelve months earlier.[^9] HubSpot weights and annualizes its rates, and changed its method in 2025.[^10] Read the definition in the annual report before comparing companies.
Why it matters
Expansion is usually cheaper growth. Winning a new customer means advertising, sales calls, and setup. Growing an existing one often needs only an account manager and a better product. A 2021 survey of private software companies by KeyBanc Capital Markets measured the gap. For 2020, the median company spent $1.67 on sales and marketing to win $1 of new annual revenue from a new customer, but only $0.63 to win $1 from upsells and expansions.[^11]
For many companies it is a large share of growth, and the share rises with size. The same survey broke down where the median company's growth came from in 2020. New customers added about 25% to the revenue it had at the start of the year, upsells and expansions added about 15%, and cancellations and downgrades took away about 13%, for organic growth of 28%. Across the survey, the median company won 36% of its new annual revenue from existing customers. The share grew with company size: 29% for the smallest group ($5 million to $10 million of ARR), and 53% for companies above $100 million.[^11] A big company runs out of new customers to win faster than its customers run out of room to grow.
Strong expansion is less common than headlines suggest. In the KeyBanc survey, the median company's net dollar retention was only about 102%. Companies that met the "Rule of 40", a common test that growth rate plus profit margin should add up to at least 40%, had a median of 109%; the rest had 99%.[^11] Compare that with the venture firm Bessemer Venture Partners, which reported in 2021 that in its data on cloud companies, the middle half had net retention above 100%, with averages of 140% for companies with $1 million to $10 million of ARR and 120% for larger ones. Companies at their stock-market debut averaged 120%.[^12] The two pictures differ partly because the samples differ. Bessemer's are mostly fast-growing companies that venture investors chose to back; KeyBanc's include many ordinary private firms. A benchmark is only as useful as the group it was drawn from.
Investors pay for it. In a study published in November 2025, McKinsey grouped business-software companies by how highly the stock market valued them from 2019 to 2024. The top quarter traded at a median of about 24 times revenue, and the bottom quarter at about 5 times. The top group's net revenue retention was about 113%, compared with about 98% for the bottom group.[^13] That is a correlation, not proof that NRR causes higher values, but it shows how heavily investors weigh it. The compounding arithmetic above is the reason: a customer base that grows on its own makes every new customer worth more.
But more spending is not always more profit. A study of customer records at five firms, in both consumer and business markets, found that 10% to 35% of customers who bought additional products were unprofitable. Those customers accounted for 39% to 88% of each firm's total losses from customers, often because they bought mainly on promotion, returned goods, or needed heavy service.[^14] Expansion creates value only if the added revenue earns more than it costs to serve. The money spent on account managers, discounts and support for existing customers has to be counted too.
The key questions investors ask
1. Direction: is expansion speeding up, steady, or fading?
The trend in NRR matters more than any single reading. Expansion is improving when retained customers grow their spending at a rising rate. It is stable when expansion and contraction stay balanced near the usual trend. It is deteriorating when downsell, lower usage, or churn increasingly offsets upsell and cross-sell.
Twilio shows all three in five years: deteriorating from 131% to 121% to 103% between 2021 and 2023, then roughly stable at 104% in 2024, then improving to 108% in 2025.[^2][^4][^8] A useful habit is to compare the direction of NRR with the direction of total revenue growth. If both fall together, the slowdown is coming from existing customers. If revenue growth falls while NRR holds, the problem is in winning new customers.
2. Drivers: where does the extra spending come from?
The same NRR can come from very different sources, and they are not equally durable:
- Usage can grow fast but also shrink fast, because customers can cut back without cancelling.
- Seats follow the customer's own hiring, and fall with layoffs.
- New products are often the stickiest source. A customer that relies on four of your products is harder to lose than one that uses one. Datadog, which sells software that watches over companies' computer systems, reported that about 58% of its customers used four or more of its products in June 2026, up from about 52% a year earlier. The share using eight or more rose from 14% to 22%.[^9]
- Price increases count too, but they are a one-time lever that customers may resist. See Price and Mix Growth.
Companies rarely split NRR into these pieces, so investors piece it together from clues: product adoption counts, comments about price changes, and whether revenue per customer moves with the customer's own business.
