How two ex-Facebook engineers built a feedback tool for product teams, bootstrapped it to millions in revenue, and taught it to read every customer conversation.
In most software companies, the single most valuable sentence a customer will ever say is not written down anywhere useful. It surfaces halfway through a sales call, or buried in a support chat at 11pm, or muttered in a Slack thread that scrolls away by lunch. Someone hears it, nods, and forgets. Canny was built on the suspicion that those lost sentences add up to a company's roadmap.
Canny is a customer feedback management platform. In plain terms: it gives product teams one place to collect what users are asking for, sort it, rank it, decide what to build, and then tell everyone what shipped. It started in 2017 as a set of feature-voting boards - the digital equivalent of a suggestion box where users can upvote each other's ideas - and has since grown into something its founders now describe as a customer intelligence layer. The through-line has never changed: help teams listen, then act on the right thing.
Sarah Hum and Andrew Rasmussen met at Facebook, where she was a designer and he was an engineer - one of the early contributors to React Native, as it happens. They left to start a company, first trying a product-review community called Product Pains that never quite caught fire. The lesson stuck with them, though: people are constantly telling companies what they want, and companies are terrible at hearing it. Canny was the second, sharper attempt at the same idea.
From the start they made an unusual choice for two ex-Facebook builders in the venture-soaked mid-2010s: they didn't raise money. Canny became profitable in under a year and has funded its own growth from revenue ever since. That decision shaped everything downstream - the size of the team, the pace, and the freedom to make calls that a board might have vetoed.
The core loop is easy to picture. A user posts a request, or votes on one that already exists. The team sees demand stack up, moves the strongest ideas onto a roadmap, builds them, and then uses a changelog to announce what's live - closing the loop with the very people who asked. It sounds obvious. Most teams still run this process across a spreadsheet, a Trello board, three inboxes, and their memory.
The product spans feedback boards, roadmaps, and changelogs, plus a thick layer of integrations - Slack, Jira, Salesforce, HubSpot, Intercom, Zendesk, GitHub, and Gong among them. In 2025 the company added Model Context Protocol access, meaning a product manager can point Claude or ChatGPT at their Canny data and ask questions in plain English instead of clicking through filters.
Here is the insight that separates Canny's second act from its first. Feedback boards only capture what users bother to type into a board. But the richest, most commercially important requests almost never land there. They live in sales calls where a prospect says "we'd sign if you had SSO," in support tickets, in app-store reviews, in the chat window. That feedback is real and it moves revenue, and it was invisible.
In 2024 Canny shipped Autopilot, an AI engine that goes and reads those conversations. It scans support threads, sales-call transcripts, and reviews, then extracts the feedback, removes duplicates, files it under the right feature area, and - crucially - can tie a request to the revenue attached to it. The suggestion box learned to walk around the building and eavesdrop, in the useful sense.
There's a quiet advantage buried in this move. When everyone in software raced to bolt AI onto their products, Canny had something most didn't: years of accumulated data about what customers actually ask for and how those requests map to money. The AI feature was strong partly because the ground underneath it was already rich.
Plenty of tools let users upvote ideas. The trouble with pure vote counts is that the loudest request is not always the most valuable one - fifty hobbyist users can out-shout the one enterprise account whose renewal depends on a single missing feature. Canny's angle is to weight requests by revenue impact, so "build this" arrives with a number attached. For a product manager defending a roadmap to the rest of the company, that number is the whole argument.
Against heavier enterprise suites like Productboard, Aha!, and UserVoice, Canny competes on being lightweight, quick to set up, and reasonably priced. Against the many small feedback-board tools - Nolt, Frill, Upvoty, Featurebase - it competes on depth: the roadmap, the changelog, the integrations, and now the AI. It sits deliberately in the middle of the market, which is a crowded place to stand but a defensible one.
| Capability | Lightweight boards | Canny | Enterprise suites |
|---|---|---|---|
| Voting boards | Yes | Yes | Yes |
| Roadmap + changelog | Limited | Yes | Yes |
| Reads calls & tickets (AI) | No | Yes | Some |
| Revenue-based ranking | No | Yes | Some |
| Fast, low-cost setup | Yes | Yes | No |
More than 100,000 companies have registered, and the paying roster reads like a directory of well-run software businesses: Ahrefs, CircleCI, ClickUp, Typeform, Mercury, Axios, Hive, Harness, Strapi, Missive. The typical user is a product manager at a SaaS company who is tired of feedback arriving from nine directions with no way to compare it. Across all of them, Canny reports having captured more than 15 million feedback items and sent over a million product updates.
Canny sells subscriptions on a tiered model - a free plan plus paid Starter, Growth, and Business tiers that scale with features and usage. Estimated annual recurring revenue sits around $3.4 to $3.5 million, run by a team that has stayed under twenty people. The company is remote, ships on a regular cadence, and treats staying independent as a value rather than a temporary phase.
The most instructive decision in Canny's history is also its most counterintuitive. When it introduced a free plan, it lost roughly half its paying customers almost overnight. That is the kind of number that gets a growth team fired. But the free tier pulled in a far larger top of funnel, and average revenue per user more than doubled as the remaining and new customers sorted into the right plans. It's the sort of move that's much easier to make when there's no board asking why the paying-customer count just fell off a cliff.
Zoom out and Canny occupies a specific seat in the software stack: the place where unstructured customer noise becomes a structured decision. Analytics tools tell you what users did. Support tools tell you what went wrong. Canny's territory is what users want next, and increasingly, what it's worth. As AI assistants become the interface through which teams interrogate their own data, being the tidy, revenue-aware database of customer desire is a good place to be standing.
The company remains small, quiet, and profitable while much larger competitors chase the same ground with far more capital. Whether that's an advantage or a ceiling is the open question. For now, Canny keeps doing the unglamorous work at the heart of good products: reading the feedback, ranking it honestly, and shipping the thing that matters.