FeverBee – Community Consultants

Headless Communities Are Here: What Does That Mean?

Richard Millington — Mon, 02 Nov 2026 09:00:00 +0000

While conducting UX research for a client’s community this month, three members said exactly the same thing.

They said they no longer visit the community.

Instead, they use ChatGPT or Claude to ask questions on their behalf, retrieve information, and find out what’s new.

Now, it would be a mistake to over-index on individual anecdotal examples.

But when combined with other signals, it portends a huge change in how people engage with communities. A small, but rapidly growing, percentage of people seem to be engaging with and benefiting from a community without ever visiting it.

It’s Not Just AI Overviews Anymore

We know that engagement in many hosted communities has already fallen between 20% and 40% year on year since Google’s March 2025 core update.

Until recently, this mostly happened because Google’s AI overviews pulled answers from communities into search results, so people got what they needed without clicking through. Now it’s a bigger behavioral change.

And this isn’t just anecdotal.

With the rollout of personal assistants like Instinct, Muse, and Claude Duet, we’re entering an era where more people will engage with a community without visiting it.

Unlike a chatbot, these assistants act on your behalf. They work in the background and message you with what they find, which is exactly what those three members described.

Welcome to the era of headless communities.

What Does A Headless Community Mean?

The idea of a headless community comes from software.

A headless CMS or commerce platform splits the backend, where the content and data live, from the frontend, where people visit, click, and browse.

This means the content and data stay in the same place, but people access them through many channels.

This is far from a new concept.

Many vendors have implemented some form of this already. We’re seeing in-product communities connect to existing enterprise communities, for example.

Some folks have already multiple custom-developed a number of different front doors for a community: perhaps one in-product, perhaps one on the website, perhaps one in an app, and other channels as well.

The difference now is organisations don’t have to jerry-rig a setup to make this work.

Platform vendors are providing it and encouraging organisations to do it. This likely means a proliferation of user interfaces all calling on the same data and content.

MCPs have already made this possible on the back end; now it’s happening on the front end.

Introducing Member-Facing MCPs

And speaking of MCPs, member-facing MCPs are starting to appear.

Mighty Networks already lets members connect their own AI assistant to a network, and more vendors are likely to follow. This essentially encourages members to engage with a community without visiting it.

They can simply submit their request through Claude, ChatGPT, Copilot, or whichever LLM they prefer.

In short, the familiar hosted community experience we’ve relied on for the last 20 years will likely fade into the background.

In its place will rise a wide range of different front-end experiences, and this has huge implications for how we think about designing and measuring a community.

Headless Communities Create New Challenges To Solve

What happens when no two members will experience the community the same way?

Well, this has some obvious implications. Let’s go through a few.

1) Gamification

If fewer people see each other’s gamification badges, points, levels, and leaderboard rankings, gamification probably becomes less valuable to the people who care about it.

While some people undoubtedly like collecting badges and points for their own sake, I suspect many of the highest contributors like the status it affords them within a community. But they will only have that status if other people visit that community.

So we need to find a new way of providing people with that status, and that will probably come from a centralised page that pulls in contributions from both on and off the community. But this will require a completely different point system and approach.

2) News And Updates

Many organisations rely on a community to disseminate news and updates, but if fewer people visit that community, members won’t see them. Instead, you will need to distribute them by email as well, and in some cases, you already do.

Or you have to create prompts that people can use to get the information they need from a community and encourage them to use it. This is the same challenge as persuading people to download and use an app: not impossible, but difficult. Many phones are littered with unused apps where good intentions met the brutal reality of prioritisation.

So it’ll be less effective to post news and announcements on the community in the future.

And this applies to anything related to discovery.

If you’re hoping to post something in a community for people to discover unintentionally while they’re browsing for something else, that’s likely to be less effective in the future than in the past. This might affect things like user groups, upcoming events, and any time-limited or time-sensitive things.

3) User Groups

I’ve long been suspicious of user groups being hosted on a single platform versus a decentralised approach where leaders can host groups on any platform that they like and that they know the audience will prefer.

Too many community platforms already host too many subgroups on life support with limited or no activity.

With fewer people visiting the community, it makes no sense to host most groups there. Instead, leaders should be encouraged to build groups wherever it makes sense. There should be a central place, similar to what Notion does, to share these groups with those interested.

4) Discussions

Two interviewees at the beginning of this article mentioned the same thing.

When they have a problem, they ask their AI tool (Claude/ChatGPT etc…) for an answer.

The agent scans all available sources to find the answer. If it can’t find a satisfactory answer, it asks a question. But it doesn’t just ask in a hosted enterprise community.

It asks a question in every channel available to it.

That might include filing a ticket. That might include the hosted community. That might include other communities on the topic. It might include Reddit, etc.

This means the same question is appearing in multiple places at once. Then it schedules a task to check the responses every hour or so and retrieve an answer.

This behaviour isn’t completely new. We’ve previously found folks tend to publish the question in as many places as possible and see what responses they get.

But now AI tools are going to enable people to automate this. But if the same question is appearing in multiple places, what is the unique value of a community? It might be filled with the same questions and responses as every other channel.

5) Superusers

It’s easy to imagine that, for most support communities, the future will be agents asking questions and other agents answering them.

This will probably start by clarifying what kind of question it is and whether the answer lives in documentation and knowledge. In which case, the organisation-created agent will simply retrieve that information and post it as an answer.

But if not, this would notify or tag the human experts (superusers or internal) to provide the answer. Internal staff would probably wait 24 to 48 hours and then pick up the questions members couldn’t answer.

But this raises an interesting question.

The major benefit of answering the question is knowing you’re helping someone. Do you get that same good feeling if you know an agent posted the question and the person didn’t make the effort to post a question themselves?

The incentives required for superusers will likely have to increase significantly to compensate for the decline in gamification and the reduced sense of helping individuals.

I suspect because of agents, we’ll have fewer superusers, but more incentivised ones.

6) Belonging

Also, what happens to belonging?

Will people still feel a sense of belonging when a growing percentage of activity is driven by bots?

This is dead internet theory playing out inside a community. Agents talking to agents in a community sounds almost Black Mirror-level apocalyptic.

I suspect if we want belonging, it won’t happen in an open, public Q&A experience like it did in the past. Instead, if we want belonging, we’ll have to work much harder to cultivate it by hosting private events and groups, and do much better facilitation of those activities. It will happen, but it probably won’t happen in the same places agents engage.

7) Measurement

One obvious problem with this, one that we’re running into already, is: how would you measure this?

This is the final nail in the coffin for measuring engagement, because there is less and less engagement to measure. Do we measure questions resolved, total answers across any surface, or the extent to which AI support leans on community? Or do we find other metrics of community retention instead?

There’s literally no point in measuring anything related to traffic if most of that traffic is bots.

I’ve written before about measuring ghost deflections and why call deflection has had its day.

Any metric built upon visits is now redundant. We need to find new means of measurement.

8) Platform Selection And Optimisation

This also significantly affects platform selection.

It’s well known that a community homepage is more important internally (showing off the community to other staff) than externally (to members) because the vast majority of people just visit a discussion beyond the homepage, get the answer they want, and leave.

But now the interface features of a community matter even less.

But backend features, like API depth, the data model, MCP support, identity, data portability, analytics, and whether the platform can expose clean, machine-readable knowledge, become far more important. Discourse, for instance, now ships native llms.txt support.

This Is All Guesswork

A lot of this is informed guesswork right now, but I think what’s clear is something big is happening, and it’s going to have a huge impact, especially for everyone that has a hosted community experience.

If we’re designing strategies for the future, we have to think about all of these kinds of questions, especially when we select the kind of platform that we want: the activities we run and what our community is going to be in the future compared to the past, because this will change a lot of things. I don’t think many of us are ready for that yet.

Let’s Set Your Community Up For The Future

I believe most organisations still make decisions based on a community paradigm that no longer exists.

This affects the budget requests they’re making, the platforms that they select, the programmes they’re choosing to run, and where their community team spends most of their time.

And that makes sense because a lot of community teams have had to do more work with less staff over the past few years. This makes it harder to look up from the day-to-day, see where things are heading, and adapt the strategy accordingly.

But the agentic era is coming. Things are changing quickly. Metrics are going haywire. Traffic is going down. Internal priorities are shifting faster than ever before.

If you’re unsure what to do, let FeverBee work through it with you.

We can audit your current community, tell you how prepared you are for current trends, and then work with you to design a strategy and plan that will guide your community into the new era: one where it’s still delivering more value than before, gains more internal support, and becomes an increasingly critical pillar of the customer experience.

If you’re interested, you can click here to contact us.

