User Feedback Examples: 10 Kinds, With the Real Replies
Shayan, Founder20 min read
We've grouped user feedback on software products into ten kinds: a bug report, a complaint that something is confusing, a performance complaint, a reaction to a price or plan change, a reaction to a redesign or a removed feature, a feature request, praise, a reason for canceling, a question the documentation should have answered, and a public review. Below is a real one of each, quoted word for word from the public thread where it was posted, along with the team's reply where there was one, how long it took, and what happened to the product afterward.
Most user feedback examples get presented as empty survey forms and aggregate star ratings, with none of the actual words a user wrote. Real feedback is messier and much more specific. Out of 54 pieces of feedback collected from public threads for this guide, 16 ended in a visible change to the product and 11 never got an answer from the team.
Messages from users come in on GitHub, a community forum, the App Store, or a public feedback board. If you're still setting up your intake channels, start with our guide on how to get customer feedback. Otherwise, here is what real user feedback looks like and how teams actually handle it.
1. A bug report
A useful bug report tells you what changed, when it started, and what happened instead of what the user expected. On February 27, 2026, a user named olavrb opened a bug report on Obsidian's forum about version 1.12.4 on Windows:
Ever since v1.12.4 on Windows, Obsidian suddenly closes both open vaults / windows when trying to close one of them.
Eleven minutes later, WhiteNoise on the Obsidian Team replied that they couldn't reproduce it and asked the user to check the troubleshooting guide. Eighteen minutes after that first reply, WhiteNoise came back with "Ok, I can reproduce." and posted an immediate workaround in the window frame settings while the team investigated. The very next day, February 28, they shipped a rebuilt Windows installer with an older Electron version that avoided the upstream bug.
Even on a one-day turnaround, the thread ended with one more piece of feedback worth reading. On March 1, olavrb came back with a fair point about silently replacing the 1.12.4 installer in place:
You should've bumped the version number for this one. Both for visibility (hey, we fixed a bug), and because replacing an existing release messes up package managers like Scoop and WinGet: Same version now gets a hash mismatch.
When a bug is alarming rather than merely annoying, users want to know how it slipped through. On May 28, 2026, NikPiermafrost filed a VS Code issue after version 1.122.0 started asking for screen-sharing permission on every launch on Linux, scaring one commenter into shutting down their PC out of fear of npm malware. About 23 hours later, VS Code team member alexdima named the exact cause and owned the testing miss:
This is caused by a new flow for reporting vscode issues (gated behind a setting
issueReporter.wizard.enabled). Unfortunately there is a small snippet of code that executes outside the gated setting. Despite us having a test plan item for linux, this has slipped our testing because the test plan item was mistakenly done on WSL and not on a proper linux distribution. I apologize for this regression.
The issue was closed 25 hours after the initial report, and a team member announced the 1.122.1 release about two hours after that. At the other extreme sits a Logseq issue filed on July 30, 2025 by davidak, who wrote "I added a new tag in a sentence and all text after it was deleted." Fifteen months later, as of November 12, 2026, no one from the Logseq team had answered in the thread, and the only comments not from davidak were two automated check-ins from github-actions[bot] asking if the data-loss bug was still relevant.
2. Something is confusing or hard to find
Usability feedback can be one person saying they stared at your page for ten seconds and couldn't tell where to click, or a grumpy review complaining that a feature vanished when it actually moved behind a new toggle.
On December 30, 2021, prescience-data opened an issue on Supabase's GitHub repository about the top navigation bar on their marketing site:
On the Supabase website, what is effectively the "login" button is named "Start your project" which can be confusing / dissonant when unconsciously Thin Slicing upon page load and expecting to see some derivation of the common
[Login] [Sign up!]top-right button pattern.
Five days later, Supabase team member kiwicopple tagged the issue "good first issue" and "help wanted", then checked with teammate MildTomato, who approved the change and suggested "Sign in" as the label, to match Supabase Studio. A community contributor opened a pull request, and kiwicopple merged it on January 6, 2022, seven days after the report.
When confusion shows up on a public review site instead of GitHub, fixing the misunderstanding doesn't always win back the rating. On July 2, 2026, didier59450 left a one-star Yoast SEO review on WordPress.org saying "Very complicated, and with last version many items disappeared in Gutemberg! I am looking for some new and efficient SEO tool!" Four days later, Plugin Support member Michael Tiña asked which items had vanished from the block editor. The next day, the reviewer found the new button at the bottom of the page that collapsed the panel by default, replied that "all improvements made seem not enough for a better appreciation", and left the one-star rating unchanged.
