ITOps Basics: Why the Same IT Problem Keeps Coming Back

ITOps Basics: Why the Same IT Problem Keeps Coming Back

Here is a scene you may know. Monday morning, the shared printer, the VPN or the invoicing app misbehaves. Someone restarts something, it works, and everybody moves on. Friday it happens again. Nobody can say exactly why, so the restart ritual becomes part of the office folklore. If that sounds familiar, the issue is not bad luck. It is a gap in what the IT world calls ITOps, and it is far easier to fix than most small teams assume.

Let us diagnose it the way a friendly mechanic would: symptoms first, then causes, then a treatment plan that fits a normal-sized business instead of a bank.

Symptom one: the same fault keeps coming back

When a glitch repeats, the temptation is to get faster at fixing it. Faster restarts, quicker replies, a shorter queue. That helps, but it never ends the loop, because speed is aimed at the wrong target.

ITOps work separates two jobs that feel identical in the moment. Restoring service is one: get people working again, even with a workaround. Finding out why it broke is another, and the vocabulary calls that second job problem management. The Ops Atlas lists it among eight core practices of the function, next to the service desk, incident handling and change control. The Atlas page on ITOps puts the removal of repeat-failure causes right in the definition, which tells you how central it is.

The treatment: keep a plain list of workarounds with dates. When one entry shows up three times, stop fixing and start investigating. Give the investigation an owner and a deadline that is separate from the daily queue. Fifteen minutes of questions on a quiet afternoon beats ten emergency restarts.

Symptom two: every request feels urgent

If everything is flagged as top priority, nothing is. Teams that live in that fog usually have no shared way to rank a ticket, so the loudest sender wins.

The cure is a tiny scoring rule. One widely used sample in a service tool combines how many people are affected with how quickly the damage grows, then subtracts one so the top level stays rare. You do not need that exact formula. You need any rule that two colleagues would apply the same way.

Then attach realistic clocks. The Atlas shows a lab setup in which the top priority gets four hours around the clock, the next gets eight business hours, then twenty-four, then forty, and the lowest has no target at all. Notice that the fifth tier carries no promise. Saying “we will get to this when we can” out loud is more honest than pretending everything has a deadline.

Symptom three: changes cause the outages

Many incidents trace back to something somebody changed last Thursday. The reflex reaction is to add approvals. Research cited in the Atlas page suggests that approval from outsiders does not lower the failure rate of changes, and that peer review works better. It also names adding process after each failure as a classic trap.

A calmer approach sorts changes into three lanes:

  1. Routine, low-risk, rehearsed changes get pre-approved, so nobody waits for a sign-off on something that has gone fine fifty times.
  2. Ordinary changes get a short look by a second person, with a written note on what could go wrong and how to undo it.
  3. Urgent changes, such as closing a hole that is being exploited right now, go ahead at once and are reviewed afterwards.

Add one shared calendar. Most self-inflicted outages come from two harmless-looking changes landing on the same day. A single forward schedule that flags clashes, plus a rule for blackout dates such as payroll week, removes a surprising share of the drama. The Atlas lab even walks through a payroll freeze in late October to show how a clash gets caught before it ships.

Symptom four: nobody knows what we own

Ask a small company how many laptops, servers and subscriptions it has, and you will hear a confident guess. Then comes the audit, and the guess turns out to be off. This is the configuration-records problem: the map of what exists has drifted from what is real.

Resist the urge to catalogue everything on day one. The Atlas warns against a record store that holds everything instead of what decisions need. Start with the things whose failure would stop work, put a named person beside every line, and diarise a recheck for any entry untouched in two months. In the sample the Atlas describes, a ten-item inventory looked mostly fine at first glance, yet two entries were stale, one had no owner and one reused a serial number. Small errors, big trust problem.

Symptom five: the quiet failure

The nastiest fault is the one you never hear about. An alert channel that silently stops, a backup that has not run for a month, a monitor pointed at the wrong address. The write-up calls out treating the alerting layer as something non-essential, because a silent failure there hides real outages behind a calm screen.

The fix costs almost nothing. Schedule a monthly “pretend it broke” test: trigger a harmless alert and see who receives it, how fast, and whether they know what to do. Treat the alert path like any other production service, with an owner and a place on the checklist.

Symptom six: the spreadsheet in the corner

Somewhere in your company is a spreadsheet that one person updates by hand, which everybody trusts more than the official system. These shadow lists appear when the official tool is slow, awkward or missing a field. They are a signal, not a crime.

Ask what the spreadsheet does that the main tool cannot, then either add that capability or promote the sheet to an owned, shared, dated record. Ignoring it leaves two versions of the truth and a very bad week when its keeper goes on holiday.

Do not over-treat

A fair warning before you buy anything. The same page lists heavyweight service management forced on one team that runs one service as a standard way to go wrong. If your whole operation fits around a single table, you do not need a board, a catalogue and a dozen forms. A shared inbox, a short priority rule, a change calendar and a monthly review will cover most of the ground.

It also helps to know that the ticketing tool is not the discipline. It records what people decide. Pick software after you have agreed the habits, not before, and be wary of anyone who claims a product will install the habits for you. Free and open options exist, and so do paid suites; the choice follows your size.

How to tell the treatment is working

Good habits show up in boring numbers. After a month, look at three things. First, how many tickets in your workaround list now point at a known cause with an owner. Second, how many changes went through the lanes without a surprise. Third, how long it takes a new colleague to find out who looks after a given service.

Be careful with percentages here. The Atlas includes a lab in which six of eight tickets met their targets, which sounds like a respectable seventy-five percent. Yet one of those tickets spent most of its life inside a weekend and a holiday that paused its clock, so the headline figure told only part of the story. Before celebrating or panicking about a score, read the calendar rules that sit underneath it.

Also keep your expectations kind. A team of five will not run like a large service organisation, and it should not try. What matters is that the same question gets the same answer whoever is on duty, and that the answer is written down where the next person can find it.

A two-week starter plan

  • Days 1 to 3: list your ten most important services and write one owner next to each.
  • Days 4 to 7: agree the priority rule and the three change lanes, and put them on one page.
  • Days 8 to 10: start the workaround list and the shared change calendar.
  • Days 11 to 14: run the alert-path test and review what you learned.

If you want to see how automation fits around this, our guide to automating routine business operations for smaller firms covers the workflow side, while a piece on streamlining business processes shows how process mapping can make the owner column easier to fill in.

The short version

Repeating problems are rarely mysterious. They come from skipping the second job, the investigation, and from having no shared rules for priority, change and ownership. Fix those four habits and the Monday-morning printer stops being a legend.