The post Headless Communities Are Here: What Does That Mean? first appeared on FeverBee - Community Consultants.

Measuring call deflection has held communities back for years. Let’s use this instead.

Richard Millington — Mon, 26 Oct 2026 12:15:43 +0000

For a long time, public relations was measured by one of two things.

Either the number of press clippings you could achieve for clients or what’s called advertising value equivalent.

Advertising value equivalent is essentially a metric where you multiply the advertising rate in each publication by the column inches you achieve in that publication. Then you combine the total amount to say how much PR is worth.

And while AVE offered a precise, scientific-sounding measure that PR firms loved to share with clients, it increasingly came to be seen as problematic.

In 2010, many of the world’s top communication professionals gathered in Barcelona for the International Association for the Measurement and Evaluation of Communication (AMEC) summit, and they agreed on a set of principles known as the Barcelona Principles.

One principle stated that AVE is not the value of public relations.

But they didn’t select a preferred option. They left the outcome to PR professionals. At the time, many worried that not being able to share a precise dollar-value return would kill the profession. But over the following years, as AVE gradually died as a measurement tool, PR firms’ popularity and success only grew.

In short, bad measurements held PR back for years, and only when they got rid of them could better techniques and processes take their place.

Enterprise community folks are now in a similar place. For a long time, we’ve used call deflection metrics. Call deflection is essentially the number of visitors who viewed answered questions, multiplied by the percentage who say they got the answer they wanted, multiplied by the average cost of a support ticket.

Call deflection value = visitors to answered questions × % who say they got the answer they wanted × average cost of a support ticket

And this ends up giving a very precise value. And typically it gives a very high value. That’s because the measure essentially gives the highest number you can possibly give.

And because the numbers are often so high, they’re often unbelievable.

I remember speaking to one exec who complained that the value of community deflections is three times higher than his entire customer support budget.

There are several reasons for this, and it’s important to understand each one.

1) Communities Attract Non-Critical Questions. When you create a community, you lower the barrier to asking a question. This means you attract a lot of questions people otherwise wouldn’t have made the effort to ask a customer rep. They aren’t ticket deflections. Their value would show up as improved customer satisfaction, product adoption, or increased sentiment and usage.

2) It’s Hard To Prove People Achieved A Resolution. The only real way to prove community resolved an individual question is to ask people if it was resolved. So organizations usually use a poll asking people, “Did it solve your question?” Sometimes, as follow-ups: “Would you have otherwise contacted support?” and “Are you in warranty or not?”

The problem is that polls are heavily influenced by response bias, and we generalize the results across all traffic.

Even if people were only on the community for a couple of seconds and there’s no way they could have received the answer, we still count that. We’ve also found there’s a curvilinear relationship between how long it takes the poll to appear and the percentage response rates.

By delaying the poll by a few extra seconds, you can double the response rate and, theoretically, double your community’s value, which undermines the entire point.

3) People Might File A Ticket Anyway. When we’ve studied how people resolve problems, we’ve found they don’t just ask a question in the community and stop. They might ask the same question in multiple communities and file a support ticket or two because dissipating the frustration matters more than the effort required to resolve the problem.

That means the community often gets credit for people who also filed a ticket.

4) Resolving A Problem Is Messy. Sometimes snippets and fragments of multiple answers are combined together to give a response. Sometimes problems have no resolution. An answer can mean people can stop looking, but it doesn’t necessarily mean there’s a resolution.

While deflection metrics are still widely used, they’re incredibly problematic and have never come remotely close to calculating the true value of a community (or, most likely, actual support costs saved).

The Final Nail In The Call Deflection Coffin

Since the rise of LLMs, the introduction of Google search snippets, and, most importantly, AI overviews, engagement rates across most enterprise communities have dropped significantly.

Many have seen declines in engagement of 20% to 40% year on year, and engagement keeps falling.

This doesn’t mean the community is resolving fewer questions.

It simply means people don’t need to visit the community for answers. It comes up naturally in search, so traditional call deflection methods will show a sharp decline in a community’s value.

At the same time, its real value is likely holding steady or even rising as knowledge spreads through more channels than ever, and this problem will only get worse.

Now we have increasingly complex and precise ways to address this. You can measure ghost deflections, as we spoke about last year, something that Microsoft has done and talked about a lot.

The problem is this is complex and difficult to prove, and sudden changes in the algorithm or Search Console calculations can wreak havoc with your stats.

And this isn’t going to change. It’s only going to get worse.

Usage of LLMs is going up. Fewer people will visit a community in the future than in the past.

People will increasingly retrieve community knowledge through other channels, so the old way of measuring deflection (any calculation that includes website visitors) is dead.

Which raises the question: what should replace it?

The Misleading Quest For A Precise Value

We try to measure the value of a community to sustain it or gain more internal support. CFOs, directors of finance, and other departments use it to decide how much of their budget to allocate to a community against the competing needs of other functions and departments.

But there’s also a misunderstanding in how these decisions are made.

These decisions aren’t based on which function delivers the highest return. They’re based on certainty, the time to achieve that return, and whether people trust the number in the first place.

The problem with call deflection for so long is that the methodology has been sketchy, the numbers are too high, and too few people actually trust that number.

What we need isn’t a higher number.

What we need is a number people can trust, and this number doesn’t need to be a dollar-value return. Most functions that an organization undertakes don’t deliver a dollar-value return. They are a couple of layers removed from that.

What we need to show is the impact on the organization’s priorities, so you don’t need to show a high dollar-value return if you can show a big impact in metrics that matter.

So drop the need to show a dollar-value return and instead focus on impact on customer satisfaction, self-service scores, resolution rates, and whatever metrics matter to the organization.

The Phoenix Measure Rising From The Deflection Ashes

For years we’ve been asking the wrong question.

We’ve been asking: how many calls or tickets does our community prevent (and what’s the value of those calls or tickets)?

The better question is:

What percentage of customer questions can now be answered because of the community?

For example, if the self-serve rate of customer questions is 55% without community knowledge (i.e., official resources/documentation) and 77% with community, that’s a fantastic outcome.

The best way of measuring this is to take a random sample of 100 questions and then answer two questions:

What % of them can be answered without community knowledge?
What % of them can be answered with community knowledge?

This gives you your additional coverage rate.

How Do We Measure This?

The ideal scenario is to pull 100 queries from your support tickets / AI logs. Then divide people into two groups at random and ask one group to solve the problem using only official documentation and another to solve it using both community and documentation.

This would give you the most reliable data, but it takes time and resources most teams don’t have.

A simpler approach is to get AI to help.

Unable to use any client’s actual support logs, I asked Claude to take 150 randomly selected Jira Cloud and Jira Service Management Cloud questions from the public Atlassian Community and then answer them in two ways. Once using official documentation and community content, and once using official documentation only.

Part of the problem is that Claude has been trained on community data and official documentation. It can’t forget it even if you ask it to. So I asked Claude to cite every source and the exact sentence it used in its response, then verify that none of the answers were unsourced.

The response was illuminating.

These findings are really interesting for a few reasons.

The resolution rate with the community is 9 percentage points higher than without. I suspect that understates the true value, but it’s worth noting.
The community either provides a better answer or the only working answer in 33% of questions. This shows the community’s unique value in an AI world.
Community adds value to questions where docs have an answer. Community often adds additional context, confirms which answers work, helps people know what to search for, etc.

When we actually drill deeper into where community added this unique value, the answers were truly illuminating:

A quick aside: I ran the same study on the Smartsheet community, and the results were surprisingly consistent.

The community gain was even larger: 84% of questions could be fully answered with official documentation alone, rising to 98.7% with community content added (a 14.7 percentage point gain).

So what does all this mean?

It means the main value of community isn’t just resolving questions that otherwise couldn’t be resolved. It’s improving the knowledge that does exist by adding context and experiences of customers who use these products every day.

You can see some examples of this below:

Notice how frequent and how valuable many of these contributions actually are!

And yet if you’re just measuring deflection, none of this will show up. That’s the problem with deflection today.

What Does This Experiment Really Show About Community Value?

The critical finding of this experiment isn’t that the community can answer questions that documentation can’t (though there are plenty of examples of that, and a clear percentage increase in resolution rate from connecting community to AI bots and including community in RAG search, etc.).

Community adds details and context that official documentation often misses.

It shows what happens when people follow instructions but still get stuck.

It highlights workarounds to problems when things don’t work as expected.

It can identify and document issues and limitations, helping customers avoid repeating mistakes others have already made.