3. A performance complaint
A performance complaint can be a cold start that takes seconds before anything renders, or lag that sets in over time even while CPU and memory usage look fine.
On June 14, 2025, jupppo opened an issue on Home Assistant's core repository after updating to 2025.6:
Since the update to the 6.0 and also 6.1 my dashboards are very laggy, sometimes displaying a value is also empty. Most of the time I have to wait up to 30sec until graphs are displayed. CPU and Memory Usage seems OK. When I do a fresh restart, it is OK for a while, but after some time it lags again
Eight days later, Home Assistant member MindFreeze stepped into the thread, which ran to 89 comments, and asked the crowd to help narrow down the shared trigger:
At first glance this looks like a recorder issue if custom charts like apex & mini-graph also cause it. I'll be able to investigate more tomorrow. In the mean time, everyone try to think about the common thing between everyone affected. Almost no one has posted their graph card YAML or their DB setup.
The next day, a frontend pull request from MindFreeze fixing duplicated requests in statistics graphs was merged. MindFreeze then traced the broader slowdown to a frontend commit range and closed the issue as fixed on June 24, ten days after jupppo filed it. Not every slowdown comes from a recent regression a team can patch right away. After justin13888 reported 6-to-10-second Android cold starts on Immich's tracker, team member alextran1502 replied in 28 hours explaining that the app fetched and parsed all server data in chunks before rendering anything, and said a refactor was under way. The issue was closed 13 months later, folded into another one.
4. A reaction to a price or plan change
When a team raises prices or ends a free plan, the announcement itself can draw pushback alongside the change.
On January 20, 2026, Bitwarden employee dwbit posted an announcement titled Bitwarden launches enhanced premium plan, noting that "Premium is now available for $1.65 per month ($19.80 billed annually)" without mentioning what it cost the day before. The next day, user marlin posted the most-liked reply in the thread:
In my opinion, every announcement of increased prices, regardless of the company or service, should include the old price for transparency. Premium: Was $10, is now $19.80 (increased by 98 %) Families: Was $40, is now $47.88 (increased by 19.7 %)
Another commenter, fresh.camp8038, wrote that the concern was how the change was communicated:
The price change was folded into a feature announcement rather than stated clearly and directly. For a company that has built its reputation on transparency and user trust, this approach felt out of character.
In that thread, dwbit answered a question about Teams and Enterprise plan features in nine minutes, and a later staff reply noted that an email goes out before renewal, while as of November 12, 2026 no Bitwarden employee had answered the complaints about the price jump or how it was framed.
Contrast that with what happened when Docker emailed Free Team organizations on March 14, 2023 to say their plan was ending, sparking a Hacker News thread where a commenter asked whether any exemptions existed for open source so their images would stay available. Ten days later, Scott Johnston published a Docker blog post reversing the sunset and promising full refunds to anyone who had upgraded in the meantime, writing: "It's now clear that both the communications and the policy were wrong, so we're reversing course and no longer sunsetting the Free Team plan".
5. A reaction to a redesign or a removed feature
An update can work as intended and still draw complaints when it drops an old behavior or switches a new default on, leaving the team to decide whether to keep the change or roll it back.
On December 13, 2025, ShrimpDuck posted on Obsidian's forum after noticing a quiet change in version 1.11.0:
In 1.10.6 and prior releases, creating a link with the link format set to relative or absolute path would automatically include an alias of that file's name, rendering the link as just the file name but retaining the exact path details. ... In 1.11.0 and above this behaviour was removed (confirmed intended on discord but not in release notes)
Thirty-three days later, on January 15, 2026, WhiteNoise on the Obsidian Team posted the entire reply:
This will be reverted in version 1.11.5. No ETAs.
Five days after that note, Obsidian Desktop 1.11.5 restored the default link display text, 38 days after the user reported it.
In August 2022, VS Code version 1.70 enabled its new 3-way merge editor by default. User JounQin opened an issue asking how to disable it with the body "It's not elegant to me personally." Six minutes later, a bot closed the issue as a Stack Overflow question, a comment that picked up 181 thumbs-down and zero thumbs-up while the thread itself gathered 487 reactions. Forty days after 1.70 shipped, VS Code team member isidorn announced in a follow-up issue that they were turning git.mergeEditor back off by default because "The large majority of the GitHub, Twitter feedback we got strongly prefers the previous inline conflict experience."
