From Shubhi K | Product & Market Analysis
Community-Led Growth: What Supabase, Vercel, dbt Labs and 3 Others Disclose
On this page
We checked what 6 developer tools companies have published about community as a channel, and only 1 put a pipeline number on it: 28%, inside a vendor case study. Community-led growth is the most cited and least measured motion in developer marketing. Supabase, Vercel and dbt Labs publish member counts, contributors and conference crowds. None of them publishes what that community is worth in revenue.
Key takeaways
- Only 1 of 6 companies audited here publishes a community pipeline figure. Census reported that over 28% of pipeline and over 50% of closed deals were community-engaged, in a case study written by its own vendor, Common Room.
- The most rigorous community numbers sit in SEC filings, not blogs. HashiCorp's 2021 S-1 disclosed 36,000+ user group members and about 100 million downloads, and GitLab's S-1 disclosed 2,600+ open source contributors.
- Private developer tools firms disclose activity, not outcomes. Vercel reports contributors and downloads, dbt Labs reports 100,000+ community members, and Supabase's own Series E post gave no community figure at all.
- Size metrics cannot tell you whether community is a channel. A channel needs a sourced or influenced pipeline number with a stated definition, and a community member count answers a different question entirely.
The short answer. Developer tools companies publicly disclose community size: members, contributors, downloads and event attendance. They almost never disclose community-attributed pipeline or revenue. In this audit, the only pipeline figure came from a vendor case study with no stated method. If you fund community as a channel, you will be building the measurement yourself.
What community-led growth claims, and what it would take to prove it
Community-led growth is the idea that a group of users who talk to each other produces customers, retention and product input more cheaply than paid acquisition. The community is the funnel, not a support forum bolted onto one.
That is a channel claim. A channel claim has a specific shape. It says this source produced this much pipeline, at this cost, over this period. Paid search can answer that sentence in an afternoon.
Community marketing usually answers a softer sentence. It says the community is large, active and growing. Those are true statements about an asset. They are not statements about a channel, and the gap between the two is what this post measures.
The test we applied is simple. For each company, we looked for 5 kinds of public number: members, contributors, event attendance, downloads or usage, and attributed pipeline. We counted only figures we could open at a primary or company source during this research.
The selection is deliberate rather than random. Supabase, Vercel and dbt Labs are the companies most often named when founders talk about developer community strategy. GitLab and HashiCorp are included because they filed S-1s, which forced their community claims into a document with legal liability. Census is included because it is the only company we found that published a pipeline share.
The community disclosure table, company by company
This is the core of the audit. Each cell is a figure the company published, with the date of the source. Where a cell says none, we did not find a number in the sources we opened, which is not proof that none exists anywhere.
| Company | Members | Contributors | Events | Downloads or usage | Attributed pipeline |
|---|---|---|---|---|---|
| Supabase | None in its Series E post | Named, not counted: a $1M community round for contributors | None | 113.8k GitHub stars, read off the public repo | None |
| Vercel | None | 570 new Next.js contributors in a year (2024) | 1,000+ at Next.js Conf 2024 | 500M+ Next.js downloads in 12 months (2025) | None |
| dbt Labs | 100,000+ data professionals | None | 2,000+ at Coalesce 2025, 10,000 online | 1,500 weekly Fusion projects (2025) | None |
| GitLab | None | 2,600+ (S-1, 2021) | None | None in the sections read | Qualitative only: free tier is "targeted lead generation" |
| HashiCorp | 36,000+ user group members (S-1, 2021) | None | 140+ user group chapters | About 100M downloads in fiscal 2021 | Qualitative only |
| Census | 7,500+ in its OA Club, about 2,000 in Slack | None | None | None | 28%+ of pipeline, 50%+ of closed deals |
Sources are listed in the references. Supabase's star count was read from its GitHub repository in October 2026 and is a platform counter, not a company disclosure. Census figures are from a vendor case study and are directional.
The two filings that put community numbers under oath
The best community metrics in this set did not come from community teams. They came from securities lawyers. A registration statement is a document where a misleading claim carries legal exposure, so the numbers in it tend to be narrower and better defined.
GitLab: contributors as engineering capacity
GitLab's S-1, filed on 17 September 2021, credits over 2,600 contributors in its global open source community as of 31 July 2021. It links them to a delivery record of a release on the 22nd of every month for 118 months in a row.
