In 56% of our callouts, the fault was not in the customer’s cabling

Why we bothered doing this

If you have ever rung your provider about a dropout, you know the routine. Restart the modem. Try a different cable. Plug into the first socket. Are you sure it isn’t your wi-fi.

Then eventually somebody tells you the problem is your internal wiring, and that it is your job to sort out.

We have spent about 30 years turning up after that phone call. Since 2015 we have filmed the jobs, so there is a record of not just what the fault was but how we found it: the symptom, what we tested, what we ruled out, and where it actually turned out to be.

The trouble with an archive like that is it stays anecdotal. You watch one video and you learn one fault. Nobody, including us, could tell you what the 324 of them said together.

So we built the AI brain.

What the AI brain actually is

It is not a chatbot, and it is not a black box that spat out an opinion. It is a knowledge base we own, made of plain text files, that an AI can read and work on.

Here is the whole pipeline. It is deliberately boring, because boring is what makes the numbers trustworthy.

Transcribe, structure, read, tabulate, count, verify. The point of working that way is that we can be told we are wrong. Every number here traces back to a specific list of jobs, and every one of those jobs is a video you can go and watch.

Watch the case videos on our YouTube channel

What actually happens at the end of a job

Across 324 cases the brain identified 56 distinct recurring fault patterns. Here is what they add up to, starting with the number that surprised us, and we are the ones who did the work.

More than half the time, a successful job does not end with us repairing anything. That sounds like a failure. It is not. It is the job working exactly as it should.

There is a boundary between your property and the carrier’s network. We are licensed to work on your side of it. We are not permitted to repair the network side, and no private technician is. So when the fault sits out there, the only useful thing anybody can do is prove precisely where it is, in a form the retailer cannot brush off, and force the people who are allowed to fix it to come and do so.

In 28.7% of all cases that proof was the entire outcome. We could demonstrate the premises were clean even though the carrier’s fault sat somewhere we had no access to.

Where along the line the fault was

Your service runs from the exchange, through a pit in the street, along a lead-in cable, across the boundary into your property, sometimes through a building frame, then through your own cabling to the device. Here is how often a fault was isolated to each point.

The lead-in cable, the run from the street to the wall, is the single most common place a fault was isolated to. It is the carrier’s asset, and in a lot of Sydney it has been in the ground for 50 or 60 years.

You have probably already been told it is your fault

118 cases, 36.4%, record that the customer had already been told the problem was on their side, in their internal wiring, or not the network’s responsibility, before we attended.

Worth being honest about that one. We do not narrate the back story on every job, so the real figure is almost certainly higher. We just will not claim what we cannot count.

The finding we did not expect: intermittent faults are where people get stuck

Of the 324 cases, 125 (38.6%) were constant faults and 87 (26.9%) were intermittent. Then we looked at how many technicians had already been through before us.

That gap is not about anyone being incompetent. It is structural, and once you see it the whole industry makes more sense.

An intermittent fault has to be present to be found. A carrier technician who arrives on a dry Tuesday, tests a line that is behaving perfectly, gets a pass result and closes the job has not done anything wrong by their own process. They have proved the line was fine at 11am on Tuesday.

Meanwhile the fault comes back on Thursday night, and the customer starts again from the beginning with a new fault reference and a new technician who knows none of the history.

Which is why our approach to an intermittent fault is not to test it and hope. It is to reproduce it deliberately, under load, on a continuous record, so it cannot hide.

Overall, 120 cases (37.0%) had at least one technician attend before us and fail to resolve it.

So where was the fault, then

Grouped into families. All percentages out of 324.

Every item on that list is something a support script will spend an hour asking you about. Together they caused 5.6% of the faults we documented.

We still check all of it, and we check it first, because it takes five minutes and costs nothing. Laptop plugged straight into the network device. A known-good lead. A spare router. Wired against wi-fi. And we pull any mobile backup SIM, because a modem quietly falling back to 4G will hide the real failure from you completely.

One thing worth knowing if you are troubleshooting your own gear: routers fail partially. We have documented wi-fi working perfectly while the ethernet ports on the back were dying. If your wireless is fine and your desk is not, do not rule out the router just because it is still lighting up.

In a house the fault is usually outside. In a unit it is usually in the comms room.