The problem, though, is: where will this value appear? What metrics will you see change as a result of this unique community value? I suspect it’s here:

Answer quality: More customer questions can be answered accurately and completely because answers can draw on community knowledge, not just documentation. This might show up as more accurate AI answers, better answer ratings, and fewer follow-up questions.
Customer outcomes: If customers get better answers from the community, they will complete more tasks successfully. They should resolve problems faster and spend less time searching for help or troubleshooting issues.
Reduced costs: Customers will be able to self-serve better, so demand for support should fall. If agents, both AI and human, can solve problems faster, resolution costs should decline. This also means that customer satisfaction and product adoption will probably improve.

These are the kinds of outcomes we definitely should be measuring. We should stop measuring how many people who visited a community might otherwise have contacted support. That’s a speculative financial metric that misunderstands a community’s real purpose and value.

Instead, a much better test is what happens when we make community knowledge available versus when it isn’t.

What does the additional coverage rate look like then?

What does problem resolution or task completion look like then?

What is the time and effort customers have to expend to resolve their issue?

And how happy are they with those resolutions?

Does AI provide better answers? Do customers complete more tasks? Are problems resolved faster? Are fewer issues escalated and reopened?

I believe the future of many enterprise communities, especially support-centric ones, isn’t about proving how many tickets we prevented. It’s about proving how much better our customers perform because of the knowledge our communities create.

Let Us Do Your Audit

The problem is that many organizations don’t have the time or ability to run experiments like this to prove the unique value their community adds beyond their documentation. If you want help, contact us and let FeverBee undertake the activity for you.

We can run the entire experiment, show the results, and help you communicate the results internally. We can also help you design systems where the organization can best capture the knowledge and deploy the knowledge generated by the community.

If you’re interested, you can click here to contact us. We would love to help you not only prove the value of your community but also ensure your organization is getting the most value from your community.

The post Measuring call deflection has held communities back for years. Let’s use this instead. first appeared on FeverBee - Community Consultants.

We Need To Do Much Better At Deploying Customer Expertise

Richard Millington — Sun, 18 Oct 2026 20:07:49 +0000

The constant challenge of building enterprise communities is that much of the value lies in places few people check.

Often, that value goes unrealized because the organization doesn’t know where to look for it, or has no system to capture it.

Let’s use retention as a good example.

How Do Communities Really Drive Retention?

If you believe the 2026 CMX Community Industry Report, retention is the primary goal for organizations building enterprise communities right now.

But if you ask people why that retention happens, you often get a fuzzy answer.

This is often something along the lines of:

“When people are in a community, talking to others about the brand, they feel better about the brand and become more loyal.”

That might work for Harley-Davidson and a tiny handful of consumer brands, but fuzzy feelings don’t cut it outside of that.

The real mechanism is that people have a better customer experience because of the community.

As a result, they want to remain customers.

This comes through the knowledge and expertise shared in the community. For example:

They can resolve their frustrations faster.
They can configure and set up their products/services the right way.
They discover novel use cases and examples which help them get more from the product.
They understand where the product and roadmap are going and want to stick around.

All these things do impact attitudes (and are worth measuring), but fundamentally it’s about using customer knowledge to improve the ability of other customers to use the product.

The more you generate, capture, and deploy that community knowledge, the higher the retention.

How Do We Successfully Deploy Customer Expertise?

The problem is that many communities generate tens of thousands, even millions, of answers, and other members have to search for the right one and find it.

If they search for the right thing, they find it. If they don’t, they don’t. And while LLMs and AI are reconstituting this knowledge in interesting and exciting ways, they aren’t really designed to capture and harness the full value of customer knowledge.

That job falls to us, and most of us could be doing it far better than we are now.

We can divide this into five levels.

Level One: The Standalone Q&A

At level one, you have a community, but it’s not really connected to anything else.

People come, ask a question, get an answer, and leave. That can be pretty good for call deflection. We’ve done studies in the past that have shown that communities have a huge impact on call deflection.

And, generally speaking, the more people that see the answer, the higher that deflection rate is. This isn’t bad. It’s where many organizations are at right now with their community. They try to optimize the quantity and quality of answers in the community.

But if that’s the extent of what you do with that answer, then you’re capturing only a fraction of the value the community could and should provide.

Level Two: Integrate It With Support Tickets

The first step up is connecting your community to a system like Zendesk, where unanswered questions become tickets.

And just like any other ticket, they will (eventually) get answered by a customer support rep.

The key is that the answer is then pushed back to the community.

This means not only will the person who asked the question receive the answer, but hundreds, even thousands, of others can see it (and search and bots can see it, too).

In my experience, this also encourages support staff to search the community for answers to support tickets and customer calls, especially for the more difficult, contextual, edge-case problems.

This makes their answers far more valuable than replies to individual customer queries.

It helps to design the community as part of the support hub from the start.

Level Three: Community Knowledge Powers Your AI Support

This should be the minimum for any enterprise community today. Community knowledge should be retrievable by your RAG systems and AI support bots. This is when customer expertise becomes far more valuable.

This means that when someone asks a question on any support portal or AI assistant, the customer’s answer can appear alongside other relevant information via RAG (retrieval-augmented generation).

Community knowledge should be ingested into your support systems.

Yes, there are arguments about the reliability of customer answers, but that’s like arguing about the reliability of Wikipedia articles.

Will there occasionally be bad, poor information? Probably.

Will the useful, accurate information far outweigh the bad? In almost every case, yes.

The challenge here is that integration isn’t easy. Discussions aren’t designed to be retrieved like other articles in a database. To ingest community content, you either have to crawl it like a webpage, export and reshape it into something that resembles a help center article, or reduce it to a question-and-answer format. All of which strips out useful metadata. But this is still far better than nothing.

This should be an immediate project for the vast majority of enterprise community professionals out there.

We’ve long argued that keeping community as a separate, standalone department limits its impact. That argument now applies to community expertise. If that data isn’t combined with other systems, it delivers only a fraction of its value.

Level Four: Feed Community Content Into Courses And Product Content

A constant problem in many enterprise communities is repetitive questions.

I’ve never measured it, but I wouldn’t be surprised to see the 80:20 rule popping up here, where 80% of questions concern 20% of topics.

There’s a lot of knowledge being shared on a comparatively small range of topics. The obvious question is what happens if we pre-empt questions by giving members the knowledge up front?

There’s plenty of discussion about combining community with education, but largely it’s about integrating the platforms, reducing friction, and letting people display course qualifications on their profiles.

But the real win is feeding community content into your courses. When the same topic and answers show up in your community, that should feed the courses, onboarding material, and product documentation itself.

This turns community knowledge that disappears into a void into a growing body of content that improves satisfaction and retention across the board.

Veterans might remember when ‘tribal knowledge bases’ were a thing (essentially turning community content into wiki articles). The problem with TKBs is that motivation to contribute was low, and the content wasn’t plentiful enough to make them broadly useful. Feeding community knowledge into existing and new lessons doesn’t have that problem. Members are already creating the content; you just need to put it to work.

One of the fantastic things about AI is that you can automate this. You can have agents that review discussions in your community, look at popular answers, and review courses and product documentation to see where the knowledge can and should be updated. It obviously helps here if they are on the same platform, but it’s not a decisive factor.

Level Five: Account Signals To CSMs

Perhaps the most powerful way community drives retention is through the signals and information it provides about customers.

Unfortunately, in most cases, this knowledge goes to waste.

In a community, you will often see customers asking questions in a certain category, appearing frustrated, or not getting answers.

Wouldn’t this be incredibly useful for the relevant customer success managers to know? (And for the organization as a whole to know?)

If you’re responsible for a customer, and you get a notification saying that customer has had two unanswered questions in the community in the past two weeks, you can set up a call with that customer and check what help or support they might need.

This requires the community to be integrated with Salesforce, Gainsight, HubSpot, or another CRM. You need that data to show up.

The funny thing about level five is that it probably has the most direct impact on retention (making contact with unsatisfied customers before they decide to leave), but it’s also incredibly hard to measure and wouldn’t show up on most dashboards.

The problem here is matching. Members often register with personal email addresses. This means a significant percentage of community activity might not link to anyone in the CRM. So before you build the integration, check what percentage of active members you can actually match to an account.

The Systems We Build Define The Results We Get

Much of our work in recent years has shifted away from programs to incrementally improve the community and towards systems that get far more value from community knowledge.

Even today, too many community teams focus on individual programs, still often using engagement metrics to track success. The real win, especially for the larger, more mature community, is building connected systems that ensure community knowledge flows back through the organization.

Customer retention doesn’t increase because people experience warm, fuzzy feelings from talking to one another. It increases when people benefit from other customers’ expertise. That means going beyond Q&A and integrating community knowledge deeply into the broader customer journey.

