Skip to content
midgeDocs
Bug reportsBug report examples: good vs bad, side by side
Bug reports

Bug report examples: good vs bad, side by side

Four realistic bug report examples, each shown as a weak report and a strong rewrite, with notes on what makes a bug report easy to act on.

A good bug report example has the same parts every time: a specific title, the environment, numbered steps to reproduce from a known starting point, the expected and actual results with exact error text, how often it happens, and evidence. Below are four realistic bugs, each written twice: first the way bugs are often reported, then a rewrite a developer could act on without asking a single question.

The examples cover the most common kinds of bugs: a flow that silently fails, a layout problem on one device, wrong data, and an intermittent error. If you want a blank version of the format, copy our bug report template. For the reasoning behind each field, see how to write a bug report.

Example 1: A flow that fails silently

A user resets their password but the email never arrives.

The weak version

Title: Password reset broken

Password reset doesn't work. Tried several times. Please fix ASAP.

This report gives the developer almost nothing. Which account? Which environment? Did the form show an error, or did it claim success? “Tried several times” hints at something, but not what. And “ASAP” without an impact statement does not help anyone prioritize.

The strong version

Title: Password reset email not sent for accounts created with Google sign-in

Environment
- URL: https://app.example.com/forgot-password
- Chrome 154, macOS 26
- Account: [email protected] (created via "Continue with Google")
- Time: 2026-11-10 09:12 UTC

Steps to reproduce
1. Go to /forgot-password while signed out.
2. Enter [email protected] and click "Send reset link".

Expected
A reset email arrives within a few minutes.

Actual
The page shows "Check your inbox". No email arrives (waited 30 minutes,
checked spam). Network: POST /api/auth/reset returned 200 with body
{"sent": false, "reason": "no_password_identity"}.

Frequency
Always for Google-created accounts (3/3). Works for an email/password
account ([email protected]).

Severity
Major: users who signed up with Google cannot set a password.

Why it works

The reporter compared two kinds of accounts and found the pattern: only accounts created with Google sign-in are affected. That one comparison narrows the cause from “email is broken” to a specific branch in the code. They also opened the Network panel and copied the response body, which shows that the server knows it did not send anything while the page claims success. That mismatch is a second bug worth noting. The exact time lets the team check the email provider’s logs. If you have never looked at network requests before, our guide on capturing network traffic in Chrome shows how.

Example 2: A layout bug on one device

On a small phone screen, a banner covers the button users need to press.

The weak version

Title: Can't click save on mobile

The save button doesn't work on my phone.

“Doesn’t work” could mean the button does nothing, shows an error, or cannot be reached. “My phone” could be any of thousands of devices and browsers. A developer testing on a large phone will probably see no problem and close the ticket.

The strong version

Title: Cookie banner covers the Save button on the profile page on small screens

Environment
- URL: https://app.example.com/settings/profile
- Safari on iPhone SE (3rd generation), iOS 26, portrait
- Viewport about 375 x 667
- Not reproducible on iPhone 15 Pro or desktop

Steps to reproduce
1. Open the URL in a private tab (so the cookie banner shows).
2. Sign in as [email protected].
3. Scroll to the bottom of the profile form.

Expected
The Save button is visible and tappable above the cookie banner.

Actual
The cookie banner (fixed, about 180 px tall) covers the Save button.
Tapping where the button should be opens the banner's "Settings" link.
Dismissing the banner makes Save reachable.

Workaround
Accept or dismiss the cookie banner first.

Attachments
- Screenshot: profile-save-hidden-iphone-se.png (banner outlined in red)

Why it works

Layout bugs depend on screen size, so the exact device, orientation and viewport are the key facts. The reporter also tested on other devices and wrote down where it does not happen, which tells the developer to look at small-screen styles. The private tab step makes sure the banner appears, so the developer sees the same screen. The workaround tells support what to tell users today, and lowers the urgency in a way the team can judge for itself.

Example 3: Wrong data

The dashboard total does not match the exported numbers.

The weak version

Title: Numbers are wrong

The revenue on the dashboard is wrong. The export shows something different.
Not sure which one is right but one of them is broken.

The reporter noticed something real, but “numbers are wrong” forces the developer to go hunting for which numbers, for which period, by how much. Data bugs need exact values, because the size of the difference is often the clue.

The strong version

Title: Dashboard revenue for October is $1,240 lower than the CSV export

Environment
- URL: https://app.example.com/reports/revenue?period=2026-10
- Workspace: Acme Demo (ID 4821), time zone set to America/Los_Angeles
- Chrome 154, Windows 11

Steps to reproduce
1. Open Reports, then Revenue.
2. Select October 2026.
3. Note the total on the card.
4. Click "Export CSV" and sum the Amount column.

