Broadband repair request speed test: the 900 technician refuses to accept it, only the business hall app counts

I only figured this out
after a speedometer screenshot caused me to get hurt. Last autumn, my broadband dropped to about 200 Mbps at night. I tested it three times with Speedtest, took screenshots and saved a photo album, called for repair, and the operator said: Please use our service hall app to test it. I got famous right then—I had a gigabit plan, an internationally renowned speed testing tool, a 900MB screenshot, and you wouldn't admit it?
The result was a test on the service hall app, 958. The numbers are pretty good. The operator said your network is working fine and there are no faults. After hanging up, I stared at the two numbers for a long time: the same router, the same network cable, within five minutes, one 900 and the other 958—which one is real?
Later, after getting to know the master who came to the door and looking through a pile of materials, he finally figured it out. It's not that one is fooling the other; it's that the rulers on both sides measure not the same stretch of road.
First, let's figure out one thing: the 900 you tested and the driver's 958 are half the distance apart
The bandwidth speed test numbers essentially refer to "the actual test of the journey to the speed test server." Note, it's this entire stretch, not the broadband itself. This section of road is split into three sections: your home WiFi or network cable, the access segment to the carrier's data center, and the backbone segment from the data center to the speed test server. Any one of the three segments gets blocked, and the numbers drop, but the reasons are completely different—you can fix your own household issues, and the last two segments are the operator's fault.
Public
speed testing tools like Speedtest assign you "servers" that are speed testing nodes deployed in each company's data center. It selects the one with the lowest latency by default, but you need to catch the automatic exhaust distribution to point you to neighboring provinces. If you take a few hundred extra kilometers on the road, it's normal for gigabit broadband to reach 700 or 800 km. You'd think something's wrong at home.
The business hall app is different. It measures data from your local operator's internal speed server, with the shortest route and highest priority, equivalent to the "half-journey" score. This number naturally looks good, but it's a standard used by operators themselves — when requesting repairs, technicians and the work order system only recognize it.
So that "900 vs 958" race was not entirely contradictory: I was testing the full marathon, while it was the first half. If you use your full marathon results to request repairs, the system won't match the accounts, and the work orders won't be issued. I also mentioned the full account of this in my article >
Only by tightening the screw at the node can your numbers be used to talk about it
Since we recognize the number of business halls, is it enough to just install this one? No, it is not allowed. It only measures half the road. If your router at home breaks or the network cable is 100Mbps, it still charges you 958—the half route doesn't even include the part at home where you are. So before reporting for repairs, you have to walk on two legs. My approach is fixed in three steps.
Step one: The business hall app tests first, which is the "official report card" for the work order system to see. Step two: also use public tools like Speedtest, but nodes must be manually selected: choose your city, the one in parentheses that matches your broadband operator, for example, for Telecom broadband, select "XX-Telecom." For automatically assigned nodes, first check their home location, then replace those that cross provinces. This step measured the total road score, compared to the number at the service hall. The difference in the middle is the 'world after leaving the operator' world. Step three: Connect the network cable directly to the router's LAN port and test once. Don't use WiFi—WiFi tests the wireless segment, which is meaningless in repair scenarios. That amount The article on why gigabit WiFi only has 600 is very detailed.
Save both numbers well. If the service hall app only shows two or three hundred, then the real proof is that it's connected to the segment or bureau, so you have plenty of confidence to request repairs; If it's a 958 or public tool for over 300 yuan, there's an 80% chance the problem is at the key exit, or at your home. Keep investigating and don't rush to call.
Slow times should be regular; one time slot chart works better than ten screenshots
The biggest taboo when reporting for repairs is the phrase 'It's very slow at night.' Operator registration is vague, and work orders are categorized as 'slow' speed, with priority placed after network disconnection. Orders are dispatched slowly, and technicians don't do their homework before arriving. I learned my lesson the second time and spent three days making a watch before the speed test.
| Time: | Servicehall app | Speedtest (same provincial node) | direct network cable connection |
|---|---|---|---|
| weekdays 10:00 | ,961 | 933 | 938 |
| weekdays 21:30 | 954 | 219 | 221 |
| Weekends 15:00 | 957 | 486 | 472 |
| Weekends 22:00 | 952 | 173 | 168 |
(Unit: Mbps, actual data from my home.) The first repair request only had the Speedtest 209 at night, which was basically nothing. )
What makes this watch so good? It draws the "shape of the fault." Three patterns are immediately clear: First, the service hall app stays stable around 950 throughout, indicating that the access route to the data center is good; Second, direct network cable connections and WiFi drops are just as bad, which means it's not a wireless issue; Third, the drop is worse during peak periods, which is a typical pattern of evening rush hour congestion at exits in the area. The operator saw the form and immediately changed the work order from "slow" to "congestion check," leaving the line "the form is properly made"—her exact words.