Compare both of those outcomes with a GitHub Community discussion posted on November 5, 2025 by cb341 after the pull-request "Add suggestion" button was replaced by a Copilot prompt: "You didn't add a new feature I could ignore. You took away something I actually use and need". Twenty-three seconds after posting, github-actions replied automatically with "Thank you for taking the time to share your insights with us! Your feedback is invaluable as we build a better GitHub experience for all our users.", and as of November 12, 2026 no one from GitHub's team had answered in the thread.
6. A feature request
Popular feature requests take months or years to build, and what keeps a community patient in the meantime is plain honesty about what sits ahead of their idea in the queue.
On June 16, 2020, an Obsidian user opened a request for a Typora-like editing mode, a thread that eventually collected 1,131 likes. In March 2021, when a commenter asked that the existing edit mode stay available as an option, ryanjamurphy on the Obsidian Team replied "That's the plan." Two months later, when the thread turned into an argument over whether mobile should have come before the editor, WhiteNoise answered directly:
Folks this is going nowhere. We don't have regrets that we picked mobile first. Once the mobile apps are shipped to the broader public and solid, the next big thing will be editor.
In July 2021, ryanjamurphy updated the thread to note it had moved from the long-term column on their roadmap to short-term, and on November 10, 2021, staff-flagged supporter argentum followed up with a note, "Implemented in v0.13.0", linked straight to the release thread. For a deeper look at collecting ideas and writing back to every status, see our feature requests guide and our templates on how to respond to feature requests.
7. Praise
When a user takes the time to write down what went right, you get to see which change or detail mattered to them and what not to break.
On June 19, 2025, davedev posted on Cursor's forum from Brazil to thank the team for switching the Pro plan from hard usage caps to rate limits:
... our minimum wage is about $200 USD, so paying $20 a month for a tool isn't like the price of a coffee, it's closer to the price of a coffee machine.
Eighteen minutes later, condor, a Community Developer on Cursor's staff, replied with thanks and asked how they were using the product:
Thats a great first post and thank you for the positive feedback. I will pass the thanks certainly directly to the Cursor Team as well. Could you share a bit what you are doing with Cursor, what models you use and how it works for you?
Sometimes a short note telling the user their praise reached the builders is plenty, like Max Tkacz on the n8n Team replying 17 hours after a user's migration post from Zapier to say "This post was shared in our internal company slack and created a lot of positive emoji reactions ...". Even unanswered praise leaves a clear signal for anyone reading the forum. On Syncthing's forum in November 2025, xymeus thanked whoever built the .stfolder circuit breaker after a runaway bash script wiped more than 20 GB on one Linux machine without propagating the deletion to the rest of the cluster, collecting 16 likes, with no reply from the Syncthing team in the thread as of November 12, 2026.
8. A reason for canceling
A cancellation note that includes someone's usage history shows the exact point where a new limit or an unfixed bug broke how they work.
On March 9, 2023, after paid search engine Kagi introduced search caps on its cheaper plans, shepherdjerred posted the most-replied top-level comment in the Hacker News pricing thread:
I cancelled by Kagi subscription because of this. I don't want to have to pause before I search and think "do I need to search for this? should I look through my bookmarks, notes, etc? is this worth one of the 7 searches my $5 plan will cover for the day?". I've been a frequent user of Kagi for the past ~9 months. Last month I made 1600 searches.
No one from Kagi answered in that Hacker News thread as of November 12, 2026. Seventy-five days later, on May 23, 2023, Kagi removed those search restrictions, writing "In response to your feedback and as part of our commitment to enhancing your experience, we're excited to announce the removal of these restrictions for all our paid plans." Then 196 days after the March announcement, Kagi brought unlimited searches to its $10 plan, explaining that "many have stressed about the cost of search and having to change their search habits when switching from an ad-supported service, so we're fixing that."
On September 16, 2024, smgdarien left a two-star WooCommerce review titled "Longtime Developer: I'm Out", naming years-old GitHub issues around inclusive shipping tax in Australia and a redirection loop in WooPayments before starting a move to Shopify and MedusaJS. Two and a half days later, Plugin Support member Sandip Mondal offered to help troubleshoot both problems, and the reviewer posted nothing further on that page.