3. Breadth: is expansion spread out or concentrated?
An NRR of 120% can mean most customers grew a little, or one giant customer grew enormously. The second is fragile. In June 2026, Datadog reported a dollar-based net retention rate in the "low-120%'s". But it also said a group it calls "AI-native" customers, which includes its single largest customer, added high-single-digit percentage points to its total revenue growth for the quarter, and that its largest customer had begun to reduce usage in the third quarter of 2026.[^9] Gross revenue retention stayed "in the mid to high 90s", management said, yet the shares fell about 15% in premarket trading.[^15]
Where available, look at results by cohort, the group of customers who joined in the same year, to see whether newer customers expand as much as older ones. Also read the risk factors for the share of revenue from the largest customers. Twilio's top ten were about a tenth of revenue; for Fastly, discussed below, they were well over a third.
Nuances
NRR, GRR, and logo retention. Logo retention counts the share of customers who stay. GRR counts the share of revenue kept. NRR adds expansion on top. A company can have high NRR and weak GRR if many customers leave while a few grow quickly. That combination deserves suspicion. Check both measures, and read Customer Retention for the stay-or-leave side.
Customer size changes what "good" means. Small businesses have less room to grow and close more often, so companies selling to them tend to have lower NRR. Bessemer gives an example: Mindbody, which sells booking software to gyms and studios, had 109% net retention at its stock-market debut on average contracts of about $2,000 a year, while Okta, which sells login and security software to larger firms, had 123% on contracts above $50,000. Bessemer also found that tools for software developers, which spread seat by seat inside a company, had the highest net retention in its portfolio, and that sales, marketing and finance software tended to be lower, because those products are often sold as a complete package up front.[^12] Compare a company with peers that sell to similar customers.
Business models differ. Usage-based businesses such as Twilio and Snowflake see NRR swing quickly. Seat-based software moves more slowly, because contracts usually renew yearly. HubSpot, which sells marketing and sales software mainly to companies with 2 to 2,000 employees, reported NRR of 103.0%, 101.8% and 103.5% for 2023 to 2025, and average subscription revenue per customer of about $11,400 in all three years. Yet revenue grew 21% and 19% in 2024 and 2025, because its customer count rose from about 205,000 to about 289,000.[^10] Low NRR is not a death sentence when a company is good at winning new customers; it just means growth costs more. Payment platforms expand when their merchants' own sales grow or when merchants move more payments to them. Adyen, a Dutch payments company, does not publish an NRR, but it said its 22% growth in net revenue in the second half of 2024 was "driven by share of wallet expansion with existing customers and new business wins".[^16] Banks track products per customer, and consumer subscriptions track average revenue per user (ARPU), which mixes upgrades with price increases. Many companies outside software disclose no NRR, so you may have to estimate expansion from revenue per customer.
Efficiency can shrink usage. For usage-based sellers, a better product can mean less revenue per job. Snowflake warns that improvements such as better storage compression "may make our platform more efficient, enabling customers to consume fewer compute, storage, and data transfer resources to accomplish the same workloads."[^17] The company is betting customers will use the savings to run more work, but that is a bet, not a given.
Expansion naturally fades with age. Young customers have the most room to grow. As a company matures, older customers make up more of its revenue, and they grow more slowly. Snowflake says it expects its NRR "to decrease over the long-term" for this reason.[^17]
A contrasting case: Fastly's biggest customer
Twilio's slowdown was broad: thousands of customers each spent a bit less. Fastly shows the opposite problem, and how the headline number can point the wrong way.
Fastly runs an "edge cloud", a network of servers around the world that delivers websites, videos and apps quickly. Like Twilio, it charges mainly by usage. In 2020, with people stuck at home, internet traffic surged. Fastly's dollar-based net expansion rate, measured over the trailing twelve months, was 137.4% at the end of June 2020.[^18] For the full year revenue rose 45% to $290.9 million.[^19][^20]
But in August 2020 Fastly disclosed that its top customer was ByteDance, the owner of TikTok, which accounted for 13% of revenue in the second quarter. Its ten largest customers together made up 33% of revenue over the previous twelve months. On August 6, 2020, the U.S. president issued an executive order banning transactions with ByteDance.[^18]
On October 14, 2020, Fastly said third-quarter revenue would be $70 million to $71 million, below its own guidance of $73.5 million to $75.5 million, and withdrew its forecasts. Usage by its "previously disclosed largest customer did not meet expectations", and a few other customers had used less than expected.[^21] The next morning the stock was down 27.7%.[^22]
Here is the twist. For the twelve months ending September 30, 2020, the same quarter as the shortfall, Fastly's expansion rate hit a new high of 147.4%.[^23] A trailing-twelve-month measure looks backward, so it still counted months when the giant customer was growing fast. Fastly's annual report later explained that its largest customer had "significantly decreased its usage in the second half of the year", that its top ten customers made up about 38% of 2020 revenue and its top five about 27%, and that many large customers' minimum commitments were "relatively low compared to their expected usage", so they could move traffic elsewhere quickly.[^19] For 2021 the expansion rate fell to 120.9% and revenue growth slowed to 22%.[^20]
An investor who read 147% as broad, durable strength had the thesis backwards. An investor who read the concentration disclosures, and knew that NRR lags, could have discounted it. The lesson is not that high NRR is bad. It is that NRR is an average, and an average can be carried by one customer.
