Every office has one: the fault that was “fixed” on Monday and is back on Thursday. The shared printer queue freezes, the booking screen shows yesterday’s data, the VPN drops at the same hour every afternoon. Each time, someone restarts something, the complaint stops and the team moves on. Three weeks later it is back. This article diagnoses why that pattern persists and sets out a precise, repeatable method to end it, drawing on the problem-handling side of ITOps as described by the Ops Atlas. It is written for coordinators, office managers and small technical teams across Chennai’s many service businesses, though nothing about it is local.
The symptom: fixing without learning
When a fault returns, the underlying cause has not been touched. Operations teams distinguish two activities that most offices blur together. Restoring service is what an incident process does: get the user working again. Finding and removing the reason it happened is the job of problem handling. The entry published at the Ops Atlas ITOps page lists both among the eight core practices and places the removal of repeat-failure causes at the centre of what the function is for. A team that does only the first will be busy forever.
Diagnose your own office with one question: if I asked for last quarter’s top three recurring faults, could anyone answer in under a minute? If not, you are fixing without learning, and the rest of this piece is for you.
Cause one: nothing records the repeats
You cannot see a pattern in conversations. If faults are reported by phone, chat and a tap on the shoulder, no one has the history. The remedy is mundane and effective: a single log, even a shared sheet, in which each fault has a date, a short description, who reported it, what was done and a category. After a month, sort by category. The repeats become visible without any analysis.
Be strict on one rule. The entry stresses that a ticketing product records the practice and does not replace it. A log full of vague lines such as “printer issue, fixed” will not reveal anything. Write what you actually did: “cleared stuck job, restarted spooler service”. Specifics are what let you spot a cause later.
Cause two: the fix addresses the symptom
Restarting a service is a legitimate incident response and a poor problem solution. To move from one to the other, ask “why” until the answer is something you can change. The printer queue freezes. Why? A large file blocks the spooler. Why does a large file block it? The print server has too little memory. Why? It was sized years ago for a smaller office. Now there is a change to make, and it is one you can plan, assess, approve and review.
Stop asking when you reach a cause that is both real and within your power to change. Do not stop at “user error” unless you have also asked why the system made the error easy.
Cause three: changes nobody connected to the symptom
A surprising share of recurring faults begin with a change: a software update, a new device, a modified firewall rule. If changes are not logged, the link is invisible. The framework summarised in the entry separates standard, normal and emergency changes and recommends one forward schedule of planned work, with conflict detection. A log of changes alongside the log of faults lets you overlay the two and ask “what changed the day before the trouble started?” Often the answer arrives within minutes.
The Atlas also records a cautionary pattern from delivery research: organisations that respond to every failure by piling on more process tend to make things slower without making them safer. The better response is to review each failed or emergency change and adjust the one control that would have caught it.
Cause four: the inventory is wrong
Diagnosing a fault on a device you did not know existed is difficult, and so is replacing a component whose model nobody recorded. In the entry’s sample health check of ten records, even a small list produced stale entries, an orphan with no owner and a duplicate serial number. Inaccurate records send people on goose chases. Fix this by sampling ten items a month and measuring the mismatch. The inventory standards mentioned in the entry put the asset list first among security controls for a reason: you cannot protect or repair what you do not know about.
Cause five: the clock hides the pattern
Teams that measure only “was it fixed within target?” miss the repeat. A fault resolved inside its time limit three times in a row meets every target and still wastes a week. This is why the entry insists that service-level reporting be understood properly, including which hours count. Pair every target with a repeat count. If a category appears more than three times in a month, it earns a problem record, regardless of how neatly each instance was resolved.
A method: the five-line problem record
Use this template for each recurring fault you decide to eliminate.
- Statement: one sentence describing the recurring symptom and its count.
- Evidence: the incident references and any changes logged near each occurrence.
- Cause: the deepest cause you can verify, with how you verified it.
- Remedy: the change required, categorised as standard, normal or emergency, with its risk score.
- Check: how you will know it worked, and the date you will look.
Keep it to a single page. Review open records for thirty minutes each fortnight. Close a record only after the check date has passed without the symptom returning.
A generic illustration of the method
The following is an invented example, not a report of any real office. A small accounts team notices that its shared spreadsheet is slow every Tuesday afternoon. The log shows five such reports in six weeks, each closed with “restarted the file service”. Overlaying the change log reveals that a backup job was moved to Tuesday afternoons during the same period. The team writes a problem record: statement, five dated incidents, the backup move as the probable cause, and a verification step of running the job once at lunchtime on a Tuesday while watching response times. The test confirms the link. The remedy is a normal change moving the backup to an evening slot, scored low risk, reviewed by a colleague and added to the shared calendar. The check date arrives; four Tuesdays pass without a complaint, and the record closes.
Notice what made that work: nothing sophisticated, only three logs that could be compared and a willingness to test the hypothesis before acting. The same steps apply to an unreliable lift in a building, a payment terminal that drops out or a website that slows at month end.
What about stopgaps?
Sometimes the real fix is expensive or must wait for a budget. In that case record a workaround, label it clearly as temporary and put the permanent change on the schedule. A workaround with no owner and no date becomes permanent by neglect. The entry’s guidance on knowledge notes inside the tracking tool applies here: write the workaround where the next person will find it when the symptom appears.
Measuring whether you are getting better
Three numbers are enough. First, the count of distinct recurring faults open. Second, the number of incidents per month attributable to those faults. Third, the share of problem records closed with a verified check. Do not pursue a perfect score; look for the trend over a quarter. A small office that removes one repeat cause a month will have removed a dozen by this time next year, and the noise level in the office will have dropped accordingly.
When the problem is bigger than your team
Occasionally a repeat fault traces back to a supplier, such as an internet carrier or a software vendor. That is where the entry’s arithmetic about agreements with outside parties matters: your promise to users has to leave room for theirs. Keep the evidence in the record, escalate with dates and counts, and note the effect on your own commitments. Vendors respond far better to a table of dated occurrences than to a phone call about a general feeling that “it keeps happening”.
For readers building their careers around this kind of work, the overview of a practical route through skills, tools and growth in DevOps sits next to this topic, and if the diagnosis leads to an offsite planning session, our guide to a first trip to Chennai may help with the logistics. The method itself is simple: log, correlate, find the cause, change one thing, check. Keep doing it and the Thursday fault stops being a tradition.