43 SaaS Community Statistics
43 figures on what SaaS communities cost and return, and which ones AI answers cite: Reddit far more than any vendor’s own forum.
On this page
- Highlights
- Most Community Teams Have Five Full-Time People or Fewer
- What Customer Communities Return in Support, Renewals and Pipeline
- The Platforms Communities Use Most Sit Behind a Sign-In
- Which Communities the AI Answers Cite
- How to Divide the Work Between Your Community and the Forums
- Frequently Asked Questions
Empact Newsletter
One email a month on what is working in AI search and SaaS go-to-market.
When Google AI Mode answers our software partners’ buyer questions with sources, it cites a Reddit thread in 9.3% of answers in the median partner’s panel. A community run by any software vendor gets 0.3%.1 If your community budget is meant to keep customers and win buyers at once, that gap matters before next year’s budget is set.
This page brings together the published research on what SaaS communities return and what they cost, and our own count of which communities the AI answers read. It is for the marketing or community lead deciding where that budget goes.
Empact Partners gets SaaS products found and tried by buyers who never talk to sales, in AI answers, forums and review sites, through product-led growth. Which communities those buyers and AI answers read decides where a community budget can do that job.
Highlights
- 16.7 times as many AI Mode citations went to Reddit as to every vendor community combined, in the median business software panel.1
- Under 1.5% of ChatGPT’s sourced answers cited Reddit, independent forums or vendor communities, each counted alone, in every one of our 13 software panels.1
- 17% of community teams have nobody working on community full time, up from 4% in 2021.3
- 21%, 20% and 16% of communities host their forums on WhatsApp, Slack and Discord, three spaces behind a sign-in.3
- 2.2 times the expansion rate went with Gainsight’s most engaged community accounts, with gross retention 3 points higher.4
The strongest returns these communities show come from customers who already bought, and the three platforms communities use most sit behind a sign-in. The visibility a buyer meets in AI Mode comes from public threads, mostly on Reddit, and ChatGPT cites communities so rarely that its shortlist gets made from other pages.
Most Community Teams Have Five Full-Time People or Fewer
Running a customer community stopped being an experiment for B2B companies some time ago. Gainsight, which sells community software, surveyed more than 200 of them with the benchmarking firm Benchmarkit in 2024:
- 31% of the companies had no customer community, so almost seven in ten had one.2
- 59% of those without one were planning to start one.2
The people who run these communities are few, and more teams have nobody full time each year. CMX, the professional network for community managers, surveyed 589 of them in May and June 2025, a little under half of them in technology companies:
- 86% of their communities host online forums, chat rooms or message boards.3
- 31% of community teams are one full-time person, and 34% are two to five people.3
- 17% have nobody working on community full time, up from 4% in 2021.3
- 60% of communities with 1,000 to 9,999 members run on two to four moderators.3
Community Teams With Nobody Working Full Time
Share of CMX respondents whose community team has no full-time person, 2021 to 2025
View the data
| Label | Value |
|---|---|
| 2021 | 4% |
| 2022 | 8% |
| 2023 | 10% |
| 2024 | 16% |
| 2025 | 17% |
Counting the teams with nobody full time, 82% of community teams have five full-time people or fewer. That sets the real limit on a community, which is how many member questions a handful of people can answer well. Spread the same team across a Slack workspace, a forum and a Discord server, and member questions wait longer for an answer in each of them.
Money follows the same pattern in CMX’s budget questions:
- 26% saw it grow in 2025, 18% saw it cut, and a quarter could not say which way it moved.3
- 37% named proving the return, the budget challenge named most often, and 20% named the cost of tools.3
- 44% expect more investment next year, about where the figure stood in 2023.3
So the scarce thing in a community is people’s time, and fewer than half of respondents expect their company to invest more next year. For a team this size, the choice that matters most is how many spaces to run. Gabby Saayman, who wrote this page and consults on go-to-market at Empact Partners, would keep it to one:
A community lives on answered questions. With one or two people, run one space where every member question gets a good answer quickly, and close the spaces nobody has time to answer in. A busy channel full of unanswered questions teaches members to stop asking.