How it connects. Expansion is one slice of Organic Growth, driven by Volume Growth or Price and Mix Growth. Signed but unused customer commitments show up in Backlog Conversion. All of these roll up into Revenue Growth. Whether expansion is broad or carried by a few accounts is also a question of Revenue Concentration, and whether it rests on customers being locked in is a question of Switching Costs.
A worked example: Snowflake
Snowflake runs a cloud platform where companies store and analyze data. Customers pay mainly for the computing and storage they use, which Snowflake calls "product revenue" (more than 90% of the total). For NRR, it takes the customers on capacity contracts (prepaid commitments) who used the platform in the first month of a two-year window, and divides their product revenue in the second year by their product revenue in the first. A customer who stopped using the platform stays in the group and counts as zero.[^17] Snowflake's fiscal year ends January 31, so fiscal 2026 ended January 31, 2026.
| Fiscal year | Product revenue | Growth | NRR at year end | Total customers | Customers above $1 million a year |
|---|---|---|---|---|---|
| 2021 | $553.8 million | — | 168% | 4,139 | 79 |
| 2022 | $1,140.5 million | +106% | 178% | 5,967 | 186 |
| 2023 | $1,938.8 million | +70% | 158% | 7,744 | 329 |
| 2024 | $2,666.8 million | +38% | 133% | 9,384 | 449 |
| 2025 | $3,462.4 million | +30% | 126% | 10,996 | 576 |
| 2026 | $4,472.3 million | +29% | 125% | 13,245 | 733 |
| Q2 fiscal 2027 (quarter to July 31, 2026) | $1,491.9 million | +37% | 126% | 14,554 | 828 |
Growth rates are calculated from reported product revenue. NRR and customer counts are the latest figures Snowflake reported for each date; the "above $1 million" column counts customers with more than $1 million of product revenue over the previous twelve months.[^24]
Direction. NRR fell steeply, from 178% to 126%, over three years. The fiscal 2024 annual report explains part of the reason: customers were working to "rationalize budgets" and "optimize consumption", for example by keeping data for shorter periods.[^25] Since mid-2025, though, NRR has held at 125–126% for five straight quarters.[^24] The trend has moved from deteriorating to stable, with a slight lean toward improving.
Drivers. Product revenue growth sped up to 37% in the July 2026 quarter, the third straight quarter of acceleration. Snowflake's quarterly report says the increase came "primarily" from "increased consumption of our platform by existing customers".[^24] Management credits both the core data platform and new AI features. One new AI tool, CoCo, was in use at more than 9,100 customer accounts, up by more than 2,000 in the quarter, and the company raised its fiscal 2027 product revenue forecast to $6.07 billion, or 36% growth.[^26] That is usage plus new products, the mix discussed under the second key question. Usage-based growth can still reverse, as in fiscal 2024.
Breadth. Expansion looks broad. Customers paying more than $1 million a year rose 27% to 828, and Snowflake added 692 net new customers in the quarter, reaching 14,554.[^24][^26] Large customers matter: the 790 Forbes Global 2000 companies among Snowflake's customers at the end of fiscal 2026 contributed about 43% of revenue.[^17] That is a big group, not one whale, which is a different risk from Fastly's in 2020. Snowflake does not publish gross retention or full cohort data, so the breadth judgment rests on partial evidence.