Many organizations still lack processes to make that happen. This information, and these incredibly powerful signals, simply get lost.

If you want to make this happen, an easy step to take is to let us review the existing shape of customer knowledge. We can help you understand what expertise exists, where it is being deployed at the moment, how valuable it is, and what’s holding deployment back.

Based on that, we can put together a clear plan, technology requirements, and the processes to bring that expertise to the forefront of your customer experience.

If you would like guidance, inquire about our support.

The post We Need To Do Much Better At Deploying Customer Expertise first appeared on FeverBee - Community Consultants.

How Do The Top 13 Community Platforms Compare?

Richard Millington — Thu, 24 Sep 2026 06:39:04 +0000

TL;DR – Our analysis of the top 13 community platforms is now available here.

What’s the best community platform in 2026?

Back in 2024, we published our most detailed breakdown of community platforms at the time.

Looking back, so much has changed. Khoros imploded, new platforms arose, and AI is becoming increasingly central to the community management experience.

Today, after six months of platform reviews, sandbox verifications, and speaking to dozens of customers across almost every platform, we can finally publish our Buzz Report 2026.

Download it here

It covers 13 platforms (vs. 6 previously) and broadens the evaluation criteria.

We evaluate each platform against nine objective platform capabilities (Discussions & Q&A, Ideation, Content and Knowledge, Events, Gamification, Moderation, AI, Mobile, and Analytics/Reporting), and two subjective criteria (Customer Experience and Vendor Trajectory).

In my opinion, the result is the most comprehensive community platform guide available today.

If you want to check that you’re on the right platform, how different platforms compare, and which platform best suits your needs, download this report!

A huge thanks to Gainsight for once again sponsoring the Buzz Report 2026 and making it available to all for free.

Note on sponsorship: Gainsight was invited to sponsor only after the scoring was complete and had no editorial input.

Note on AI usage: AI was used extensively for the design and some points of technical verification (notably on AI features/MCPs, etc.), not for the main editorial.

The post How Do The Top 13 Community Platforms Compare? first appeared on FeverBee - Community Consultants.

Should You Build Your Community On Gradual?

Richard Millington — Mon, 31 Aug 2026 11:00:00 +0000

Gradual is the newest platform in this series, and you can feel it. Nearly every other tool I review grew out of a forum. They started as a place to post threads, then added events, content, and everything else over the years, and an old discussion engine still sits at the centre of them. Gradual doesn’t carry that. It was built recently, around events and content first, and the community grew up around those.

That single fact explains most of the scorecard. I scored it on the same nine capabilities I use for every platform in this series, so you can put it side by side with Higher Logic Thrive, Gainsight CC, and Higher Logic Vanilla. It came out as a specialist. One eight, a band of sevens, two fives, and one real gap at a three. It is ahead of the pack on the things communities have moved towards, and behind on a couple of the fundamentals.

The other thing to get straight early is what Gradual is actually closest to. It isn’t Discourse, and it isn’t Higher Logic Vanilla or Gainsight. It’s Circle.

The Newest Platform In The Series

Gradual is a young company, and for a platform you commit to for years, the vendor matters as much as the features. It started small, powering the occasional event, roughly one a month, for a handful of big early accounts. OpenAI and Perplexity were among the first communities on it, and it has widened out from there to Scale, HeyGen, GE HealthCare, Cognition, and Zwift, among others. A lot of those are the sharp, AI-native, technically demanding companies that tend to be fussy about what they put their name on.

The order is the whole point. Most platforms in the series began as a forum and added events later. Gradual began around the events and the content, and grew the community around that, which is why those parts feel the most finished. It’s a small team with a tight focus, building one well-designed platform rather than a sprawling suite. It’s close to the opposite of Higher Logic Thrive in that regard. It’s also SOC 2 certified, so the enterprise basics are in place.

The downside of being new is that some things aren’t built yet. The analytics are thin, the AI is early, and the ideation barely exists. You get a platform that is ahead on the modern things and behind on a couple of the basics.

The Closest Comparison Is Circle

Think of Gradual as a more customisable, more capable upgrade on Circle. It’s the same all-in-one shape, community and courses and events and content in one good-looking place, but it goes further on most of what I score here and reaches into the enterprise level.

On the nine-capability card, Gradual matches or beats Circle on eight of the nine, and trails it only on mobile by a point. The widest gaps run in Gradual’s favour on events and gamification, and on real questions and answers, where Circle handles RSVPs but not ticketing, has no gamification to speak of, and gives you a feed rather than a proper Q&A. Circle has strengths this card doesn’t score, the paid-membership and creator side especially, and it’s the simpler product. But if you’re weighing Circle, weigh Gradual next to it. Gradual is the more capable tool, and that comes with a higher price tag.

What Gradual Does Well

One capability scores an eight, and three more score a seven: events, content, gamification, and moderation.

Events. This is the highest score in the whole series. Events in Gradual are the core of the product, not a module bolted onto a forum. When you create one you pick from six native formats, from a meeting or a livestream to a webinar, a hybrid event, an in-person event, or a one-to-one match for mentoring. Open a single event and there’s a full builder underneath: days, tracks, sessions, speakers, sponsors, an attendee manager, and a ticketing setup with ticket types, promo codes, and a separate virtual ticket. It sends the registration and reminder emails, records the sessions and publishes the recordings back into the community’s content, and shows you live attendance while the event runs. The limit is that this is events built for a community, not a standalone events platform at the scale of a dedicated tool like Bevy. For most community teams, that’s the right amount.
Content and knowledge. This scores a seven, and again it’s native. The content hub covers blogs, videos, podcasts, resources, external content, and collections, and the presentation is good. But the score comes from the courses. Gradual has a real learning platform built inside the community, with video lessons, per-lesson analytics, completion tracking, ratings, an instructor, and a certificate at the end, and you can chain courses, events, and community activity into learning paths that flow through to Credly. Every other platform in this series lost a point on content because none of them did courses natively. Gradual has it built in. What keeps it a seven is that there’s no structured knowledge base in the support sense, so if the job is deflection, it’s the wrong shape.
Gamification. This scores a seven, which is stronger than I expected. Points, badges, and a leaderboard sit in one place, the badges are designed rather than generic, and there’s room for fifty of them plus your own, each triggered by real activity across events, courses, and the community. The part that stands out is the rewards store. Members can exchange points for physical items, actual products with inventory tracking, a shipping policy, and an order screen to fulfil them, an Influitive-style setup where a hoodie arrives in the post. Where it stops short of the top is the gap Vanilla and Discourse fill: it rewards people, but it doesn’t use that reputation to run the community for you with trust levels that auto-promote reliable members.
Moderation. This scores a seven, better than you’d expect from a platform this focused on events. You get AI content classifiers for discrimination, sexual content, violence, and personally identifiable information, each with a lenient, standard, or strict setting, plus a profanity filter, a blocked-URL list, and a progressive ban ladder that escalates on its own. There’s a manual review layer behind the automatic one, and event channels get their own moderation. What holds it back is the same trust-level gap as in gamification, and no thread merge or split for tidying messy discussions.

The Best-Looking Platform In The Series

This isn’t one of the nine scores, but it’s a reason people pick Gradual, so it earns its own section. Gradual is the best-looking platform in this series. The graphics, the event pages, the way content is laid out, and the member-facing polish are all a step ahead of the tools that outscore it on raw features. For a brand, or an events and education community, or a marketing-owned one, that isn’t a nice-to-have. It’s part of the job, and it’s something most organisations would happily put their logo on out of the box.

The downside is that the polish comes with limits. When you go into the branding admin you can set logos, favicons, and a single brand colour, and there’s a menu, a footer, and a widgets-and-nudges builder. It stands up fast, but it is not the code-level control of Vanilla’s widget builder or an open Discourse. You’re buying a product that looks great and is limited in how far you can bend it. For most brand communities that trade is fine.

The Middle: Discussions And Mobile

Two capabilities sit in the middle at a six.

Discussions and Q&A. Gradual’s forum is competent and current. You get up to fifty boards across community, chapter, and topic-level scopes, and six real post types, including answered and unanswered questions tracked separately, so the persistent Q&A is there. Posts carry reactions, views, and a watch count. What it’s missing against Discourse is the modern-forum depth: read-position tracking, real-time presence, wiki mode, and thread merge and split. It’s a very good forum, just not the best one here.
Mobile. Gradual has native iOS and Android apps, white-labelled to match your community, which puts it ahead of Vanilla and Gainsight, both on a three with no real native app. That’s a quiet advantage worth naming, because almost nobody else in this series has it. Confirm the apps are generally available on your plan when you talk to them, but on the card it’s a real plus.