Gabby Saayman on LinkedInOpens in a new tabWhat Customer Communities Return in Support, Renewals and Pipeline
If the cost is people, the return has to be something those people can point to. CMX asked each team for its single top business objective, and most of the answers sit after the sale:
The Top Business Objective Community Teams Name
Share of CMX respondents naming each as their community's single top objective, 2025
View the data
| Label | Value |
|---|---|
| Support for customers | 24% |
| Customer success | 22% |
| Acquisition and marketing | 20% |
| Retention and engagement | 19% |
| Product feedback | 6% |
| Content creation | 6% |
Support, customer success and retention together are the top objective for 65% of teams, against 20% whose community exists mainly to win new customers.3 Most communities are built for the people who already pay, and the evidence on what they return comes from those people too.
The clearest number comes from a community vendor measuring its own customers. In October 2023 Gainsight matched its community activity to customer accounts and compared the most engaged accounts with those barely or never engaged:
Engaged accounts showed gross retention 3 percentage points higher and expanded at 2.2 times the rate.4 Both gaps carry the usual caveat. Customers who are already doing well may simply join communities more, so the comparison shows the two moving together, not that the community caused it.
Revenue from new customers is measured far less often than retention, and CMX’s survey shows where the measuring stops:
- 72% of retention-focused teams measure customer retention directly.3
- 34% of acquisition-focused teams measure the revenue their community influences.3
- 24% of all teams can confidently put a money value on their community, up from 16% a year earlier.3
- 34% have a CRM that links community and customer data.3
Of the 59 teams that could put a value on it, 36% reported a million dollars or more.3 The teams with a linked CRM were more than twice as likely to call their community extremely successful, 24% against 11% where the CRM is not linked.3 Matching members to accounts and comparing them with similar accounts that stay away is how most teams can show finance what a community returns.
Support is the job teams name first, and Gartner’s December 2023 survey of 5,728 customers, across industries, shows how often self-service fails them:
- 14% of service issues are fully resolved in self-service.5
- 36% of the issues customers call very simple get resolved there.5
- 43% of failures came down to the customer not finding content relevant to the issue.5
A community can answer the questions a help center never covered, in the words customers search with. That only helps a customer who can find the answer, and where a community lives decides who can.
The Platforms Communities Use Most Sit Behind a Sign-In
A community earns its return with the customers inside it. A private one stays out of sight of a buyer who has not chosen a product yet, and of the AI engines that buyer asks for a shortlist. CMX asked where communities host their forums and message boards. The three answers named most often were chat apps:
Scroll horizontally to see all columns when the table is wider than this page.
| Platform | Share of communities using it, 2025 |
|---|---|
| 21% | |
| Slack | 20% |
| Discord | 16% |
| Custom platform built in-house | 10% |
| Facebook group | 10% |
| Khoros Communities | 8% |
| LinkedIn group | 7% |
| Salesforce | 7% |
| Circle | 6% |
| Discourse | 5% |
WhatsApp, Slack and Discord keep their conversations behind a sign-in, and a private Facebook or LinkedIn group does the same. A search crawler can’t read what sits behind a sign-in, so an AI answer has nothing from it to quote. WhatsApp rose from 8% of communities in 2023 to 21%, and Slack fell from 28% to 20%.3 Both are closed spaces, and so is Discord, the third most used.
A closed chat app is a fine home for a customer community, whose members are already signed in. It means a community on one of these platforms was never going to be read by a buyer researching the category. Community managers on Reddit say the same about search:
Tips from Reddit
What community managers say about search and where conversations live
For forums and similar platforms, SEO is crucial for driving organic growth. However, for communities hosted on platforms like Discord, SEO is less relevant because the content isn’t as easily discoverable through search engines.
In support-based communities, you usually don’t have to stress too much about SEO because the users and the content tend to handle that pretty well on their own.
The reality with community is that it is messy. Conversations happen everywhere. The best strategy is to embrace it, move with it, learn with it.