Splitting the growth. Put NRR next to total growth. In fiscal 2026 product revenue grew 29% and NRR was 125%. Roughly, then, about 25 of the 29 points came from customers Snowflake already had, and about 4 points from everyone else, mainly customers who joined during the period. The split is approximate, because NRR covers only capacity-contract customers, not those who pay as they go. Still, it shows that Snowflake is mostly an expansion story: four-fifths or more of its growth comes from existing customers using more. The same arithmetic on Twilio's 2025 results gives a similar cross-check. Its revenue rose $609.2 million, of which $242.0 million came from new accounts, so about $367 million, or 60%, came from existing ones, close to what an expansion rate of 108% on $4.46 billion of prior-year revenue implies.[^8]
Peers and targets. Snowflake's 126% compares with Datadog's "low-120%'s", Twilio's 108% for 2025 and HubSpot's 103.5%.[^9][^8][^10] Each company defines the number differently, so treat this as a rough ranking, not a precise contest. Snowflake has not set a public NRR target. Its own warning that NRR will fall over the long run is the closest thing to one.[^17]
A lag. NRR compares a full year with the year before, so it moves slowly. Growth accelerated for three quarters while NRR rose one point. If the acceleration is real, NRR should follow. Fastly in 2020 showed the same lag working in reverse.
A note on restated counts. Snowflake adjusts past figures. It first reported 5,944 customers for the end of fiscal 2022 and later showed 5,967; it first reported 9,437 for fiscal 2024 and later 9,384.[^24] The changes are small, but they are a reminder to take every figure from the latest report.
The thesis. An investor could state: "Snowflake's existing customers will keep expanding at least as fast as they do today." To make it testable: "Snowflake reports a net revenue retention rate of 125% or higher as of January 31, 2028, in its fiscal 2028 annual report." As of September 2026, that report should arrive around March 2028, about 18 months later. The thesis bets that new AI workloads will offset the long-term decline that management itself expects.
How new evidence updates it. Each quarterly report is a checkpoint:
| Next quarterly reading | What it would suggest |
|---|---|
| NRR of 127% or more, and product revenue above guidance ($1,588 million to $1,593 million for the October 2026 quarter) | Expansion is improving; the thesis becomes more likely |
| NRR of 125–126%, revenue roughly in line with guidance | Stable; little change to the odds |
| NRR of 123% or below, especially with renewed comments about customers optimizing spending | Deteriorating; the thesis becomes less likely |
The quarterly guidance comes from Snowflake's September 2026 results release.[^26] Product revenue that beats guidance is an early signal, because NRR reflects it only later. A sudden drop in the count of $1 million customers, or a disclosure that one customer accounts for an unusual share of growth, would be a warning even if NRR held.
Common traps
- Shifting definitions. Companies measure NRR differently, and revise it: Snowflake first reported fiscal 2024 NRR as 131% and later showed it as 133%.[^24] HubSpot changed its method in 2025, which moved its 2024 figure from 102.2% to 101.8%.[^10] Compare like with like.
- Different averaging. A rate averaged over four quarters, like Twilio's, and a rate measured at one date, like Snowflake's, can tell different stories about the same year. So can a rate based on ARR versus one based on revenue.
- The lag. NRR looks back over a year. It can hit a record just as the trend turns, as Fastly's did in September 2020. Watch quarterly revenue alongside it.
- New customers counted as expansion. Revenue per customer can rise just because new customers are bigger. NRR tracks a fixed starting group.
- Acquisitions dressed up as expansion. Buying a company brings its revenue and customers. Good definitions keep acquired customers out until a year has passed, as Twilio's does.[^3] Check that the one you are reading does too.
- Price disguised as loyalty. A one-time price rise lifts NRR without anyone using more.
- Usage that vanishes. Usage-based expansion can reverse when budgets tighten, as at Twilio.
- One whale. A single huge customer can carry the whole figure, as ByteDance did at Fastly. Read the customer concentration numbers in the risk factors.
- Benchmarks from the wrong crowd. A venture firm's portfolio and a broad survey give very different "typical" NRR. Compare with companies of similar size, selling to similar customers.
- Gamed cross-sell metrics. From 2012 to 2016, Wells Fargo promoted its "cross-sell" metric, a measure of selling more products to existing customers. The SEC found it was inflated by accounts that were "unused, unneeded, or unauthorized". The bank paid $500 million to the SEC, part of a combined $3 billion settlement with the SEC and the Department of Justice.[^27]
- Expansion that loses money. Heavy discounts or high service costs can make added revenue unprofitable.