The Three Gaps

Three capabilities are where being new shows.

Ideation. This scores a three. Ideas is one of the six forum post types, and members can react and vote, but that’s the extent of it. There’s no dedicated ideation board, no status workflow moving an idea from under review to planned to shipped, and no weighting by account value. If feeding your product roadmap is the main job of your community, Gainsight is miles ahead here at an eight. This is a real gap, not a matter of framing.
AI. This scores a five, and it’s the thinnest area. The AI you can see today is in moderation, the content classifiers doing a first pass on every post, plus a member-facing system that answers from the community’s own content. It’s early, and it’s a narrower story than Discourse and Gainsight, who are shipping whole suites of AI agents. If AI readiness is the thing driving your decision this year, this isn’t the platform leading on it.
Analytics. This also scores a five. The overview dashboard is okay, and the per-event and per-course analytics are good. The reporting area gives you a set of ready-made reports you generate and export, covering members, events, content, forum posts, comments, and search history. What it doesn’t give you is the build-your-own dashboard that Vanilla does. If your execs want a board-ready view they can assemble themselves, you’ll feel that gap.

The Verdict: Should You Build On Gradual?

Buy Gradual if your community is built around events and education, if content and design matter to you, and if you’d rather have events, courses, and community in one good-looking product than stitch three separate tools together. It’s a strong fit for a brand or marketing-owned community, and for a customer education motion where the courses and the events are the point.

Look elsewhere if your community is mainly support deflection and product feedback, if mature build-your-own analytics are non-negotiable, or if AI readiness is the one thing you’re deciding on this year. The shape of the scorecard is the whole story. This is a specialist with a clear identity, focused on events and content, with discussions almost secondary to that.

On pricing, Gradual don’t publish numbers, so you’ll be talking to their sales team. As a rough floor, a community of a few thousand members starts around 20,000 dollars a year and scales up from there. The single biggest thing that moves that number is whether you use the platform’s own video calls and webinars, which cost more on Gradual’s side, so leaving those out drops the price a fair bit. Onboarding, training, and support are usually included, and the pitch is that it replaces several point tools at once. As with every platform, get the quote and don’t take a number off a slide, including mine.

Where To Go Instead

If Gradual isn’t the fit, here’s where I’d point you. Pick Higher Logic Vanilla if you want the cleanest question-to-answer-to-knowledge loop for support. Pick Gainsight CC if you’re B2B SaaS and want your roadmap tied to your customer accounts. Pick Discourse if your community is discussion-led and technical. And if you want the simpler, creator and paid-membership version of the same all-in-one shape, look at Circle.

For a brand, events, and education community that wants one good-looking product rather than three stitched together, Gradual is a strong answer. For a lot of teams, it’s the best platform they haven’t heard of.

More In This Series

This is part of my 2026 series reviewing the major enterprise community platforms against the same nine-capability scorecard. If you’re weighing a move, these go together:

The post Should You Build Your Community On Gradual? first appeared on FeverBee - Community Consultants.

The Rise of Community Marketing

Richard Millington — Mon, 17 Aug 2026 10:00:00 +0000

The Rise of Community Marketing

An increasingly common refrain today is that ‘community should support marketing goals‘.

And there’s obvious logic in that. The budget for support costs is very much finite, while the budget for growth is theoretically infinite.

The problem is you can’t simply say you’re now supporting marketing goals and expect anything to happen. That’s because this isn’t a semantic exercise. It’s a fundamental change in the techniques, platforms, and internal level of collaboration. It takes more than words to turn the oil tanker around.

But the change begins with an understanding of Community Marketing as a concept.

Today we’re publishing our free strategic briefing outlining three key things:

What is Community Marketing (and why it’s on the rise today)
Three Types Of Community Marketing and When To Use Each
How to Prove The Value of Community Marketing

The strategic briefing is 100% free.

Thanks to Bettermode for supporting this strategic briefing.

The post The Rise of Community Marketing first appeared on FeverBee - Community Consultants.

So, You’ve Inherited a Community: Three Big Decisions

Richard Millington — Tue, 11 Aug 2026 13:00:00 +0000

Three times in the past year, we’ve been asked to evaluate an organisation’s community by its new leader. And I’m frequently surprised by how often our results, those rooted in data and evidence, differ so greatly from those of the folks responsible for the community.

I suspect the result comes down to how we evaluate.

Here’s a common scenario.

You’ve just taken over a community you didn’t build. Perhaps you’re the new community head, VP, or CMO – community is now a part of your role.

Maybe you’ve worked your way up through the tiers of community, or perhaps you’ve been reassigned or acquired community responsibility outside of your work.

But, quite quickly, you’re going to get a lot of opinions about the community from colleagues, members, and peers.

The platform is dated, engagement is down, the team is stretched.

This naturally leads into recommended actions. Someone wants a redesign, someone wants a new platform, someone wants to shut the whole thing down, etc…

There’s a real danger here of accepting these views as fact and taking the wrong course of action entirely.

Before you do anything, you need to determine what you’ve actually inherited here.

This has to be an evidence-based process, rather than opinions. Because you’re going to be faced with one of five major strategic choices, and you need to get it right.

You can find the full version in our Community Strategy Guide. But, for simplicity, we can strip everything down to three simple questions.

If you’re following the ‘first 90 days‘ formula, I’d recommend undertaking these three questions and then making the one critical decision about what to do next.

Key Lesson: Don’t Start With The Platform

I’ve noticed that the less familiar new community execs are with the concept of community, the more likely they are to focus on the platform.

That makes sense; it’s the most visible aspect of community.

And if you’re used to modern CMS or LMS experience, community platforms can look pretty ugly.

So the instinct is always to think about updating it. This is a problem for two reasons.

Migrating a platform is significantly more difficult and costly than you might think.
It’s rarely the root cause of major problems in a community.

This doesn’t mean you shouldn’t ever do it, but it shouldn’t be the first thing you look at.

More often than not, the community is set up fine, and the problems lie in how it’s run, who it’s for, and whether anyone ever agreed on what it was supposed to achieve.

You can’t know that until you’ve looked.

If you do end up needing to compare options later, our platform comparison tool is there. Just don’t lead with it.

It’s far better to focus on the real questions first.

Question One: What Is The Experience For Members Really Like?

I’m constantly amazed how frequently those closest to the community don’t objectively evaluate the experience of being a member.

Go through the community the way a member does. Sign up as a new user. Ask a question. Try to find something. Read the last week of activity. Hardly anyone does this on a community they’ve just taken over. You will quickly become familiar with the experience and likely problems.

If you want a complete picture, follow FeverBee’s benchmarking process across a set of categories: liveliness, how content is organised, the quality of answers, events, navigation, look and feel, the mobile experience, and whether there’s a clear call to action.

Each one gets a rating, so you can get a clear scorecard. Our Community Benchmarks & Standards set out what “good” looks like in each category.

A couple of specific things to think about:

How quickly do questions get a useful reply? If your median is slower than 24 hours, members have usually gone elsewhere for the answer before it arrives.
What share of questions ever get resolved? In a healthy support community, you expect a decent proportion to reach a marked solution. If few questions are being marked as resolved, that can indicate a problem.
Where do new members drop off? Onboarding is where most of the loss happens, and most onboarding effort is spent in the wrong place. Our onboarding journey resources cover what actually moves the number.

It’s worth reviewing key steps to improve the community experience and some common misunderstandings about building successful community experiences.

If you don’t know whether the current community experience is good, it’s very difficult to decide what to do next.

Question Two: What Do Members Actually Want?

I can’t stress enough the importance of engaging directly with members to find out what they really want. Communities are often launched upon assumptions – which may or may not be true – but which are then never updated.

Many of the assumptions about what members want, how they want to engage, and why they want to engage are remarkably out of date.

Two weeks of listening here will save you a year of guessing.

Do three specific things here:

Interview eight to twelve members. Not a survey, actual conversations. Include people who left, people who lurk, and a couple of your most active members. You’re listening for the goal they were trying to achieve and where the community helped or got in the way. Our member interview questions give you a script to start from.
Run a short survey to put numbers behind what the interviews surface. Keep it tight and behavioural, not a satisfaction poll. Our example member survey is built for exactly this.
Segment what you learn. You’ll usually find the community is serving three or four different audiences, not one, and each wants something different. Our segmentation template and persona template help you set that out so the rest of the organization can see it too.

You end this question with a clear picture of who your members are and what each group is really there to do. You’ll use it in every decision that follows.