Which Communities the AI Answers Cite
If the customer community is out of the engines’ reach, the useful question is which communities they do read. Every week our Empact Panel asks ChatGPT, Google AI Mode and Claude what each partner’s buyers ask them, and stores every answer with the pages it cites. This count uses ChatGPT and AI Mode, because for Claude the Panel stores the pages its search returned rather than the pages its answers cite.
We took 13 partners in software categories and their sourced answers, the ones citing at least one page, collected from 30 October to 16 November 2026: 66,266 from AI Mode and 69,684 from ChatGPT. Every cited page went into one of four groups: Reddit, independent forums and Q&A sites, communities run by software vendors, and everything else.1
The two engines read communities very differently:
AI Answers Citing Each Kind of Community, by Engine
Median of 11 business software panels, share of answers citing it, Oct to Nov 2026
View the data
| Label | ChatGPT | Google AI Mode |
|---|---|---|
| 0.1% | 9.3% | |
| Vendor communities | 0.4% | 0.3% |
| Independent forums and Q&A | 0% | 0.1% |
- 0.3% of AI Mode answers cited a vendor community in the median business panel, from almost none to 3.2% across the 11.1
- 11 of 11 panels saw AI Mode cite Reddit more often than every vendor community combined, by a median of 16.7 times.1
- 0.1% of AI Mode answers cited an independent forum or Q&A site, such as Stack Overflow or Quora, in the median panel.1
- Under 1.5% of ChatGPT’s answers cited Reddit, and the same held for independent forums and for vendor communities, in all 13 panels. The vendor-community median was 0.4%.1
For AI Mode, the community that reaches a buyer’s shortlist is Reddit, and a vendor’s own forum hardly registers. ChatGPT barely cites communities of any kind, so for ChatGPT the work moves to other kinds of pages, such as the listicles AI engines trust.
The gap is not one category pulling the median. Across the 11 business panels, the share of AI Mode answers citing each kind of community ranged like this:
How Often AI Mode Cites Each Kind of Community
Lowest to highest of 11 business software panels, share of answers, Oct to Nov 2026
View the data
| Label | Range |
|---|---|
| 1.7% to 15.3% | |
| Vendor communities | 0% to 3.2% |
| Independent forums and Q&A | 0% to 0.3% |
Reddit’s lowest panel, at 1.7% of AI Mode answers, sits close to the highest panel for vendor communities, at 3.2%. In the two panels whose partners sell to developers, Reddit reached 16.3% and 32.1% of AI Mode answers, against 0.4% and 0.9% for vendor communities.1
When AI Mode did cite a vendor community, it was almost never the partner’s own. In 12 of the 13 panels, no AI Mode answer cited the partner’s own community, and the ones cited belonged to rivals and to other software vendors.1
Our numbers can’t settle whether a busier vendor forum would close the gap. The vendor communities AI Mode did cite were already public, and they still trailed Reddit in every panel. The cut also counts citations, not signups, over two weekly collections.
A public, well-kept forum of your own still isn’t the community AI Mode quotes to your buyers.
The Panel’s Reddit Lift card compares how often a partner is recommended when the cited threads mention it and when they don’t. On one software partner’s panel over the last 90 days, 172 answers cited Reddit threads, all but one of them in AI Mode:

Answers whose cited thread mentioned the partner recommended it about twice as often as answers whose threads left it out.1 The two move together on the card, which does not prove the thread caused the recommendation. It does show that the mention sat in a public thread the partner doesn’t host.
Buyers’ own answers point the same way on a different question, which sources they trust. In a survey Reddit commissioned from SurveyMonkey around the turn of 2026, 1,202 US business decision-makers said:
Weigh the Reddit figure with its sponsor in mind. Trust is a different measure from the Panel’s citations, and it leans the same way, with peers above vendors. 48% of these buyers also said they struggle to find real user testimonials at all.6
Put side by side, the two kinds of community reach different people at different moments:

Run one community well for the customers you already have, and put the visibility work into the public threads the answers already cite.
