Bug triage: how to turn bug reports into a clean backlog
A practical bug triage method: merge duplicate reports, write each bug as a clear task with steps and a done-when check, and rank what to fix first.
Free app. Runs on your Mac with the Claude or Codex plan you already have.
What bug triage is
Bug triage is the work of reading incoming bug reports, deciding which are real and distinct problems, and putting each one in the order it should be fixed. The word comes from emergency medicine: sort first, treat in order. Without it, a bug backlog fills with duplicates, vague one-line reports and feature requests filed as bugs, and the team stops trusting it.
The method below is how an experienced product manager runs issue triage on a batch of reports from a tracker, a support inbox or a folder of notes. The result is a short list of tasks a developer can pick up, plus a clear record of what was set aside. Each step says what to do and why, then how 10folds carries it out.
What it does for the business
- Fix what hurts most users firstA ranked backlog sends developer time to the bugs that block money, data or sign-in for many people. Cosmetic problems wait their turn instead of competing for attention.
- Developers start work instead of investigatingA task with steps to reproduce, the environment and the evidence saves the first hour of every fix. The developer can confirm the bug and start on the cause.
- See the true size of a problemTwenty-three reports of the same broken button become one task with twenty-three sources. That shows reach clearly, which a list of separate tickets hides.
- Planning meetings get shorterWhen duplicates are merged and each task has a priority with a reason, the team discusses the few borderline cases instead of reading every report aloud. The set-aside list also shows which reports went nowhere and why.
Bug triage, step by step
Put every report in one table
You cannot spot a duplicate until every report sits in the same place.
What to do
- Export open bugs from the tracker, the support inbox and any error tracker.
- Give each report a stable id, keeping the source's own id where it has one.
- Record the date, the channel, the title and the full text for every report.
- Note what each report lacks: steps, expected and actual results, version or environment.
- Look for candidate duplicates by close wording and shared error messages, codes, URLs or stack frames.
Why
Reports arrive through different channels and in different words, so the same bug looks like three problems. A single table with full text lets you compare them side by side. Shared error signatures are the strongest clue that two differently worded reports are one bug. Listing what each report lacks at this stage also tells you early which reporters you will need to contact.
Decide what is the same bug, and what is not a bug
Two reports are one bug only when they describe the same failure in the same place.
What to do
- Read every candidate group and every report in full.
- Merge only when the faulty behavior and its location match, for example the same button failing with the same error in the same browser.
- Keep reports in the same area with different behavior as separate bugs.
- Set aside feature requests, questions and bugs already fixed, each with a one-line reason.
- For reports that cannot be acted on, write exactly what to ask the reporter.
Why
Over-merging is as costly as under-merging. Two different bugs folded into one task means one stays broken after the fix ships. Duplicate bug reports also hide in plain sight: a report with no error message can match one with a full stack trace if the steps and timing agree.
Write each bug as a task a developer can start
A good bug title says what to fix, not only what went wrong.
What to do
- Write a title that names the fix, such as "Fix Pay now on Safari checkout".
- Give the steps to reproduce taken from the reports, then expected and actual results.
- Add the environment, the first-seen date or release, and every source report id.
- Write a done-when check that someone else could verify.
- Mark anything no report gives as "not reported" and list the open question.
Why
Steps to reproduce and expected and actual results are what turn a complaint into a task. A done-when check, close to acceptance criteria, stops arguments about whether the fix is complete. Invented steps or versions send developers chasing the wrong cause.
Set priority from impact and reach
More reports raise confidence in a bug, not its severity.
What to do
- P1: money, data, security or sign-in affected for many users, or with no workaround.
- P2: a main feature fails for some users, or the workaround is poor.
- P3: cosmetic or rare problems.
- Write the reason for each priority in one line.
- Treat a link to a recent release as a clue to the cause, not a reason to raise priority.
Why
Clear bug priority levels make triage repeatable, so two people sorting the same report reach the same answer. Severity and priority are often confused: severity is how bad the failure is, priority is how soon to fix it given who it reaches. A written reason lets the team challenge a rank quickly.
Check the backlog, then hand it over
Every report should be accounted for exactly once.
What to do
- Confirm every report appears once in the triage table.
- Check that each merge is justified by the reports' wording.
- Check that each task's facts can be found in its source reports.
- Check that nothing set aside as needing information could in fact be acted on.
- Import the tasks and send the questions to reporters.
Why
A report that falls through triage is a customer whose problem is never looked at. A second reader catches merges made too quickly and tasks that drifted from their evidence. It also checks the set-asides: a report parked as needing information is sometimes clear enough to act on once read carefully.
Common mistakes
- Merging on title aloneSimilar titles often hide different bugs in different places.
- Ranking by report countLoud bugs are not always the ones that cost money, data or sign-ins.
- Vague titles"Checkout broken" tells a developer nothing about what to fix.
- Inventing stepsSteps nobody reported send developers down the wrong path.
- Leaving feature requests in the bug listThey inflate the backlog and get compared with real failures.
What a good result looks like
- Every report accounted for once, as a task, a merge or a set-aside with a reason.
- Task titles that say what to fix.
- Steps, expected and actual results, environment and evidence on each task.
- A priority and a one-line reason on each task.
- A short list of exact questions for reports that need more information.
And then what? Using the result
- Import the backlog file into your tracker.
- Start the next sprint with the Fix now column.
- Send the questions to reporters whose bugs need more information.
- Move feature requests to the product roadmap discussion.
- Run triage again on the next batch of reports.
When to run it
- Every weekA regular pass on new reports keeps the backlog short and stops duplicates from piling up.
- After a releaseNew reports often cluster around a change. Triage shows quickly whether they are one bug or several.
- When the backlog has grown out of controlA full pass on a few hundred reports gives the team a clean starting point.
- Before planning a sprintA ranked list with reasons lets the team agree on what to fix in minutes.
Questions people ask
What is bug triage?
Bug triage is reviewing new bug reports to decide which are real, distinct problems, merging duplicates, setting aside what is not a bug and ranking the rest by when to fix them.
How do you prioritize bugs?
By impact and reach. Bugs that affect money, data, security or sign-in for many users, or that have no workaround, come first. Main features failing for some users come next. Cosmetic or rare problems come last.
What is the difference between severity and priority?
Severity describes how bad the failure is. Priority describes how soon it should be fixed, taking into account how many users it reaches and whether there is a workaround.
How often should you triage bugs?
Weekly suits most teams. A short regular pass keeps the backlog manageable, and a pass right after each release catches new problems while the change is fresh.
Does it change my issue tracker?
No. It prepares files only. It does not create, close, label or comment on anything in a live tracker and does not reply to reporters. You import the backlog file when you are ready.
What does it need from me?
An export of the bug reports, such as a CSV of open issues, a support inbox export or a folder of notes. A few facts about which product flows carry revenue help it set priorities. If no business context is given, priorities are set from the reports alone and the handoff says so.
More jobs in Developers
Turn bug reports into a clean backlog, on your own Mac.Free app. Runs on your Mac with the Claude or Codex plan you already have. Your files stay on your computer.
Download for Mac