Aside, a common mistake is to only speak to the most engaged people in your community (these are the folks most likely to respond to requests for interviews). It really helps here to speak to the desired audience who aren’t participating in your community.

Find out what their needs are, the level of their desire, frequency of need, where they go today, and why. The results illuminate what you aren’t doing today but should be considering before launching any other programme. You can map it out like this:

Reflect carefully here if the community is truly delivering an experience that matches the needs of your members today – and to what extent is that need evolving?

Question Three: What Is This Supposed To Produce?

There’s no shortage of organisations investing six- to seven-figure sums annually within their communities without ever being entirely clear on the desired outcome.

In some organisations, stakeholders we speak to highlight competing objectives for the community. In others, people give vague ideas (support/marketing) without being crystal clear in the role the community plays.

The key here is to ask the people who fund it what they think they’re buying. You’ll often get three different answers from three stakeholders, and that gap reveals a communication and alignment problem to resolve later.

You’re checking three things:

Is there a valid north star? Does the community have a clear north star, and is everyone aligned on it? Does it clearly support deflection, retention, product feedback, advocacy, etc?
Is the mechanism for those results clear? How specifically does the community achieve that goal? Does it change how many support tickets are resolved? Do members learn how to do more with the product? Does it improve member sentiment about the company or its products, etc?
Is there a valid measurement process? Is there a measurement model in place that would survive the sniff test? How is attribution calculated? Is the data accurate and good? Are the results clearly displayed for everyone to see?

You don’t need the full measurement model in ninety days. You need to know whether the community has a model and whether the model is reliable.

Our 46 questions for uncovering community goals, our guide to building a community dashboard, and our community ROI guide take you further when you’re ready.

The dream community dataset shows what a complete picture looks like.

Create Your Problem Statements

Based on this, you should be able to turn your observations into some clear problem statements.

There are some good and bad versions of this.

Bad is: “engagement is low and the platform feels dated,”

That’s a vague opinion. Good names the segment, behaviour, evidence, and cost. For example compare:

Bad: “Onboarding needs work.”

Good: “Around 60% of new members never return after their first visit. They arrive with a question, don’t get an answer within a day, and find it elsewhere. That’s roughly [X] members a month we acquire and immediately lose.”

The latter is a measurably clear problem to act on.

Likewise,

Bad: “Members aren’t very engaged.”

Good: “Around 30% of members told us they want to connect with peers in the same role, but there’s no way to find people like them, so they post once and leave. Our most valuable segment has no reason to stay.”

Again, it’s specific, measurable, and clear that there’s a problem which needs to be resolved.

One more example:

Bad: “We’re not showing value to the business.”

Good: “The community resolves product questions that would otherwise become support tickets, but we’ve never counted them, so at renewal we can’t say what it saves. Support leadership currently assumes the answer is zero.”

Your numbers might be your own, but the shape is what matters. You get the idea – give clear, specific problems that affect an identifiable audience.

Don’t worry about solving them yet; properly identifying the problems is more important at this point – and most problems naturally point to obvious solutions.

What Type Of Strategy Do You Pursue?

Once you have all this information, you can decide what kind of strategy makes the most sense for you.

You can learn more about which type of strategy each is here.

Essentially:

Optimise what you already have. If members’ needs aren’t changing and the environment is stable, work on incremental improvement based on your results above.
Repair. If the community isn’t successful, but the environment is stable, then identify the issues and repair them.
AI-Ready. If the community is doing well, but the environment is changing (noticeable due to AI), align the community to those goals.
Community Belonging. If the community isn’t doing well and the environment is changing, one of the two approaches is to focus on belonging. In a world of AI, belonging can be scarcer.
Community Everywhere. If belonging isn’t the right answer, the final approach is to engage members where they participate. This is a more decentralised approach.

If it helps, use your research above to complete this short survey.

This will guide you to the strategy options which may make the most sense to you.

After this process, you should be able to walk into any meeting with a clear direction and vision for the community – and defend it with real data and coherent thinking.

Or Hand The Project To Independent Consultants

While you can run a short version of the process above yourself, you will come up against two problems.

Limited time. Interviewing enough members to be sure, running the survey, and benchmarking your community against others in the same category is a lot of work, and it lands on top of everything else in your first quarter. It’s the part that often gets skipped. You likely need someone you can outsource some or all of this project to.
Bias. You own this community now, so you have a stake in the answer before you start. Without meaning to, you’ll lean towards the evidence that fits the plan you already like, and towards protecting the parts you’ve inherited. Colleagues and members often sense this and align their responses to what they feel the interviewer wants to hear.

This is why I believe it typically helps to hand the diagnosis to us and get a clean, unbiased evaluation and the problem statements back. It saves time and typically delivers better results.

If you’ve inherited a community you know is underperforming but can’t yet say why, that’s where we start.

The post So, You’ve Inherited a Community: Three Big Decisions first appeared on FeverBee - Community Consultants.

Should You Build Your Community On Higher Logic Thrive?

Richard Millington — Thu, 30 Jul 2026 14:45:00 +0000

Higher Logic Thrive usually gets filed under one heading, an association platform. That heading is true, but it undersells the product. Thrive is the one platform in this series that tries to be the whole stack in one place: community, email marketing, AI, learning, and events, all switched on from one admin and often run by a very small team.

I scored it on the same nine capabilities I use for every platform in this series, so you can put it side by side with Higher Logic Vanilla, Khoros, and Gainsight CC. It came out as a breadth play. A wide band of sevens, a couple of sixes, and one soft spot at a five.

The first thing to get straight is which Higher Logic you’re actually looking at.

Two Products, One Company

Higher Logic sells two community platforms, and buyers mix them up. There’s Higher Logic Vanilla, which I reviewed earlier in the series, a more traditional forum built for support and heavy moderation. And there’s Higher Logic Thrive, the product here, an all-in-one engagement suite built for members having ongoing conversations with each other. It’s the same company, but two very different products.

The design choice underneath is the sharpest way to tell them apart. Vanilla, like most platforms in the series, opens into a forum: categories, then threads, then posts. Thrive opens onto a wall. When you log in, the first thing you see is a personalised feed built from the groups you’ve joined, with your profile and your points beside it. Higher Logic call it My Activity Feed, and they’re deliberate about it. They’re not chasing the attention economy. They’re surfacing the latest from your groups to bring people back daily.

That’s the real line between the two products, and it isn’t associations against corporate anymore. It’s whether you’re building a community that lives on questions and answers, or one that people visit every day to see what’s new. Thrive is built for the daily visit, and that explains most of the scorecard. It’s strong everywhere daily engagement lives, and softer where the traditional question-and-answer forum is strongest.

How Higher Logic Thrive Got Here

The history goes back further than most people realise. Higher Logic was founded in 2007 out of the association world, when the CIO of a risk-management society wanted better tools to engage members. For close to two decades it’s been the dominant platform in the association space, and it grew by buying the pieces around the community: a run of email and marketing companies, Informz among them. That’s why marketing automation is so native to the platform rather than a bolt-on. Vanilla only joined in 2021. Thrive is the brand that pulls everything into one suite.

The knock, until recently, was the back end. This is close to a twenty-year-old platform, and it felt its age: an older engine under a community that kept getting new features piled on top. That’s changed. Higher Logic launched Thrive Core at their Super Forum event, which adds AI tuning, official responses, and a task system. And they’re rolling out a single new admin experience over the next six to seven months to replace the old portal. The thing that used to be the weakness is now the thing that’s moved the most.

What Thrive Does Well

Six capabilities score a seven: AI, content, events, mobile, gamification, and analytics.

