What Are We Optimising For Anymore?
For the best part of twenty years, we’ve been optimising enterprise communities for engagement; more posts, more replies, more members, more activity, more deflection.
If the graphs went up – the majority of people didn’t think too deeply beyond that.
There are generally two ways to increase engagement;
- Increase people’s motivation to engage.
- Reduce the friction to engage.
The former is much harder than the latter, so over the course of twenty years, we’ve been on a relentless mission to lower the friction to engaging.
This meant:
- Single sign-on. Removing the separate login so members could access the community with their existing credentials.
- Gamification. Giving points and badges for every post, reply, login etc..
- One-click reactions. Adding likes, upvotes, and emoji reactions so people could participate without writing anything.
- Email notification triggers. Sending digest emails, “you haven’t visited in a while” reminders, and reply notifications to pull people back in.
- Removing onboarding steps. Cutting profile completion, interest selection, and community guidelines from the sign-up flow to get people posting faster.
- Auto-suggested content. Showing trending topics, recommended threads, and “people also asked” to reduce the effort of finding something to engage with.
- Lowering moderation barriers. Pre-approving new members, reducing content review queues, and letting posts go live instantly rather than holding them for approval.
The more members, content, and activity we had, the better. The graphs went up!
There are obvious benefits of more engagement, too. You get a greater diversity of responses, more people to connect with, and a lower response time.
But these are now outweighed by bigger factors
- A greater volume of content makes it harder for people to find the answer they’re looking for.
- More content confuses the knowledge-retrieval AI and leads to greater cleanup effort later.
- More people discourage people from feeling a sense of connection with one another.
None of the five possible futures for enterprise communities depends on driving as much engagement as possible.
Having a big party doesn’t mean everyone had a meaningful conversation. You just have a big cleanup afterwards.
So what if we begin adding friction?
Two Big Changes We Made To A Client’s Community
We’ve been trying exactly this with a mid-tier SaaS community over the past two months.
We made two specific changes:
- We began approving members. We asked members to complete a short form to be approved for membership. This asked for some basic details about them and what they could contribute to the community. We also asked them to agree to two specific things. First, to take the effort to respond to the answers they received without disappearing, and two to take the time to construct questions people could answer.
- We required people to complete a different process to ask a question. Instead of letting members ask a question in a box, we created a question wizard. Essentially, we would ask THEM questions to describe the issue (product type, what they have already tried, ideal outcome, any constraints, etc.).
(Our first attempt let members bypass the wizard and write a question as normal. Almost everyone did exactly that. So we removed it).
These weren’t random interventions. We spent time analysing this community’s data to understand where the quality breakdown was actually happening.
The question wizard was designed around the five most common reasons questions went unanswered in this particular community: missing product version, no description of what the member had already tried, vague problem descriptions, duplicate questions that had already been answered, and questions posted in the wrong category.
The approval form was shaped by patterns we’ve seen across dozens of client projects about what separates members who contribute long-term from those who post once and disappear.
The two commitments we asked for (respond to your answers, take time to construct your question) weren’t arbitrary. They mirror the behaviours we’ve consistently seen in the most valuable community members.
The bypass lesson was also worth noting. If you offer the easy path alongside the hard path, people will always take the easy path. You have to commit to the friction. Half-measures don’t work.
The Results
We’ve only been running this for two months, but comparing March 2025 to March 2026 gives us the cleanest year-on-year comparison.
The results have been genuinely interesting.
- Total post volume dropped by around 20%. Probably due to the following, and the increased difficulty of asking a question.
- New member registrations declined, too. The approval filters deterred some people from even trying to get through them, which filter out those who aren’t willing to invest even a small amount of effort.
But here’s what went up:
- Average question length increased by 87%. Members were providing more context, more detail about what they’d already tried, and clearer descriptions of their situation. This turned out to be a surprisingly insightful thing to measure.
- The answer rate improved by 23%. A higher percentage of threads received at least one response.
- The accepted solution rate increased by 73%. More threads ended with a marked solution, meaning the answers given actually resolved the problem (or the new members were simply more inclined to mark them as accepted solutions).
- Member retention improved by 60%. A significantly higher percentage of approved members went on to make a second post, with no further increase in email prompts.
- Community’s content started being retrieved more frequently in the organisation’s internal RAG search as citations, up roughly 55% compared to the pre-friction* average. The AI tools were pulling from community threads more often.
* Community content was being retrieved roughly 55% more frequently in RAG search, though other RAG improvements were happening simultaneously, so this one is harder to isolate.
Your mileage will vary depending on your community model, your audience, and how much friction you can introduce without losing critical mass. There’s a balance to find, and it’s different for every organisation.
But the broader point stands. For twenty years, the default assumption has been that making it easier to participate makes a community better. That assumption is probably wrong, and it’s been wrong for longer than most of us have been willing to admit.
The communities that will matter most over the next five years won’t be the biggest ones. They’ll be the ones where the quality of what’s shared is high enough that both humans and AI systems treat it as a trusted source.
If you’re rethinking what your community should be optimising for, or you want to explore what adding friction could look like in practice, that’s exactly the kind of work we do at FeverBee.