How to Divide the Work Between Your Community and the Forums
Buyers now find a product in AI answers, forums and review sites before they ever sign up, so both halves of the split matter to product-led growth. Our work with a partner opens with an audit and a roadmap. For a partner, the Empact Panel puts its buyers’ questions to ChatGPT, Google AI Mode and Claude each week, and stores the answers along with the threads and forums they cite.
Our advice is to keep the customer community to one space, run for customers and judged on renewals and support. In the public threads, our Reddit marketing team reads each community’s rules first. Then it writes replies that deal with what the person asked, and mentions your product only where the answer needs it. Every week you get a report on those threads and on which of them Google and the AI answers cite.
Each space then gets measures of its own, and a team can check every one of them:
What each space is judged on
- The customer community: renewals and expansion of member accounts, set against similar accounts that never joined
- The customer community: how many member questions get a good answer, and how quickly
- The public threads: whether ChatGPT and AI Mode cite them when your buyers ask
- The public threads: whether the cited threads name your product, and how often they come back week after week
Keeping the two measures apart also makes each easier to defend in a budget review:
Report the two on separate lines. The customer community earns its budget in renewals and in questions answered. The public threads earn theirs when an AI answer cites one that names you. Put both on one slide and the community gets blamed for visibility it was never built to win.
Gabby Saayman on LinkedInOpens in a new tabThe public half also runs through your own site, because a buyer who meets you in a thread lands there next. The PLG Readiness Score (PLG stands for product-led growth) reads up to fifteen of your public pages. It scores ten signs a buyer looks for before signing up alone, visible community activity and customer reviews among them:
In the threads themselves, the work starts with the ones the answers already cite. That step has a playbook of its own, on putting your product into the threads AI Mode and ChatGPT draw on when your buyers ask.
Set the two budgets separately, and review each against its own measures every quarter. If the cited threads that name you grow while the community keeps renewing, the split is working.
Frequently Asked Questions
How many B2B companies have a customer community?
Almost seven in ten, in Gainsight and Benchmarkit’s 2024 survey of more than 200 B2B companies, and 59% of the rest planned to start one. Read it with its sponsor in mind, since a company that sells community software ran the survey. If your customers need one, the useful question is how many spaces your team can keep answered.
How big is the team behind a typical community?
Small. In CMX’s 2025 survey of 589 community professionals, counting full-time people only, 31% of teams were one person, 34% were two to five people, and 17% had nobody full time, against 4% in 2021. Communities of 1,000 to 9,999 members most often ran on two to four moderators. Plan the number of spaces you run around those people, not around the number of channels your members already use.
What does it cost to run a SaaS community?
Mostly people’s time, and few teams can count on more of it. Of CMX respondents who see their community budget, 26% saw it grow in 2025, 18% saw it cut and a quarter could not say. The challenge named most often was proving the return (37%), ahead of the cost of tools (20%). Before adding a platform or a second space, price the hours of the person who will answer in it every week.
Do members of a customer community renew more often?
In Gainsight’s own data, yes, with a caveat. In its analysis of its own customers, the most engaged accounts showed gross retention 3 percentage points higher and expanded at 2.2 times the rate of accounts with little or no engagement. Committed customers may also join communities more, so compare members with similar accounts that never joined before you credit the community with the difference.
Can a customer community generate pipeline?
Some can, but few teams measure it. In CMX’s survey, 20% of teams named winning new customers as their community’s top objective, and only 34% of those measured the revenue it influenced. If pipeline is the goal, connect community activity to your CRM first. Teams whose CRM links the two were more than twice as likely to rate their community extremely successful.
Does ChatGPT cite vendor community forums?
Rarely, and it rarely cites Reddit either. Across our 13 software panels, neither Reddit, independent forums nor vendor communities reached 1.5% of ChatGPT’s sourced answers in any panel. Its answers cite other kinds of pages, so in our panels a forum on your own domain rarely shows up in them at all. If ChatGPT leaves you out, start with the pages it already cites for your category.