9. A question the documentation should have answered
When a smart user reads your launch post or pricing page and still has to ask how a basic unit works, that question is really a bug report against your docs.
On September 24, 2025, when Zed switched its AI plans to token-based billing, oakesm9 commented on the Hacker News announcement:
I completely get why this pricing is needed and it seems fair. There's a major flaw in the announcement though. I get that the pro plan has $5 of tokens and the pricing page says that a token is roughly 3-4 characters. However, it is not clear: - Are tokens input characters, output characters, or both? - What does a token cost?
Ten minutes later, morgankrey, who noted in the thread that they work at Zed, linked the model-pricing docs and wrote: "Thanks for the feedback, we'll make sure this is linked from the pricing page. Think it got lost in the launch shuffle." After oakesm9 replied that it was still hard to picture what a million tokens looked like, morgankrey came back 53 minutes after the original comment with a live anchor on the pricing page: "Now live: https://zed.dev/pricing#what-is-a-token. Thanks for the feedback".
Not every doc gap gets closed in under an hour. In April 2024, Badestrand filed a Next.js docs issue asking how to update a page title from a client component in App Router:
Just to clarify, I am not even looking for an answer here, I just want the docs to be updated so that others like me can find the answer quicker, or at all.
The first reply from Next.js member samcx took ten months and the documentation update landed 28 months after the report.
10. A public review
App Store and marketplace reviews carry extra weight because every prospective buyer sees them. A reply that misses the complaint sits out in public too, while fixing the bug behind a one-star review can bring the rating back up.
On October 22, 2022, an Apple App Store reviewer named Elvite left a review on Spark Mail after an update appeared to force long-time users into a "Sent with Spark" email signature behind a paywall, writing "This was a 5 star app now reduced to a 1 star app that I will be forced to stop using due to this problem." A developer response signed Anastasiia addressed the exact bug:
As an existing customer, you maintain free use of all the features you had access to in the previous version of Spark, including disabling the Sent with Spark tagline. Our development team has identified a bug that forces existing users into a paywall and we've already released a hotfix.
Elvite came back, edited the top of the review to note "Spark has since patched this app, so the bottom complaint is no longer relevant ... This app is a good app again though.", and bumped the rating from one star up to four stars. Unlike Google Play, which prints dates on both the review and the developer's reply, the Apple App Store shows no date on a developer response, so there's no way to tell from the page how fast Spark answered.
Set that beside a Todoist App Store review from April 23, 2023, where AppleUserID18 wrote that "paying close to $50 a year for these basic functionalities unfortunately makes the app unfeasible for me." The developer response asks for a ticket number the review never mentions: "Hello! We are sorry to hear this! Could you please let us know the ticket number so we can look into it? ..."
Product feedback examples that changed a decision
Sometimes a piece of feedback challenges a core product bet rather than a bug or a button label, and reversing course takes much longer than a patch release.
On April 6, 2022, swsnider opened issue #900 on Warp's GitHub repository, which picked up 506 reactions, pushing back on having to sign in to use a local terminal:
I know I cannot be the only developer who would never do a cloud login to use a terminal, let alone do it without any indication why it's needed or what it implies will be collected/transmitted to the warp team.
About two years and seven months later, on November 22, 2024, Warp published Lifting the login requirement and closed the issue the same day, writing: "If you were one of the hundreds of developers who let us know that the login requirement was keeping you from using Warp, we heard you. Thank you for speaking up."
The other reversals we looked at moved faster than Warp's, though on very different timelines. Two terms-of-service blowups got fixed almost overnight. Zoom published a blog post and clarified its AI terms one day after an August 2023 uproar, and Mozilla reworded Firefox's Terms of Use in two days in February 2025. On pricing and plan changes, Docker reversed its Free Team sunset in 10 days and Kagi removed its search restrictions in 75 days. Unity announced its per-install Runtime Fee in September 2023 and ended up canceling the Runtime Fee in September 2024, 366 days after that announcement.
How much feedback teams get, and who reads it
No software company we checked publishes how its incoming feedback splits between bug reports, feature requests, and praise. A few teams do publish their raw volume and who sits on the front line to read it.