Expected
Dashboard total and CSV sum match.

Actual
Dashboard: $48,310.00. CSV sum: $49,550.00. Difference: $1,240.00.
The difference equals the three orders placed on Oct 31 after 17:00
Los Angeles time (IDs 99812, 99815, 99820).

Frequency
Always for October. September totals match.

Notes (guess)
Looks like the dashboard groups by UTC date and the export by
workspace time zone.

Why it works

The report gives both numbers, the difference, and the orders that make up the difference. That last step took the reporter a few minutes, and it saves the developer much longer. The guess about time zones is clearly labeled as a guess, kept apart from the observations. Mentioning that September matches shows the bug is tied to specific data, not to the whole report.

Example 4: An intermittent error

Uploads sometimes fail, and nobody can say why.

The weak version

Title: Upload fails sometimes

Sometimes uploads fail with an error. Happens randomly.

“Randomly” usually means “for a reason nobody has found yet”. Without the error text, file details or a count, there is nothing to investigate.

The strong version

Title: Uploading files over 10 MB fails with "Upload failed (413)"

Environment
- URL: https://app.example.com/projects/218/files
- Chrome 154 and Firefox 158, Windows 11, office network

Steps to reproduce
1. Open a project, then the Files tab.
2. Click "Upload" and choose a PDF larger than 10 MB
   (tested: 10.4 MB, 12 MB, 25 MB).

Expected
The file uploads and appears in the list.

Actual
Progress bar reaches about 90 percent, then "Upload failed (413)".
Network: POST /api/files returned 413 Request Entity Too Large,
response body is an HTML error page, not JSON.
Console: Uncaught (in promise) SyntaxError: Unexpected token '<'

Frequency
Always for files over 10 MB (6/6). Never for files under 9 MB (0/8).

Attachments
- Screen recording of a failed upload (40 seconds)
- Sanitized HAR file of one attempt

Why it works

The “random” bug turned out to depend on file size, and the reporter proved it with counts on both sides of the threshold. The 413 status and the HTML response point to a size limit somewhere in front of the app, and the console error explains why the user saw a generic message: the app tried to read an HTML page as JSON. The HAR file is sanitized, because HAR files carry cookies and tokens; the HAR sanitizer removes them in your browser.

Turning a weak report into a strong one

If you triage bugs for a support or QA team, you will receive plenty of reports like the weak versions above. Rewriting them is faster than sending them back, as long as you can get answers to five questions:

  1. What exactly did you see? Ask for the error text or a screenshot, not a description of it.
  2. Where were you? The URL or screen name, and the account.
  3. What did you do right before? The last two or three actions are usually enough to start reproducing.
  4. Does it happen every time? If not, roughly how often, and does anything seem to make it more likely?
  5. What were you trying to get done? This tells you the expected result and the real impact.

Ask all five in one message. Reporters, especially customers, answer one round of questions willingly; three rounds and they give up. When you have the answers, try to reproduce the bug yourself and add what you learn, like the comparisons in the examples above. A reproduced bug with your own evidence attached moves through the queue much faster than a forwarded email.

What the good examples have in common

  • Titles that describe the symptom and the condition, not just the feature.
  • A known starting point and the exact data used.
  • Exact numbers and error text, never “wrong” or “an error”.
  • A comparison: another account, device, month or file size where it works.
  • Evidence attached, with secrets removed.
  • Guesses labeled as guesses.

Capturing the evidence faster

Most of the strong versions above needed the console and the Network panel. Midge, a free Chrome extension, captures both while you record your screen or take a screenshot. Every recording includes console logs, network requests, user actions and device info. Passwords, card numbers, tokens and cookies are redacted in your browser before anything leaves your device. You still write the title, steps and expected result; the technical details come attached.

Frequently asked questions

What does a good bug report example look like?

It has a specific title, the environment, numbered steps from a known starting point, expected and actual results with exact error text, how often it happens, and evidence such as a screenshot, console error or failing request.

What is the most common mistake in bug reports?

Missing or vague steps to reproduce. Without them, the developer has to guess how the bug happened, which often leads to a closed ticket marked cannot reproduce.

Should a bug report include a screenshot?

Yes, when the problem is visible. Add the error text as well, because text can be searched and copied, and add a short recording when the bug involves several steps or timing.

Where can I get a blank version of this format?

Use the copy-and-paste bug report template on the Midge docs. It has the same fields as the good examples on this page.