Notice how GitLab frames the community. It is presented first as development capacity, and second as a sales input. The filing says its free tier exists to "encourage use", "solicit contributions" and "serve as targeted lead generation for paid customers". That is a channel claim in plain words, with no number attached to it.
The filing reports customers instead. Its Base Customers, those above $5,000 of annual recurring revenue, grew from 1,662 in January 2020 to 3,632 by July 2021. The link between contributors and those customers is asserted, not measured. The same filing also warns that users may stay on the free product rather than convert.
HashiCorp: the most complete community funnel on record
HashiCorp's S-1, filed on 4 November 2021, is the closest thing to a community funnel any company in this set has published. It reports about 100 million downloads in fiscal 2021 and an estimated 11,000 organisations that had downloaded at least one product.
It reports 36,000+ members in more than 140 user group chapters across 50+ countries. It reports over 10,000 people certified on its products. Then it reports the commercial end: 558 customers above $100,000 of ARR, up from 419 a year earlier.
Read those numbers in sequence and a funnel appears. The filing describes it in words too, saying the open source footprint lets its enterprise sales teams "target and sign" customers. What it never states is the conversion rate between the stages. HashiCorp published the top and the bottom of the funnel and left the middle to the reader.
That still beats every blog post in this audit. If your board wants a model for community disclosure, a 2021 prospectus is a better template than any community report written since. A companion analysis of what developer tools companies disclose in their filings about agents found the same pattern: filings define their terms, press releases do not.
What private developer tools companies choose to disclose
Private companies owe the public nothing. What they choose to publish is therefore a signal of what they think flatters them. In every case below, that is scale.
Supabase: community named, never counted
Supabase is the company most associated with community as a growth engine. Its Series E post, published 3 October 2025, announced $100M at a $5B pre-money valuation. It said the community "continues to grow" and promised a $1M community round open to early customers and contributors.
That post contains no community figure at all. No member count, no contributor count, no attribution claim. The widely repeated figure of 4 million developers appeared in press coverage of the round, including Pulse 2.0, but not in the company's own announcement that we opened.
The only hard community number we could read was on GitHub. The main Supabase repository showed 113.8k stars and 16.9k forks when we checked in October 2026. Stars measure attention. They do not measure use, and they certainly do not measure purchase.
The community round is the most interesting fact here. Letting contributors buy equity is a strong statement of how much the founders value them. It is also a statement made in dollars of cap table, not dollars of pipeline.
Vercel: contributors and downloads, carefully dated
Vercel publishes the most disciplined activity numbers of the private firms. Its Next.js Conf 2024 recap, dated 25 October 2024, reported 1,000+ attendees in San Francisco, 4,000+ merged pull requests and 570 new community contributors since the previous conference.
Its Series F post, dated 30 September 2025, said Next.js had been downloaded over 500 million times in the prior 12 months. It said the AI SDK went from 446,012 to 3,209,817 weekly downloads in a year. Reuters, covering the same round, reported that Vercel's revenue rose 82% and its user base doubled.
So Vercel discloses the open source top of funnel and the revenue bottom, separately. What it does not say is how much of that 82% came through Next.js users. That number almost certainly exists inside the company. Its absence from public material is a choice.
dbt Labs: the biggest member count, no conversion
dbt Labs runs one of the largest communities in B2B software. Its own community guidelines say the dbt Community includes over 100,000 data professionals. Its October 2025 Coalesce post reported more than 2,000 people at the general session in Las Vegas and another 10,000 online.
The commercial side is reported elsewhere. When dbt Labs and Fivetran announced their all-stock merger in October 2025, they said the combined company would be approaching $600M in ARR. Neither release we found split out how many dbt Cloud customers started in the open source community.
dbt's community page uses the phrase "active members" without defining active. That word is doing a lot of work. Slack counts accumulate. People join for one question and never leave, which makes a member total closer to a lifetime visitor count than a measure of a live audience.
The only community pipeline number in the set
Census, the reverse ETL company, is the only firm we found that published community pipeline attribution. It did so in a case study on the website of Common Room, a community analytics vendor and Census supplier.
What the Census case study says
The case study reports that Census built its Operational Analytics Club to more than 7,500 data practitioners in two years, with about 2,000 in Slack. It reports that more than 50% of closed deals and over 28% of pipeline were community-engaged.
It also reports that deals involving a club member closed 40% faster than deals without one. The measurement method named is Common Room's business impact reporting, connected to Census's CRM.