There are three rules for making watches. Measuring at the same location, different from today's living room and tomorrow's study; Each test takes the median three times, and the single test number doesn't count. If the speed test server gets busy, the same session can differ by 30 to 40 trillion; Time periods must cover both "good times" and "bad times." Just having the bad half sheet cannot prove that the problem changes over time. Just the table with evening rush hour data will only reply, "That's how it is during evening rush hour."
On the day the technician arrived, his appearance and your app each checked separately
The technician brought an optical power meter, which measured another ledger: how many dBm of light signal degradation from the data center to your home. Your app can't measure this account at all. So don't point out speed-testing screenshots on your phone at the driver—you can't help with what they need to see, and they won't ask what you should have prepared in advance—but if you do, your work will be done twice as fast.
Prepare two items in advance. First, copy down the light degradation readings in the backend of the optical modem: enter 192.168.1.1 in your browser. For telecom devices, usually find the 'Optical Power' or 'Receiving Light Power' section on the 'Status' or 'Network' page. If it's negative, it's healthy if it's below -27; if it's above -28, you should be cautious. I wrote a complete table about the meaning and classification of these numbers in the article light decay; just follow the instructions. Just copy it down first, and the technician will test it with the instrument for comparison. Whether it's a wire issue or a splitter, you can tell on the spot.
Second, note down the
work order number in the repair phone number. This is the account the technician visits to verify the account. Afterwards, they need to rush orders, rework, and if the same issue fails within 30 days, they can handle it for free—all thanks to it. How to proceed and how to say the procedure, The repair process article is more detailed, so I won't repeat it here.
There's another thing worth doing that day: once the technician finishes the repair, don't rush to sign off. Use the business hall app to test it on the spot, then retest the "bad times" condition on your schedule—for example, if he came during the evening rush hour, test again during the evening rush hour, and only when the numbers come back is it truly repaired. My first request was because I signed too quickly, and that night it dropped back to 200 MB. The second work order took another day and a half in line.
After repairing, save the data as a file, saving half the argument when reporting next time
I currently have a folder on my computer called "Broadband Files." It contains three items: a photo of the contract page for the broadband installation (package and speed in black and white), the work order number and results for each repair request, and a benchmark speed test screenshot of the normal home network hours. Normally, it doesn't seem useful, but when something happens, it really saves the trouble.
Benchmark speed test screenshots are especially valuable. When the home network is stable, I pick a workday morning to test and save the store app and Speedtest once each. If the internet is slow any day, I'll retest it and compare with the benchmark: if 20% is lost, it's the equipment at home; if 80% drops, then report it for repair. This step can block most of the "fake faults"—my mother-in-law called to say the network was broken. I dug up a pair of benchmarks: her grandson was streaming 4K in the living room, and the uplink was all consumed. Every downloaded package had to wait for confirmation from the uplink direction. I also suffered losses when working from home—the uplink got snatched and the downlink collapsed too. The accounts were in The speed test article had algorithms.
Here's another easily misunderstood point: the "standard" broadbands for 2000M, 1000M, and 500M are not the same. Gigabit wired is fully charged around 930, don't chase after 990; There are many 2000M checkpoints: the optical modem, cable level, and computer network card are all blocked one by one. My device measured 943MB from 2000M, and after some search, it turned out to be about the network port specifications. The checklist for all four checkpoints is recorded in the 2000M speed test article. If you take a screenshot of 940 to report 'gigabit not meeting the standard,' the technician will just think you're making things difficult for him.
Finish by sticking the order in place: first self-check, then leave evidence, then make a phone call
The
whole process runs smoothly in just three sentences. Before making a call, test the network cable directly connected to the router and clean up the section at home—the optical modem is plugged incorrectly, the network cable is 100Mbps, the router port is 100Mbps, these are all our own pitfalls. < article listing four places where speed is stolen > a href="https://www.shunot.com/en/luyou/762.html" one by one, all of which are free actions. Then spend three days making a time slot list, using the business hall app and public tools as a dual caliber, covering both the good and bad time slots. After making the call, they report the package rate, schedule schedule, and work order number. Before the technician arrives, they prepare screenshots of the light decay readings and reference points.
If the order is reversed, it's wasted: the first time I requested a repair, I just took a screenshot and typed it, the work order got stuck on 'normal on the user side,' and waited two days with no response. The second time, the meter was handed over, and the order was assigned that day. On the third day, the technician brought the instrument to the site and found that one port on the corridor splitter was oxidized, replaced in ten minutes. Before leaving, he told me the truth: I took a screenshot of your report and sent it to the backend. We're willing to take this kind of work order—you can tell exactly where to check.
When it comes to speed testing, the numbers themselves aren't worth much; what's valuable are the origins and patterns of the numbers. Clearly explain the origin and draw out the pattern; only by reporting for repairs can you wrap things up in one go.