Houses made up 194 of the 324 cases (59.9%), apartments and strata 67 (20.7%), business and commercial 36 (11.1%). The fault profile is genuinely different between them, and it is worth knowing which one you are before you start.

If you live in a unit and your service has never worked properly since you moved in, the odds are strongly on the frame downstairs.

Which of these can even happen to you

This part gets skipped everywhere and it matters, because a lot of these faults are physically impossible on some connection types. Where your fibre stops is what decides which faults can reach you.

FTTN and FTTC: fibre stops in the street

From the node or the kerb, your service runs over the old Telstra copper, through a pit, along a lead-in cable, and into your house. The NBN never replaced any of that. It just changed what runs over it. So the full list is available to you: bridge taps, stray voltage, wet pit joints, damaged lead-ins, noise, breaks and shorts, plus anything wrong inside the house.

FTTC is where bridge taps concentrate. Of our 60 FTTC cases, 21.7% involved one.

FTTB: fibre stops in the basement

This one gets confused constantly. In a fibre-to-the-basement building the fibre runs all the way into the building’s comms room and terminates there. From that point down to your unit it is the building’s own internal copper risers.

That means the street copper faults cannot apply to you. No lead-in cable in your path, no pit, no gel joint, none of the CAN 2000 legacy described below.

The data backs that up cleanly. Not one of our 25 FTTB cases involved a lead-in or a pit fault. What they involved instead was bridge taps in 32% of FTTB cases, missing frame jumpers in 24%, and the documented path not being the real path in 20%. Exactly the profile fibre to the basement should produce.

FTTP: no copper in the path at all

Fibre runs to a box on your wall. None of the copper faults in this article can reach you.

HFC: coaxial cable, with its own failure modes

The copper phone network is not in your path either, but coax has its own list, and it is mostly about connections rather than cable. Of our 35 HFC cases, 8 of them, 22.9%, came down to a coax fitting, splitter or connection device on the customer’s side of the tap:

• Loose, hanging, disconnected or redundant coax connections, the most common of the group
• Old or legacy splitters left in the path from a previous pay TV or cable setup
• Corroded compression ends, particularly anywhere exposed to weather
• Faulty or ageing isolation devices
• The HFC connection device itself failing

Coax is far more sensitive to a poor physical connection than most people expect. A single corroded end will produce exactly the same intermittent, load-dependent dropout you would get from a copper fault, and it will pass a casual visual inspection. We have found a splitter with the cores twisted together hidden in a wall cavity behind an alarm panel.

Everyone, regardless of technology

A faulty router or lead, NBN equipment installed in the wrong room, a cable cut during renovations, sockets left dead after the cutover, and a service that syncs but passes no traffic.

Bridge taps: the most common single fault we find

28 cases, 8.6% of everything. The largest specific named fault in the archive, and the leading fault on both FTTC and FTTB.

In plain English

Somewhere in your place there is a length of phone cable that goes nowhere useful. An old socket in a bedroom nobody uses. A second wire left connected when something was changed years ago. An extension run to the kitchen for a phone that has been in a drawer since 2011.

On an old landline, that dead leg did nothing at all. Everything worked.

On a modern NBN service running over copper, it is a genuine problem. The signal travels down that dead leg, hits the end, and bounces back into the line it came from. That reflection interferes with the real signal. It hurts the high frequencies worst, and the high frequencies are exactly where your speed lives.

How you know you have one

• The line looks perfectly fine when you are not using it
• Put load on it, a big download or a stream, and it drops
• Often worse at night, or after rain
• Remove the extra legs so there is one clean path, and it settles down

For the technicians reading

The bit we do differently is the proof, not the fix. Anyone can pull a bridge tap. The problem is that “it seems better now” is worthless to a customer three weeks later when it comes back.

1. Start PingPlotter and leave it running on one continuous trace throughout
2. Apply load deliberately and capture the failure
3. Remove the bridge tap
4. Apply the identical load again, same trace, same conditions
5. Compare

One record, before and after, same instrument, same load. That is not an opinion, and a retailer cannot argue with it.

More on how Australian homes came to be wired this way

Lead-ins and pit joints: the faults that follow the weather

53 cases, 16.4%. The lead-in is the cable running from the street to your house. In a lot of Sydney it has been in the ground for 50 or 60 years.

22 cases, 6.8%, explicitly tie the fault to weather, rain in 16 of them. That is the count where the video says so outright. Corrosion and water damage on their own were not counted, so the true number is higher.