Checklist
When you study a company's existing-customer expansion, ask:
- Does the company report net revenue retention, and exactly how is it defined (revenue or ARR, one date or an average, which customers are included)?
- Has the definition changed, or have past figures been restated?
- What is the trend over the last three to five years, and is it moving in the same direction as total revenue growth?
- How far apart are net and gross retention, and is churn hiding behind expansion?
- Is expansion coming from usage, seats, new products, or price?
- Is it broad across customers and cohorts, or driven by a few large accounts? What share of revenue comes from the top five or ten customers?
- How much of total growth comes from existing versus new customers?
- How does the figure compare with peers that sell to similar-sized customers?
- Could the lag in a trailing measure be hiding a recent turn? What does the latest quarter's revenue say?
- Is the expansion profitable after discounts and service costs?
Key takeaways
- Existing-customer expansion is growth from current customers: more seats, usage, products, or price, minus downsell and churn.
- Net revenue retention measures it by comparing what last year's customers spend now with what they spent then, leaving out new customers. Gross retention ignores expansion and never exceeds 100%; see Customer Retention.
- Expansion is usually cheaper than winning new customers, it becomes a bigger share of growth as companies grow, and investors reward it. But it can fade, reverse, or be concentrated in a few accounts, as Twilio and Fastly showed in different ways.
- NRR is an average that looks backward. Read the definition, compare it with recent revenue, and check who is doing the spending.
- Judge it by direction, drivers, and breadth, against peers that sell to similar customers.
Data needed to support this thesis
To build and keep updating an existing-customer expansion thesis, you need the following data.
Required
| Data | What it tells you | Where to find it |
|---|---|---|
| Retention metrics | Net revenue retention (and gross retention, if given): how much last year's customers spend now, with churn and downsell included | The key metrics section of annual reports (Form 10-K) and quarterly reports (Form 10-Q); results releases |
| Customer revenue history | How spending by the same group of customers has changed over time | Revenue and retention figures across at least three comparable periods of the same filings |
Helpful
| Data | What it tells you | Where to find it |
|---|---|---|
| Cohort data | Whether groups of customers who joined in the same period keep growing | Investor presentations and earnings calls; few companies publish it in full |
| Product adoption | Whether customers are taking up new products, which drives cross-selling | Results releases and earnings calls, which often name adoption counts for new products |
| Contract data | Contract length, renewal dates, and pricing terms that shape future spending | The revenue and remaining performance obligations notes in filings; the count of large customers |
The comparisons must be like-for-like. Companies define retention differently, and sometimes revise past figures. Use one company's latest definition across all periods, and keep new customers out of the group you are tracking. When comparing companies, compare those that sell to similar-sized customers, and remember that each one's formula is a little different.
The thesis is usually updated by three kinds of events: earnings releases (new retention and revenue figures), customer metric updates (new retention or customer counts), and product launches (new products that existing customers can adopt). A typical horizon for checking the thesis is about a year and a half.
In the Snowflake example, the required data was its net revenue retention rate and product revenue for each fiscal year and for the latest quarter, from its annual and quarterly reports and earnings release.[^17][^24][^26] The helpful data was the count of customers paying more than one million dollars a year, the total customer count, the Forbes Global 2000 revenue share, and the number of accounts using a new AI tool, which are partial stand-ins for contract and product adoption data.[^24][^26][^17] The peer comparison drew on Datadog, Twilio and HubSpot filings.[^9][^8][^10] The example did not have cohort data or gross retention.
