Why feature voting boards fail (and when they still work)
Short answer. The main downsides of a feature voting board like Canny are that votes count clicks and not the value behind them, that free or vocal users can outvote paying or quiet ones, that an upvote carries no reason, and that a public board turns every unanswered request into a visible promise. Board vendors have addressed some of this: Canny syncs revenue from CRMs and scores posts with weighted factors, and its Autopilot pulls feedback from tools like Slack and Zendesk and merges duplicates. So the honest version is that a board is a good fit when customers will vote in public, and a weak system of record when it is your only source of demand. Modem (our product) takes the other approach: it reads the conversations where requests already happen and groups them into topics. It has no public board, so it does not replace one where public voting is the point.
The pattern is familiar. A team puts ideas on a public page, lets customers vote, and builds from the top. After a quarter or two the board is partly a graveyard of stale requests, and the top ten is a list nobody plans to build. The problems are structural, so it helps to separate the ones a board can fix from the ones it can't.
“export is broken for us”
“any update on SSO?”
“we hit the rate limit again”
Read + clusteredDownside 1: votes count volume, not value
On a board with a free tier, free users usually outnumber paying ones, so the ranking tracks how many people voted and not who they are. One developer essay, Vote Counts Are a Vanity Metric, makes this case. Its author sells a weighted-vote product, so read it with that in mind. Canny's own feature-voting best practices also acknowledges that more technical or vocal customers can dominate a board while quieter ones go unheard.
Ten votes from free-tier users and one quiet email from your largest account are not ordered the way a board orders them.
What boards do about it. The fixes are real, and they take configuration:
- Canny documents a RICE-style Score whose factors can include revenue, opportunities and votes from key segments, each with a weight, divided by effort. It can sync monthly or annual spend from HubSpot and Salesforce.
- UserVoice's marketing says it enriches each submission with account, ARR and contract data.
- Productboard's Salesforce integration attaches opportunity value to features. Which plans include it was unclear in the pages we could read, so check before you buy.
- Savio documents public, private or internal boards that can hide upvotes and randomize order.
- Other fixes in circulation: capping votes per user, or limiting votes to paying or active customers.
Downside 2: an upvote hides what people meant
A board entry is usually a proposed solution ("add a Gantt view"), and the problem underneath ("I can't see which projects will slip") never makes it onto the page. Jason Evanish's essay argues that upvotes do not mean the voters want the same thing, and that there is no conversation behind them. His fix is to treat a request as the start of a conversation, not the end of one. Voters also can't see your strategy or the engineering cost, so the ranking they produce ignores both.
What boards do about it. Comments help, and merging duplicates keeps them together. But a board can't make people explain themselves. You do that by asking, which is manual work no matter which tool you use.
Downside 3: the board only sees what reaches it
Real feedback arrives as a sentence in a Slack channel ("is there any way to export this? kind of a dealbreaker for our finance team"), a support ticket, or a remark on a sales call. Asking that person to also find your board, search for an existing post and vote is a funnel, and not much feedback survives it.
This is the downside the tools have changed most. Canny's Autopilot extracts feedback from sales and support conversations (Gong, Intercom, Slack, Zendesk, Help Scout, Freshdesk, Zoom and tl;dv, with others through Zapier) and merges repeat requests. Featurebase has an AI duplicate finder, and merging brings votes and comments to the parent post. So "most feedback never reaches the board" is less true than it was, if the capture is switched on and covers your sources. What remains is that the output is still a board post with votes, and the raw conversation is one step removed.
Downside 4: a public board is a public commitment
Everything on the board is visible. Customers can read statuses and timestamps, so a top-voted request sitting at "Open" for three years says something whether you meant it to or not. Declining a request in public takes more effort than leaving it open, so decide in advance how you will say no. That is a maintenance cost that comes with the format.
This one cuts both ways. A commenter in an old Hacker News thread argues that customers get frustrated when requests vanish, and that a public forum with visible progress and explained declines is the answer. We found no source showing what happens to customer trust when a team removes a board, so we won't claim either direction. The practical point is that a public board works if someone keeps it current.
Downside 5: the price can track the crowd
Canny bills by tracked user, meaning anyone with a post, vote or comment attributed to them (pricing). Reviewers on G2 and Capterra say this fits poorly when a business has many free or trial users, since every voter counts. Featurebase prices per seat instead, with unlimited end users on every plan (pricing explained). Check the current numbers on each vendor's page.
When a public board is the right choice
A board is a good tool when:
- Your customers form a community that will vote in public. Public-roadmap devtools, games and consumer apps fit. Voting is the point there, and the board is a fair use of the format.