AI. It’s strong on the breadth of search rather than on agents that take actions. The member-facing assistant, Floyd, answers from your own community content and cites where each answer came from, searching across discussions, the library, events, and courses in one place. When it isn’t confident, it doesn’t guess. Ask it something it has no source material for, and it tells you so and points you back to the community. On the admin side you can build custom search agents scoped to a topic or a language, deploy an assistant onto an external site, and set a confidence threshold so weak answers get held back. You can also weight results by recency so old posts stop resurfacing as current advice, which is a control most community AI doesn’t give you. What keeps it off an eight is the ceiling on what the AI does: there’s no MCP yet, no AI moderation agent, and no sentiment analysis. The agents search and answer, they don’t take content actions.
Content and knowledge. It covers more content types than most platforms in the series: per-community libraries and document sharing, member blogs, video, CMS-built pages, and Thrive Learn, an LMS that puts real courses, learning paths, and continuing-education credits inside the community. Every other platform in this series lost a point on content for the same reason, no courses, and Thrive has them. The limit is that this is a broad library rather than a structured, versioned knowledge base, and the LMS comes through a learning partner rather than being native.
Events. This is the one that surprised me, because most platforms in this series do a poor job of events. The events admin is a proper tool: more than ten pre-built event types, from a single webinar up to a multi-day conference with multiple tracks, plus paid registration, discount codes, presenter roles, and a native calendar. There’s a dedicated mobile events app for the in-person side, with session check-ins, badge QR scanning, and session polling, and an abstract-submission and review workflow. That’s real conference management, not an RSVP bolted onto a forum. The limit is that Bevy and Gradual go deeper on virtual and hybrid production.
Mobile. This one is worth flagging, because it’s unusual here. Most enterprise community platforms have no native app at all, and Vanilla and Gainsight both score a three, responsive web and little else. Thrive has a properly branded native app on both the Apple and Google stores, covering community, events, push notifications, learning, surveys, and gamification, with a separate events app for conferences. The downside is that the app is powered by a partner, Results Direct, and the deeper white-labelling sits on the paid Mobile 365 tier rather than coming free with every plan.
Gamification. It’s deeper than the standard points-and-badges box-tick. Out of the box there are 46 ribbons and badges, and behind them an automation-rules engine driving a tiered contributor-point system, bronze up to gold. Points come from across the platform, blog posts, course completions, best answers, not just raw posting, and members get an automated message the moment they earn a badge. It’s tied to contribution and recognition rather than vanity activity.
Analytics. It opens onto a set of dashboards across community, benchmarks, discussions, users, and events, with ideation, events, and the smart newsletter each carrying their own reports and a benchmark view that compares you against a wider set. The limit is that there’s no build-your-own dashboard tool of the kind Vanilla offers, and the strongest version of the reporting gets better the more you feed it from your association-management or CRM data.

Marketing In The Same System

This isn’t one of the nine capabilities, but it’s a single reason people pick Thrive, so it earns its own section. Thrive has a full marketing automation system sitting inside the community, in the same admin, on the same data. It’s their Thrive Marketing product, built on Informz, and it ships pre-built smart campaigns for the jobs a community manager actually needs: onboarding, profile completion, and re-engaging lapsed members.

Take that last one. You can re-engage everyone who signed up more than ninety days ago and never came back. On most platforms that’s a whole project. You brief your marketing team, they build it in HubSpot or Salesforce, they pull the community data across, and then you wait your turn behind the sales campaigns. In Thrive it’s a smart campaign that triggers on inactivity. You switch it on. That matters most for customer success and education teams who own onboarding and the learning journey and don’t want to queue behind a marketing department chasing prospects. The downside is that it’s another system to learn and run, and marketing automation done badly is just more email nobody opens.

The Middle And The Soft Spot

Two capabilities sit in the middle, and one is the platform’s real soft spot.

Discussions and Q&A. It scores a six. It’s a clean, functional threaded forum with categories, tags, and accepted answers. What holds it back is that the Q&A isn’t a distinct, purpose-built content type the way it is on Vanilla or Discourse, and it’s missing the modern-forum polish like deep threading and read-position tracking.
Ideation. It also scores a six. It’s a proper module with a status workflow, categories, and its own reports, where members submit and vote and watch ideas move toward the roadmap. What keeps it out of the top tier is what Gainsight has and this doesn’t: the ability to weight ideas by the revenue of the accounts asking, and a built-in link out to product tools like Jira or Productboard.
Moderation. It scores a five, the one real soft spot, and Higher Logic agree. What’s there is solid and standard: a review queue, watch words, flood control, and pre-moderation of first-time contributors, with the automation-rules engine able to trigger moderation actions. What’s missing is the modern layer: no AI classification doing a first pass on incoming posts, no sentiment analysis, and no trust levels that auto-promote reliable members to reduce the load. If heavy moderation is what you need, Higher Logic themselves point you to Vanilla, because Vanilla is built for support and sentiment-heavy communities and Thrive is built for members talking to members, where you need less of it.

How Far You Can Shape It

This isn’t one of the nine scores, but it’s becoming one of the most important questions to ask any vendor: how far can a non-developer shape the platform? Thrive goes a long way. Every page is built in a drag-and-drop editor with a large widget library, and you can drop in your own HTML, JavaScript, or images. You can spin up microsites and spaces with their own URL, navigation, and colour scheme under the main community, so a product line or a chapter feels like its own place. And you can personalise any widget by audience, so the same page shows one thing to a logged-out visitor, another to a member, and another to an MVP.

The limit worth naming is that this is configuration flexibility, not code-level flexibility. There’s no MCP yet, and the platform underneath is opinionated. If your plan is to build something completely bespoke, an open-source platform like Discourse gives you more room. Thrive’s whole pitch is that you give up that last mile of control to get a product that’s fast to stand up and hard to break.

The Verdict: Should You Build On Higher Logic Thrive?

There’s no single peak on this scorecard. No eight or nine, and that’s the point. Thrive is the widest all-in-one product here, good at almost everything rather than best at any one thing. It’s strong nearly everywhere, with one real gap, and that gap is the one its own maker points you to Vanilla for.

Buy Higher Logic Thrive if you want breadth out of the box, community and marketing and AI and learning and events in one system, and you’d rather switch capabilities on than build and integrate them. It’s a strong fit if you run a small team, or close to no team, because it’s designed to be run that way. And it’s a natural home for associations and member organisations, where the management-system integration and the all-in-one model line up with how you already work.

Look elsewhere if moderation is the centre of your community, where Vanilla, Gainsight or Discourse fit better, or if you want heavy code-level customisation, where Discourse gives you more room. And if virtual events are the point, Bevy and Gradual go deeper. The all-in-one model cuts both ways. If you want one system that does most things well and can be run lean, that breadth is the appeal. If you want the best-in-class tool for one specific job, Thrive isn’t it.

On pricing, Higher Logic don’t publish numbers, so you’ll be talking to their sales team for a real figure. As a rough shape, Thrive tends to land somewhere between 20,000 and 100,000 dollars a year, driven by the modules you switch on and the size of your community. The marketing automation, the LMS, and the mobile app are the parts most likely to move that number, so scope what you’ll actually use before you buy the full suite.

Where To Go Instead

If Thrive isn’t the fit, here’s where I’d point you. Pick Higher Logic Vanilla if you want the cleanest question-to-answer-to-knowledge loop for support. Pick Gainsight CC if you’re B2B SaaS and want your roadmap tied to your customer accounts. Pick Discourse if your community is discussion-led and technical. And if events are the whole point, look at Bevy or Gradual.

If you’re leaving Khoros, the real question isn’t whether to go, it’s which of these you land on. For an all-in-one association or members’ community, Higher Logic Thrive is a strong answer.

More In This Series

This is part of my 2026 series reviewing the major enterprise community platforms against the same nine-capability scorecard. If you’re weighing a move, these go together:

The post Should You Build Your Community On Higher Logic Thrive? first appeared on FeverBee - Community Consultants.

Community Platform Overage Fees Are Becoming A Big Problem

Richard Millington — Mon, 20 Jul 2026 11:07:45 +0000

Have You Been Hit With An Unexpected Invoice? What Should You Do?

Over the past six months or so, several organizations have contacted me about billing errors from one specific platform vendor. A handful are being charged $100k+ for a single quarter in additional, unexpected bot traffic.

This wasn’t the result of a new campaign, a favorable change in algorithms, or an influx of new customers. It was sharp spikes that arrived without warning and vanished just as fast.

In recent months I’ve reviewed the data of three clients and came to the same conclusion. The traffic which presents itself as human most definitely isn’t.

But this presents a problem. What should you do in this scenario?

Bot Traffic Now Exceeds Human Traffic

Over the past year or so, most communities reached a point where bot traffic exceeded human traffic.

I analysed a few of our clients and discovered some have seen bot traffic reach 90%+ of all traffic to their community. As best as I can tell, the majority sit somewhere between 50% and 80%. That’s a staggering increase in usage for the vast majority of communities.

And it comes at the exact time that human visitors are declining because of the collapse of search traffic.

For major sites like Reddit and Stack Overflow, the problem became so extreme they’ve struck paid licensing deals with the AI companies, and Wikipedia’s parent now sells a paid enterprise API for bulk access.

Communities don’t have that leverage. And they don’t want to block everything anyway, because they still want the useful crawlers: Google, so members can find them in search, and the AI engines that cite their sources and send readers back.

What they don’t want is to pay to be scraped by the extractive crawlers that take everything and return nothing.

This is the catch with blocking. It isn’t free. Block the wrong crawler and you lose your search visibility or your place in AI answers. So the goal isn’t to shut bots out, it’s to stop paying for the ones that give you nothing back.