What is missing from it
Almost everything a finance team would ask for. The page does not define "community-engaged". It gives no time window and no deal count. It does not say whether 40% is a median or a mean. It is undated, and it is published by the vendor whose product produced the number.
There is also a selection problem that no tool can fix. People who join a vendor's community are already more interested in that vendor. Deals with engaged prospects close faster everywhere, community or not. Without a comparison against similarly engaged prospects who skipped the community, a 40% speed gap is correlation.
We still think the Census figure is useful. It is the only published attempt to put community on the same scorecard as other channels, and it names its tooling. Treat it as a proof that the measurement is possible, not as a benchmark for what yours should show. The same problem with vendor-written proof is covered in the piece on what makes a case study verifiable.
Measured claims versus narrative claims about community marketing
Most writing on community led growth mixes two kinds of statement without marking the difference. Separating them is the most useful thing you can do before you read the next community report.
| Claim type | Example from this audit | What it can support |
|---|---|---|
| Counted, defined, dated | GitLab: 2,600+ contributors at 31 July 2021 | Statements about the size of the asset at a point in time |
| Counted, undefined | dbt Labs: 100,000+ community members, "active" not defined | Order of magnitude only |
| Platform counter | Supabase: 113.8k GitHub stars | Attention, not usage or revenue |
| Attribution with method undisclosed | Census: 28%+ of pipeline community-engaged | Directional evidence that measurement is possible |
| Narrative | GitLab: free tier serves as "targeted lead generation" | The company's own theory of its funnel, nothing more |
Our view is blunt. If a community claim cannot be placed in the top 2 rows or the attribution row, it should not appear in a budget memo. Narrative claims belong in strategy discussions, where everyone knows they are hypotheses.
The pattern across the 6 companies is consistent. Activity is counted, outcomes are narrated. That is not because developer tools companies are hiding something. It is because community touches pipeline in ways the standard attribution stack cannot see, a problem examined in detail in the post on dark discovery and attribution measurement.
Where this argument is weakest
We would be guilty of the thing we are criticising if this section were missing. The audit has real limits.
The sample is small and chosen, not random
Six companies is a sample picked to cover the names founders cite and the firms whose filings forced disclosure. A broader sweep might find a company that does publish community attribution well. We also searched only what was public and opened during this research. Conference talks, podcasts and investor letters may carry numbers we did not reach, and a blank cell in our table means not found, not absent.
Measurement may be the wrong frame for community
The strongest counter-argument is that community is not a channel at all. It may be closer to brand or product quality: something that raises conversion across every channel without owning any of it. On that view, demanding sourced pipeline from community is like demanding sourced pipeline from good documentation.
We take this seriously. Census itself chose not to assign revenue targets to its community, according to the same case study. That is a defensible choice. The problem is that companies cannot have it both ways: either community is a channel and should be measured like one, or it is infrastructure and should be budgeted like one. Most community budgets are pitched as the first and defended as the second.
The private firms also have good reasons to stay quiet. Publishing a community conversion rate hands competitors a benchmark. The absence of a number is weak evidence that no number exists inside the company.
The community metrics worth tracking as a channel
If you are deciding whether to fund a developer community strategy, the useful question is which numbers would change your mind. Here is the set we would track, ordered from cheapest to most persuasive.
| Weak metric | Version that works | Why it is better |
|---|---|---|
| Total members | Members active in the last 28 days, with active defined | Census used a 28-day rolling average, which filters out one-visit joiners |
| GitHub stars | Organisations running the open source product, by company domain | HashiCorp's 11,000 downloading organisations is a stage in a funnel, stars are not |
| Event registrations | Accounts with an attendee that opened an opportunity within 90 days | Connects an event to a deal with a window you can audit |
| Community-touched pipeline | Pipeline where a member engaged before the opportunity opened | Timing rules out members who joined after sales was already involved |
| Deal velocity, members versus everyone | Deal velocity, members versus equally engaged non-members | Removes most of the selection bias in the Census-style comparison |
| Contributor count | Contributors employed at paying or target accounts | Shows whether the people shipping code are also people who buy |
Two of those rows matter more than the rest. The timing rule on pipeline is the one that separates sourcing from tagging. Without it, any deal where someone once joined your Slack counts as community, and the number inflates as the community grows regardless of whether it causes anything.