On Basecamp's Signal v. Noise blog, Noah reported that their support team closed 2014 with 123,350 cases, bringing their median response time across the year down to three minutes and their 95th percentile down to two hours. At a smaller scale, Thomas Paul Mann wrote on Raycast's blog that they put the whole team on a weekly rotation: "Each week, another person of the team reads, filters and replies to every piece of feedback."
When incoming volume outpaces what a team can close, the backlog keeps climbing. Vlad Chesnokov wrote on Gradle's blog in July 2026: "On average, we receive about 110 issues per month, and given that we close 80 issues per month on average, that leaves around 30 additional issues each month." That steady gap pushed Gradle's open issue count from around 2,200 to 3,300. If you're choosing where requests should live once a tracker fills up like that, we compare eight options in the best customer feedback tools.
Keep user feedback tied to the outcome.
Collect requests on a public board, catch duplicates as people type, and email everyone who posted, voted, or commented whenever a status changes. Free to start.
Keeping user feedback connected to the outcome
Once feedback starts coming in, the work is making sure the bug report or idea someone sends in chat doesn't vanish before the fix ships, and making sure everyone who asked hears back without a teammate hunting through six-month-old threads by hand.
We built UserJot to keep that loop together in one workspace. Every plan, including our $0 Free plan with unlimited users and posts, gives you a public feedback board where customers can post, vote, and comment without creating an account. As someone types on the board or inside the in-app widget, duplicate detection suggests matching posts in real time so they can open an existing thread and add their vote instead of filing a duplicate. Every request carries one of six statuses, Pending, Reviewing, Planned, In Progress, Completed, or Closed, and every status change automatically emails everyone who posted, voted, or commented.
If you turn on Conversations, which brings widget live chat and forwarded support email into one inbox and is free on every plan while in beta, our support agent Juno answers routine questions from your docs and the sites you point her at before handing off to your team. When a customer describes a feature idea in chat, Juno searches your public board, shows them the matching request, and files a new post or adds their vote once they confirm. When you outgrow three team members or two boards on Free, Starter is $29 a month and Professional is $59 a month, flat for your whole workspace.
Frequently asked questions
What are examples of user feedback?
Real user feedback includes bug reports like an Obsidian user flagging that closing one vault closed both windows, usability notes like a visitor telling Supabase its "Start your project" button hid the login link, and pricing questions like a Hacker News commenter asking Zed what a token cost. It also covers performance complaints, reactions to removed features, feature requests, praise, cancellation reasons, and public app store reviews.
What are the types of user feedback?
In this guide, we group written user feedback on software products into ten types: bug reports, confusion or usability friction, performance complaints, reactions to price or plan changes, reactions to redesigns or removed features, feature requests, specific praise, reasons for canceling, questions that expose documentation gaps, and public reviews. Teams rarely receive these neatly sorted, since one app store review or forum post often mixes a bug report with a threat to cancel.
How do you respond to negative user feedback?
Address the specific thing that broke or changed, state the real reason without corporate filler, and tell the user what happens next. When a Linux regression in VS Code 1.122.0 prompted screen-sharing popups on launch, a team member explained which ungated code snippet caused it, noted why WSL testing missed it, and apologized, and the 1.122.1 release was announced about four hours later. When the answer is a reversal, spell out what happens to users who already acted on the change, as Docker did when it kept its Free Team plan and promised full refunds to anyone who had upgraded in the meantime.
How fast should a team reply to user feedback?
In the public threads collected for this post, the team's first reply ranged from 9 minutes on a plan question in Bitwarden's thread, 10 minutes on Zed's pricing question, and 11 minutes on Obsidian's Windows vault bug to 23 hours on VS Code's Linux regression, 8 days on Home Assistant's dashboard lag, 33 days on Obsidian's removed link alias, and ten months on a Next.js docs issue. Some threads never got one. For dedicated support queues, Basecamp reported a 3-minute median response time across 123,350 cases in a year. Speed helps on regressions and launch-day confusion, while on deeper product changes a clear answer a month later still beats an instant bot reply followed by silence.
What do you do with positive feedback?
Share it with the team so builders know which details mattered, and reply with a quick question while the customer is engaged. On n8n's forum, a staff member let a user know their post got shared in the company Slack, and on Cursor's forum, a staff member replied in 18 minutes to ask what the user was building and which models they relied on. Even when you don't need a follow-up call, specific praise tells you which quiet safeguards, like Syncthing's .stfolder deletion check, are worth protecting in future rewrites.
