Quick answer
Troubleshooting is a systematic search for the cause of a problem, followed by fixing it and confirming the fix. It works backwards from a symptom: define it, reproduce it, narrow down where it is, test a cause, fix and verify, then record what was found.
What troubleshooting means
Troubleshooting is a systematic search for the cause of a problem, followed by fixing it and confirming the fix. It starts from a symptom (something does not behave the way it should) and works backwards to a cause, one tested guess at a time. The word is most often used for machines, software, networks and support work, but the method is the same anywhere: say what is wrong, narrow down where, test a cause, fix it, check it, write it down.
It is not the same as guessing and changing things until the symptom goes away. A change that hides a symptom without a known cause is a workaround, and the problem usually comes back.
| Term | What it is | How it differs from troubleshooting |
|---|---|---|
| Troubleshooting | Finding why something is not working, then fixing it | The whole process, from symptom to confirmed fix |
| Debugging | Finding and removing a defect in code | One kind of troubleshooting, limited to software defects |
| Root cause analysis | Asking why until the underlying cause is found, often after the fact | A deeper step; troubleshooting may stop at the fix |
| Trouble ticket | The record of a problem and its handling | The paper trail, not the activity |
The six steps
Google's site reliability engineering book describes troubleshooting as repeatedly moving between observations and hypotheses, and breaks the work into stages from problem report to cure. The six-step version below follows the same logic in plainer words.
| Step | Question it answers | What to do | Output |
|---|---|---|---|
| 1. Define the symptom | What exactly is wrong? | Write expected behaviour, actual behaviour, when it started and who is affected | One sentence anyone can test |
| 2. Reproduce | Can I make it happen on demand? | Find the shortest set of actions that triggers it | A repeatable test, or an honest "not reproducible" |
| 3. Narrow down | Where is it? | Split the system in halves, test each, repeat; look at what changed recently | A small area to examine |
| 4. Test a cause | Why is it happening? | Pick the most likely cause, change one thing, observe | A confirmed or rejected cause |
| 5. Fix and verify | Is it really gone? | Apply the fix, repeat the step 2 test, check nearby things | Evidence the symptom is gone |
| 6. Record | Will we know next time? | Note the symptom, cause, fix and how it was found | A searchable entry |
Two rules run through every step. Change one thing at a time, or you cannot tell which change mattered. Check what changed first: a release, a settings change or a new integration just before the symptom appeared is the most common cause, though timing alone does not prove it.
A good symptom statement in step 1 has four parts: what you expected, what happened instead, when it started and who or what is affected. "Checkout is broken" is not testable. "Since Tuesday morning, 1 in 3 iPhone users who tap Pay see no confirmation page; web checkout works" is, because anyone can try it and say whether it still happens. Write it before you touch anything, and you can reuse it as the test in step 5.
A worked example: conversions dropped 32%
Sample data: a store sees purchases fall from 1,250 a day last week to 850 a day this week.
Step 1, define. Expected: about 1,250 purchase events a day. Actual: 850. Started Tuesday. The drop is 1 − 850 ÷ 1,250 = 32%.
Step 2, reproduce. A test purchase on the website is recorded. A test purchase in the iPhone app is not.
Step 3, narrow down. Split by platform, before and after:
| Platform | Last week | This week | Change |
|---|---|---|---|
| Web | 700 | 690 | −1% |
| iOS app | 400 | 20 | −95% |
| Android app | 150 | 140 | −7% |
| Total | 1,250 | 850 | −32% |
Check: 700 + 400 + 150 = 1,250 and 690 + 20 + 140 = 850. The sales fall is almost all in one platform. The store's iOS release 4.2.0 went out on Tuesday, so that is the recent change to examine.
Step 4, test a cause. Hypothesis: the new build sends the event under a different name. The event list shows a new name, order_completed, with 380 hits this week and the old purchase nearly gone (380 + 20 = 400, the same as last week's iOS purchases). Cause confirmed: a renamed event.
Step 5, fix and verify. Ship 4.2.1 with the original name. Repeat the test purchase and check that the count for purchase rises on iOS. Events already sent under the wrong name stay under it, so note the gap in the report.
Step 6, record. "4.2.0 renamed purchase to order_completed; 4.2.1 restored it; add an event-name check to release testing." The next person who sees a platform-only drop has a place to start.
This is not a sales drop and the marketing budget did not need to change. The troubleshooting prevented a wrong conclusion.
Common troubleshooting mistakes
- Fixing before defining. If the symptom is "it's broken", nobody can tell when it is fixed. Write expected and actual behaviour first.
- Changing several things at once. If you restart the server, clear the cache and update the app together, you cannot say which one worked.
- Trusting the recent change without testing it. A release shipped the same day is a lead, not proof.
- Stopping at the workaround. If you added a retry and the failure went away, you still do not know why it failed.
- Skipping the record. The same symptom will return. Teams that write down the cause turn a one-off fix into a runbook. For a related pattern with broken links, see how to fix 404 errors.
- Never escalating. Set a time box before you start, for example 30 minutes for a first pass. If you have not narrowed the area by then, hand over what you have: the symptom statement, what you tested and what you ruled out. Passing on a clean record saves the next person from repeating your steps.
- Skipping the user's side. The customer's report is the first piece of evidence. Ask what they expected, what they saw and how to repeat it.
Troubleshooting data and support problems in TapCub
Take the iOS example above and follow it in TapCub.
- See the symptom early. Add a monitor in All-platform analytics › Operations › Monitors on the purchase event count, so a drop reaches you by email or chat tool instead of a weekly report. A monitor only notifies; it does not diagnose or fix anything.
- Narrow down by platform. In All-platform analytics › Behavior › Event analysis, group the purchase event by platform and compare the two weeks, as in the table.
- Look for new names. New event names land in All-platform analytics › Data › Event inbox before they join the event definitions, so a renamed event shows up as a new row.
- Verify after the fix. In All-platform analytics › Data › Tracking plan, turn on Live verification (24 h) to see hits in the last 24 hours and completeness of required properties for each planned event.
- For live chat problems, the help page on a missing bubble lists what to check first (switch, allowed domain, mobile hiding). See chat bubble not showing.
TapCub shows the data and the checks; it does not decide the cause. The reasoning in steps 1 to 6 stays with you.
Sources
- Google, Site Reliability Engineering, Effective Troubleshooting.
- Wikipedia, Troubleshooting.
Was this explanation clear?
We use your answers to fix and expand this page.
Couldn’t send. Please try again in a minute.
Thanks for the feedback. We’ll use it to improve this page.You’ve already sent feedback on this page. Thanks.

Product Analyst
Focuses on event analysis, conversion funnels, and retention, using clear metric definitions and structured analysis to explain product performance.