The matched comparison on deal velocity is the other. It is harder to build, and it is the only version that would survive a sceptical CFO. If your community vendor cannot produce it, ask why. A similar test applies to community mentions as a driver of AI search visibility, where the same activity-versus-outcome gap shows up.
We would also publish whatever you measure, with the method stated. As the table above shows, the bar for community disclosure is currently low enough that a single defined, dated pipeline figure would put your company ahead of the 6 in this audit. That is the same logic behind publishing original research as a content moat: a specific number with a stated sample gets cited because almost nobody else offers one.
One more consideration for founders weighing community against product-led motions. Agents are starting to evaluate and adopt developer tools on behalf of humans, a shift traced in the analysis of product-led growth when the user is an agent. Agents do not join Slack. If more of your top of funnel becomes automated, the share of buyers who ever touch your community may shrink, and you will want a baseline from before that happened.
Frequently asked questions
What is community-led growth?
Community-led growth is a go-to-market approach where a community of users who interact with each other drives acquisition, retention and product feedback. Instead of paid ads or outbound sales generating demand, members recommend the product, answer each other's questions and contribute code or content. It is common in developer tools, where open source projects, Slack groups, user meetups and conferences often sit at the top of the funnel.
How do you measure community as a channel?
Measure community as a channel by joining community activity to your CRM and counting pipeline where a member engaged before the opportunity opened. Track 28-day active members rather than total members, and compare deal velocity against equally engaged prospects who did not join. Without a timing rule and a matched comparison, a community pipeline figure mostly measures interest that existed already.
Which companies publish community-attributed pipeline numbers?
Very few. In our audit of 6 developer tools companies, only Census published one, reporting that over 28% of pipeline and over 50% of closed deals were community-engaged. That figure appeared in a case study by its community analytics vendor, Common Room, without a definition, time window or sample size. Supabase, Vercel, dbt Labs, GitLab and HashiCorp published community size and activity numbers instead.
How big is the dbt community?
dbt Labs says its community includes over 100,000 data professionals, according to its community guidelines read in October 2026. The company does not define what counts as a member or an active member. Its Coalesce 2025 conference drew more than 2,000 people to the general session in Las Vegas and about 10,000 online, according to a dbt Labs post from October 2025.
What are the best B2B community metrics to track?
The most useful B2B community metrics are 28-day active members, organisations using the product by company domain, opportunities opened within 90 days of an event, pipeline where community engagement came before the opportunity, and deal velocity compared against similarly engaged non-members. Total members, GitHub stars and event registrations are easy to report but say little about revenue on their own.
Where to start
If you run a community today, pull 1 number this week: the share of opportunities opened in the last 2 quarters where a contact engaged with your community before the opportunity date. If you cannot produce it, the gap is in your data model, not your community, and that is the first thing to fix.
If you are deciding whether to fund one, write down the number that would make you stop before you spend anything. A community that is measured against a stated kill line can earn a bigger budget later. One that is only ever reported by size will be the first line cut when money gets tight.
Related in this series
Community is one of several channels whose influence escapes standard attribution. For how that gap affects budget decisions, read the analysis of dark discovery and what to measure instead.
References
- HashiCorp, Inc., Form S-1 Registration Statement, filed 4 November 2021. Used for downloads, user groups, certification and customer figures.
- GitLab Inc., Form S-1 Registration Statement, filed 17 September 2021. Used for contributor count, free tier description and Base Customer figures.
- Supabase, Supabase Series E, 3 October 2025. Used for funding terms and the community round.
- GitHub, supabase/supabase repository, read October 2026. Used for star and fork counts.
- Vercel, Recap of Next.js Conf 2024, 25 October 2024. Used for attendance, contributors and downloads.
- Vercel, Series F, 30 September 2025. Used for Next.js and AI SDK download figures. Revenue growth of 82% is from Reuters coverage of the same round, a secondary source.
- dbt Labs, Community Rules of the Road, docs.getdbt.com, undated, read October 2026, and dbt Labs, Coalesce 2025: Rewriting the Future, 14 October 2025. Used for community size and event attendance. Combined ARR approaching $600M is from the Fivetran and dbt Labs merger announcement, 13 October 2025.
- Common Room, How Census used Common Room to scale growth and show ROI, undated. Used for all Census figures. Vendor-published.
The weakest part of this source base is that the only pipeline figure is vendor-published and undated, and the two filings that define their terms are from 2021. Community numbers in fast-growing companies go stale quickly, so re-check every figure before citing it.
Related reading