Does Google AI Mode cite our community or Reddit?
Reddit, by a wide margin. In the median business software panel, 9.3% of AI Mode answers that cited any page cited a Reddit thread, and 0.3% cited a community run by any software vendor. In 12 of our 13 panels, no AI Mode answer cited the partner’s own community at all.
Which communities besides Reddit do AI answers cite?
Independent forums and Q&A sites such as Stack Overflow, Quora and Hacker News, but far less often. In the median business software panel they appeared in 0.1% of AI Mode answers. For tools bought by developers they appeared a little more, in about 0.4% of AI Mode answers in each of our two developer panels, and Reddit still led there at 16.3% and 32.1%.
Should we make our private Slack or Discord community public so AI engines can read it?
Usually not. Members joined a private space and talk differently there, and in our panels even the vendor communities that were public trailed Reddit in every case. Keep the private community for customers. Publish the answers buyers ask about in places anyone can read, such as your docs and the public threads where buyers already compare tools.
Is one community better than several?
For most SaaS teams, yes. A community lives on member questions that get answered. Unless question volume is low, a team of one to five full-time people struggles to answer quickly in a Slack workspace, a forum and a Discord server at once. Run one space well, set a response time your team can keep, and send members to it from the other places they ask.
How much of a branded community should be about our product?
As a rule of thumb, keep about four posts in five useful to members with no product angle, such as answers to their questions and how-tos. A community that reads like a sales channel loses the members whose answers made it worth joining.
How do we know whether the public threads are working for us?
Check two things for your buyers’ questions every week. First, whether ChatGPT and AI Mode cite the threads at all. Second, whether the cited threads name your product. On one partner’s panel, answers whose cited thread mentioned the partner recommended it 58.6% of the time, against 27.9% when the threads left it out.
Sources (6)
- Empact Partners, "Empact Panel, ChatGPT and Google AI Mode answers to the unbranded buyer questions of 13 partner panels in software categories, each panel’s latest two weekly collections after 9 October 2026" (2026)Empact Partners' own data↩1↩2↩3↩4↩5↩6↩7↩8↩9↩10↩11↩12↩13collected 30 October to 16 November 2026, every cited page sorted into Reddit, independent forums and Q&A sites, communities vendors run, and everything else. Each panel counted once. The Reddit Lift card is one software partner’s panel, last 90 days, captured 16 November 2026. Read 16 November 2026.
- Gainsight and Benchmarkit, "The State of Community in Customer Success 2024" (2026)↩1↩2of more than 200 B2B companies, results page. Read 16 November 2026.
- CMX, "The 2025 CMX Community Industry Trends Report" (2025)↩1↩2↩3↩4↩5↩6↩7↩8↩9↩10↩11↩12↩13↩14↩15↩16↩17↩18↩19of 589 community professionals, 1 May to 22 June 2025. Read 16 November 2026.
- Gainsight, "Community Engagement Drives Post-Sales Results" (2023)↩1↩2↩3Gainsight Community, 15 November 2023, Gainsight’s analysis of its own customer accounts, October 2023. Read 16 November 2026.
- Gartner, "Survey Finds Only 14% of Customer Service Issues Are Fully Resolved in Self-Service" (2024)↩1↩2↩319 August 2024, survey of 5,728 customers, December 2023. Read 16 November 2026.
- SurveyMonkey and Reddit, "The hidden B2B journey" (2026)↩1↩210 March 2026, survey of 1,202 US business decision-makers, 23 December 2025 to 7 January 2026. Read 16 November 2026.
Keep reading
Statistics 48 AI Search Citation Statistics
See which pages AI answers cite, why your own site is only part of the picture, and where to focus your SaaS visibility budget.
Statistics 45 SaaS SEO Statistics
45 figures on what organic search is worth to a SaaS company, and where AI answers about software get their sources.
Statistics 37 Subreddit Self-Promotion Rule Statistics
Promotion, reply, karma and AI-text rules in 111 subreddits that AI answers cite, and what they let a SaaS team do.