The giveaways repeat so reliably they are almost a checklist:

• Dropouts or crackling after rain
• Faults when it is windy
• Speed dropping suddenly and staying down
• Hum, static, or a dead phone line
• It improves briefly right after somebody disturbs a joint, then goes again
• Several technicians have already been out without ever opening the right pit

That last one is the intermittent fault problem from earlier, showing up in physical form.

The way you prove it is dull and effective. Separate the internal cabling from the lead-in, then run identical tests on both sides. Clean readings inside plus failing voltage or insulation readings outside is a much stronger case than one odd number on its own.

In apartment buildings, the paperwork is usually wrong

49 cases, 15.1%. If you live in a unit, your service runs from a main frame in the basement, sometimes through further frames on each floor, and then to your wall. Every one of those points is a place where a wire is joined to another wire by hand, usually years ago, often by someone in a hurry.

What we find: double jumpers, pairs daisy-chained between units, pairs quietly borrowed for someone else’s connection, unit tags that are wrong or missing entirely, punch-down contacts damaged by a screwdriver, and retailer paperwork describing a pair that matches nothing on the frame at all.

A tone is a signal you inject at one end and listen for at the other. It tells you where the wire actually goes. The label tells you where somebody once intended it to go.

So we tone from the wall socket back toward the main frame, check every frame in between, and follow the real path one segment at a time. When the documentation and the tone disagree, the tone wins.

And more often than not the right answer is not to repair a chain of undocumented jumpers. It is to build one clean path on a verified spare pair and leave it better than we found it.

The best first question is not about the symptom

33 cases, 10.2%, came down to something that had been done to the building, not a component wearing out.

So the question that narrows a fault down fastest is almost never what it is doing. It is: what changed just before this started?

A renovation. Digging in the street or the yard. Switching providers. The NBN cutover itself. A previous technician’s visit. Any one of those will cut the search area down faster than an hour of testing will.

If you are booking a technician, have that answer ready. It genuinely saves you money.

Most jobs have more than one thing wrong

117 cases, 36.1%, carried two or more fault patterns. Setting aside the general proven-network-side tag and counting only specific named faults, 57 cases, 17.6%, still had two or more.

A bridge tap and a bad termination. An internal bridge tap and a wet pit joint. A faulty router and a completely unrelated frame problem. A corroded socket found and repaired, with a lead-in fault still sitting underneath it that had nothing to do with the socket.

This is why finding something wrong is not the same as finding the cause, and it is why we do not stop at the first defect. We document it, fix it if that is the right call, then run the original test again to see whether it actually changed anything. If the fault is still there, we keep going.

It is also why a job that was fixed last month can come back. The first fault was real. It just was not the only one.

The method, in ten lines

1. Start with the symptom, not the retailer’s diagnosis
2. Use history and weather correlation to pick the likely area
3. Simplify the circuit down to one clean path
4. Change one variable at a time
5. Substitute known-good leads, sockets, modems and devices
6. Tone and test every physical segment rather than trusting records
7. Reproduce intermittent faults under controlled load
8. Test the premises side and the carrier side with the same instrument
9. Retest after every single repair
10. Document enough evidence that escalation is actually actionable

What this data is, and what it is not

The numbers only mean something if you know what they describe, so here it is plainly.

The set

324 attendances we filmed and documented, from 16 January 2015 to 25 August 2026.

How it was counted

Every case was read in full and tagged into a table with one row per job. Every figure on this page was then computed from that table by script, not estimated by reading. A sample of 25 cases was independently re-tagged by a second pass working blind, with 96.7% agreement across 150 field decisions. The five disagreements were all borderline judgement calls.

What it is not

It is not every job we have done in 30 years, and it is not a random sample of Australian internet faults. It is weighted toward the hard ones. A job gets filmed when it is worth filming, which usually means the customer had already been told the problem was theirs, and it was not.

That is the population these findings describe. If you have been told your wiring is at fault and you do not believe it, this data is about you specifically. If your internet has been flawless for six years, it is not.

Where it is thin

A third of cases, 32.4%, do not record the access technology. Around 63% do not record whether the customer had been blamed beforehand, so that figure is a floor. Pattern counts overlap, because one job can contain several faults. Anything in the archive at four cases or fewer has been kept out of the headline numbers rather than dressed up as a finding.