Part of the challenge is that building and running crawlers has become cheaper and easier than ever. On top of that, AI agents that fetch pages live, on demand, have multiplied the number of requests hitting your community. User-triggered crawling grew more than 15x last year alone.

Part of it is just the incentive to grab this data before more organisations lock it down.

Two things are happening here.

The baseline is rising permanently. Every new model generation re-crawls the archives, and agent traffic climbs as more people use AI to answer questions, so the floor under your usage keeps moving up and won’t come back down.
Spikes are becoming a problem. They’re the thing clients keep reporting, and the thing that actually blows the budget. These are sharp, sudden surges of traffic pretending to be human, arriving without warning and pushing them over their usage allowance in a matter of days.

What Are These Bots Exactly?

A bot is essentially just software that requests pages in the same way a person does – but there’s no human reading the other end. For the best part of two decades, they were a good thing. Googlebot would crawl your community, index valuable content, and then surface them when people requested relevant information.

Over the past few years, however, training crawlers like GPTBot, ClaudeBot, and CCBot pulls entire archives in bulk to feed their language models. Worse yet, retrieval and agent bots will fetch pages in real time when someone asks a question, rather than just sending them a link to the potential answer. The numbers vary, but AI might be retrieving 30+ pages to effectively answer a single query.

The Real Scourge Isn’t The Good Bots

But the real scourge here isn’t the clearly labelled LLMs or search bots, although they do make a lot of data requests. The real challenge is the growing share of bots, sometimes as high as 30%, that pretend to be humans (or other bots).

This means they use generic browser identifiers and datacenter connections to look like ordinary visitors.

Their goal is to consume knowledge and information without honouring blocks or, in some cases, being forced to pay Reddit, Stack Overflow, or Wikipedia. They don’t honour robots.txt, because that would defeat their entire objective.

And this is where the real problem begins.

Because if you are on a usage-based platform package, as most of you are, and this dark bot traffic causes you to go over your agreed usage allowance, you may get hit with significant overage fees.

There’s a nasty irony here, and it drives everything below. The bots that are honest about being bots are the easy ones to deal with. It’s the ones that lie about being human that end up on your invoice.

Be fair to the vendors here. Most of them aren’t villains and most want to help. But they get charged for this traffic too, and that charge is the same whether the visitor is a human or a bot. When a vendor passes an overage on to you, it’s often a bill that landed on them first.

Historically, most smart vendors, when confronted with obvious proof that traffic is non-human, waive the overage charge for the benefit of the relationship. The vendor, in essence, eats the cost and tries to do a better job filtering and blocking it in future. They do this for the benefit of the client relationship (or, ideally, because of carefully negotiated contractual terms, more on that later).

But not all vendors are alike.

Today a growing number of organisations are being threatened with legal action (or even closure of their community) if they refuse to pay the overage fee. And this creates a big problem, because it messes with budgets and calls the community’s viability into question.

What Should You Do If You’re Hit With An Overage Charge?

The most important thing to do when you get an overage request is not to rely on the platform vendor’s analytics or tools like Google Analytics. Instead, you need to request access to the raw server logs. This is a reasonable request to make.

The reason is simple. Your analytics and your server logs show very different things. Google Analytics and similar tools run on JavaScript tags, and most crawlers don’t execute JavaScript, so they’re invisible to Google Analytics. Your server logs, however, or your CDN logs, see every single request. Every time a bot requests a page or data, it shows up in the logs.

Separate Out The Expected Bot Traffic

Now within this, there are some things you can count exactly.

GPTBot, ClaudeBot, and CCBot all publish their user agent string, so you can filter the logs for them and have a hard floor.

Most importantly, you can verify them against the published IP ranges.

OpenAI, Anthropic, and Google all publish those, or you can use reverse DNS. That separates the real bots from the ones trying to pretend to be them.

Create A Range For Dark Bot Traffic

The hard part is the undeclared ones. These you infer from behaviour rather than counting them, but there are some very specific tells to look out for.

Request rate. A single IP pulling hundreds of pages a minute isn’t a person.
Hitting deep archive threads no human would navigate to.
Fetching only HTML and skipping the CSS and images a browser would load.
Requests coming from datacenter IP ranges (AWS, GCP, Azure) instead of residential ISPs.

So, for example, if you see a datacenter hammering your old forum posts at machine speed, that’s a bot, even if it claims to be Safari. This gives you an estimate, not a clean number.

An even better route, when available, is to have a CDN (content delivery network) or WAF (web application firewall) classify the traffic for you.

Tools like Cloudflare and Fastly are already pretty good at labelling verified bots, AI crawlers, and traffic that is likely automated, and giving you a breakdown. It’s probably the most credible, independent number you can use.

Then you can compare the metering number the vendor charges you against your human analytics. For example, if your invoice is built on one million sessions and Google Analytics shows 400,000, that gap is your automated traffic. It’s not exact, but it gives you a rough estimate to work with.

Three Terms In The Contract Are Critical Here

In the contract (or Master Services Agreement) there are three things to pay attention to.

What is the usage allowance? This is typically how many page views and API calls are allowed per month or quarter. This number is often in the low millions to begin with, but can rise significantly for larger organisations.
How is usage defined and calculated? What counts as a view? Many contracts define a view in a way that is meant to exclude bots, often with a phrase along these lines: Traffic not detected as originating from a browser is not counted as page views.
What happens when that usage is exceeded? Do you simply move up to the next tier indefinitely? Or do you get billed separately (often excessively) for the overage?

Of the three, it’s the definition of a ‘view’ that matters most, and it’s worth understanding why it doesn’t protect you as well as it looks.

A definition like the one above sounds reassuring. Bots are excluded, so you only pay for humans. But look at what it actually does. The honest, declared crawlers, GPTBot and the like, announce themselves as bots rather than browsers, so they’re easy to detect and get excluded.

The disguised ones, the dark bot traffic wearing a browser identity, and the AI agents like ChatGPT’s and Perplexity’s browsers that literally are browsers rendering the page, are all detected as browsers. So they count.

In other words, the standard definition protects you from the bots that tell the truth and bills you for the bots that lie. Which is exactly backwards, because the lying ones are the fastest-growing share of the problem.

The definition also puts the burden of proof on you. By this wording, it’s on the client to demonstrate a spike is dark bot traffic and not real humans. And that usually requires the server logs. If the vendor won’t give you access to them, you can’t make your case.

The vendor also has leverage. If you don’t pay your bills, they can threaten to close your community down.

Negotiate These Conditions At Contract Signing And Renewal

All of which means the critical thing to do at renewal is to negotiate all three of the terms above.

Before the specifics, one shift matters more than any single clause: what counts as proof. You will almost never be able to demonstrate, request by request, that a spike was automated. What you can produce is an independent CDN or WAF classification. So the real win at renewal is getting the vendor to accept that classification as sufficient evidence. That moves you off an impossible standard, proving every request was a bot, and onto a fair one, where the independent tooling makes the call.

Negotiate what counts as a view. It’s not enough to rely on ‘non-browser’ traffic when the worst bots are pretending to be browsers. You need a fair way to determine whether a spike or increase is bots posing as humans, and to have that traffic excluded. Language worth pushing for looks something like this:

Page views shall exclude any traffic the Customer can reasonably demonstrate, through server logs or CDN classification, to originate from automated or non-human sources, whether or not that traffic presents as a browser. Where a sudden increase in traffic is shown to be predominantly automated, it shall be excluded from usage calculations and shall not, by itself, trigger overage fees.

Secure timely access to server logs. On request, you should be able to obtain the raw server logs within a defined window, say [x] business days. Without this you can’t prove a spike is automated, and the definition above is worthless to you.
Agree what happens when you exceed the allowance. Pin down whether you simply move up a tier or whether you’re billed per unit at a punitive rate. Cap the overage rate, and build in a waiver or review process for spikes that are shown to be automated rather than human.

Summary

Overage invoices from community platforms are rarely a billing error. They’re the result of bot traffic, most of it now non-human, and increasingly of dark bots that disguise themselves as human visitors to get past blocks and avoid paying for what they take.

If you’re hit with one, don’t argue from the vendor’s dashboard or Google Analytics. Get the raw server logs, separate the declared bots you can count from the disguised ones you have to infer, and use a CDN or WAF classification as your independent number.

Then fix it in the contract. Negotiate the usage allowance, the definition of a view, and what happens when you exceed it. These are the three terms that decide who pays for the bots.

Right now, the standard ‘non-browser’ definition has you pay the cost; renewal should be the time to change that.

The post Community Platform Overage Fees Are Becoming A Big Problem first appeared on FeverBee - Community Consultants.