[^1]: Twilio, Form 10-K for 2020 (dollar-based net expansion rate of 137% for 2020). SEC filing. [^2]: Twilio, "Twilio Announces Fourth Quarter and Full Year 2021 Results" (February 2022). Twilio. [^3]: Twilio, Form 10-K for 2023 (Lawson described as co-founder; definition of the dollar-based net expansion rate, including the treatment of acquisitions; customer industries; ten largest accounts at 10%, 12% and 11% of 2023, 2022 and 2021 revenue). SEC filing. [^4]: Twilio, "Twilio Announces Fourth Quarter and Full Year 2023 Results" (February 2024), with 2022 figures from the 2022 results release. Twilio. [^5]: Twilio, "Twilio Announces Fourth Quarter and Full Year 2022 Results" (February 2023). Twilio. [^6]: The Motley Fool, "Is Twilio Stock a Buy After It Crushed Expectations?" (August 23, 2023). Motley Fool. [^7]: Twilio, Form 8-K (January 8, 2024), resignation of Jeff Lawson and appointment of Khozema Shipchandler as CEO. SEC filing. [^8]: Twilio, Form 10-K for 2025 (2023–2025 revenue, growth, active customer accounts and dollar-based net expansion rate; $242.0 million of 2025 revenue growth from new accounts). The existing-customer share is calculated. SEC filing. [^9]: Datadog, Quarterly Report on Form 10-Q for the quarter ended June 30, 2026 (net retention definition and rate, product adoption, AI-native cohort, largest customer). SEC filing. [^10]: HubSpot, Form 10-K for 2025 (customers, average subscription revenue per customer, net revenue retention for 2023–2025 and earlier figures under the previous method, revenue growth). SEC filing. [^11]: KeyBanc Capital Markets, 2021 Private SaaS Company Survey, 2020 data for companies above $5 million of ARR (customer acquisition cost ratios for companies above $5 million of sales and marketing spend); median net dollar retention 101.8%; share of new ARR from upsells and expansions by ARR size; Rule of 40 comparison. PDF. [^12]: Mary D'Onofrio, Ethan Ding and Atlas Editors, "Scaling to $100 Million," Bessemer Venture Partners (September 21, 2021). Figures are from Bessemer's own sample of private cloud companies. Bessemer. [^13]: McKinsey & Company, "The net revenue retention advantage: Driving success in B2B tech" (November 19, 2025). McKinsey. [^14]: Denish Shah, V. Kumar, Yingge Qu, and Sylia Chen, "Unprofitable Cross-Buying: Evidence from Consumer and Business Markets," Journal of Marketing 76, no. 3 (2012). SAGE. [^15]: Datadog second-quarter 2026 earnings call transcript (August 6, 2026). Investing.com. [^16]: Adyen, Shareholder Letter H2 2024 (February 2025). Adyen. [^17]: Snowflake, Form 10-K for fiscal year ended January 31, 2026 (NRR definition and outlook, efficiency risk, Forbes Global 2000 customers). SEC filing. [^18]: Fastly, Form 10-Q for the quarter ended June 30, 2020 (ByteDance at 13% and 12% of revenue for the three and six months; ten largest customers; executive order; trailing-twelve-month expansion rate of 137.4%). SEC filing. [^19]: Fastly, Form 10-K for 2020 (revenue; largest customer's reduced usage; top ten and top five customers; minimum commitments). SEC filing. [^20]: Fastly, Form 10-K for 2021 (2019–2021 revenue; dollar-based net expansion rate of 120.9% for 2021 and 143.6% for 2020). Growth rates are calculated. SEC filing. [^21]: Fastly, "Fastly Provides Preliminary Third Quarter Revenue Results" (October 14, 2020), Exhibit 99.1 to Form 8-K. SEC filing. [^22]: The Motley Fool, "Why Fastly Stock Plunged Today" (October 15, 2020), share price as of 9:59 a.m. EDT. Motley Fool. [^23]: Fastly, Form 10-Q for the quarter ended September 30, 2020 (trailing-twelve-month expansion rate of 147.4%). SEC filing. [^24]: Snowflake Forms 10-K for fiscal 2022 (fiscal 2021 customers; fiscal 2022 first reported as 5,944), 2023 (fiscal 2021 NRR and $1 million customers; fiscal 2022 customers), 2024 (fiscal 2022 NRR, 177% in the prior 10-K, and $1 million customers; fiscal 2023 customers; fiscal 2024 customers first reported as 9,437), 2025 (fiscal 2023 NRR and $1 million customers; fiscal 2024 customers) and 2026 (fiscal 2024–2026 NRR and $1 million customers, fiscal 2024 NRR first reported as 131%; fiscal 2025 customers), and the Q2 fiscal 2027 10-Q (quarterly NRR, fiscal 2026 and July 2026 customer counts, $1 million customers, drivers of growth). [^25]: Snowflake, Form 10-K for fiscal year ended January 31, 2024. SEC filing. [^26]: Snowflake, "Snowflake Reports Financial Results for the Second Quarter of Fiscal 2027" (September 2, 2026). SEC filing. [^27]: U.S. Securities and Exchange Commission, "Wells Fargo to Pay $500 Million for Misleading Investors About the Success of Its Largest Business Unit" (February 21, 2020). SEC.