- You want customers to see progress. Canny and Featurebase email voters when a post's status changes, and Featurebase's GitHub sync does the same when a linked issue completes. That loop is built in and tied to the people who voted. Productboard and Savio document their own ways to notify requesters.
- Volume is low and the audience is yours. A free tier such as Featurebase's (one seat, with unlimited end users) makes a board practical for a small project.
- You will maintain it. Merge duplicates, update statuses, and decline in the open.
Where a board is weakest is as the only place demand is recorded, for a B2B product whose largest accounts talk to support and account managers and never visit a portal.
What to do instead, or alongside
Capture where feedback already happens. Read Slack, Discord, support tickets, email and call transcripts instead of asking customers to come to a page. If you already have a board, treat it as one input among several.
Count people and companies, not clicks. Merge the same request across channels and keep the list of who asked. "Export to CSV: 19 requesters, 4 enterprise accounts" is a decision input. "Export to CSV: 214 votes" tells you little about who wants it.
Where Modem fits. Modem reads conversations from the sources you connect and groups reports of the same thing into topics by meaning. Each topic keeps the original messages and the people and companies behind them, and people are matched across tools by exact email plus platform account. Accounts without an email, such as Discord, stay separate until someone merges them by hand. It can also ingest a Canny board as a source, so your existing posts join the same topics. See how that import works.
Its limits matter if you are comparing it with a board:
- No public board, voting, roadmap page or voter emails. For those, use a board, or run one on top.
- Priority is not a vote count. It is an AI-assigned score that weighs severity, how many companies are affected, how often a topic recurs and more. There is no sort by customer count or by revenue.
- Revenue is context, not a ranking rule. Stripe matches customers to people and companies, so billing data attaches to the companies behind a topic. Salesforce data is a read-only sync the agent can query. Neither feeds the score, and the agent joins them to topics when someone asks, for example "which paying customers reported bugs this month?"
- Your account guidance is manual. If your top-tier accounts should count more, you write that into a prioritization instructions box in settings. Someone has to keep it current.
- Grouping is probabilistic. It matches by meaning, so it can over-merge or over-split. You can merge topics or change priority by hand.
We built it this way because the same ask shows up in different words across a dozen channels, so nobody sees its real size. Two reports belong together when the same change would fix them, not when they share a theme. Revenue arrives as context and agent answers instead of a fixed weight.
Close the loop on your own terms. A board notifies voters for you. Without one, you notify the people who asked. Modem has an opt-in automation that posts an internal Slack note naming who asked when a linked GitHub pull request merges. Merged is not shipped, so it notifies your team and nothing goes to a customer by itself. Replies to customers are drafted when a teammate asks, and in Slack and, by default, in the dashboard they wait for approval first. See how to close the feedback loop with customers.
Thirty minutes with a spreadsheet
Skip the tool decision for now. Create one private spreadsheet and spend 30 minutes pulling every feature request from the last month of support tickets and Slack into it, deduplicated, with requester and company names. Keep it private, since it holds customer names. Compare that list with your board's top ten, if you have one. The gap between the two is the argument this guide has been making, and it tells you whether your board is a good sample of your demand or a sample of your most active voters.
For a head-to-head on specific tools, see our comparison of Canny alternatives and Modem as a Canny alternative.
FAQ
What are the main downsides of feature voting boards?
Votes measure popularity among the people who use the board, not value. Free or vocal users can outvote paying or quiet ones, an upvote carries no reason, and a public board makes unanswered requests visible. Billing by tracked user can also make a board with many voters more expensive.
Do voting boards still make sense?
Yes, when your customers will vote in public and you will keep the board current. Public-roadmap devtools, games and consumer apps are common fits. Canny and Featurebase also email voters when a post's status changes.
Can a board weight votes by revenue?
Some can. Canny documents weighted Score factors that can include revenue, opportunities and votes from key segments, and can sync spend from HubSpot and Salesforce. UserVoice enriches submissions with account and ARR data. Setting it up and keeping it accurate is on you.
Does Canny capture feedback from outside the board?
Yes. Canny's Autopilot extracts feedback from tools like Gong, Intercom, Slack, Zendesk, Help Scout, Freshdesk, Zoom and tl;dv, and merges repeat requests.
Does Modem have a public voting board?
No. Modem reads conversations from connected sources and groups them into topics. It has no public board, roadmap page or voter notification emails, so teams that need public voting use a board, either instead of Modem or alongside it.
How does Modem rank topics without votes?
Priority is an AI-assigned score that weighs severity, how many companies are affected, recurrence and more. It is not a vote count, and there is no automatic sort by revenue. You can add written guidance in settings, and the agent can join Stripe or Salesforce data to topics when asked.