Evidence your provider cannot wave away
"My internet is slow in the evenings" is a feeling, and support is trained to answer feelings with a modem reboot. A number with a place and a time attached is a different conversation: packet loss starts at your third hop every night after nine, here are the runs. Three measurements do that job — where the loss begins, how latency behaves under load, and how far and how unstable the servers you actually use are. Every one of them exports as text or PDF, and NetDiag+ can also put your daytime and evening Health Score checks into one report for your provider (the report itself is in English).
Three measurements that name a place and a time
- Locate the loss with MTR. A traceroute that keeps pinging every hop shows where packets start disappearing. Loss that begins at one hop and persists to the very end is real, and that hop shows which network it starts in. Loss at a single hop that vanishes at the next is that router rate-limiting its own replies — normal, not evidence. Run it at the bad hour and note the hop.
- Add latency under load and the latency table. A bufferbloat test tells you how much latency rises while the line is busy and in which direction: an upload spike points at your own router, a download spike at the provider's equipment — and that direction is what tells you where the problem is. The Latency Dashboard adds median, jitter and loss to the named servers you use, with your own router measured first, so the report also shows whether your Wi-Fi was clean.
- Export at the right hours, more than once. Every tool has Share → text or PDF report: targets, timestamps, every number. One run proves nothing; three at the bad hour on different days plus one at a quiet hour as a control is a pattern. Health Score checks from the daytime and the evening go into one report for your provider, with the two averages side by side. The bufferbloat trend chart keeps the worst spike of every run over 7, 30 or 90 days, so the pattern is visible without you keeping notes. History keeps everything you ran.
Rule out your own Wi-Fi before you point at the line
The most common reason a ticket bounces back is that the problem was never on the provider's side. The phone running the test measures the delay on its own Wi-Fi link, so a busy 2.4 GHz channel or a weak signal can look like a fault on the line — even when every other device in the home is fine.
Separate the two before you write. The test shows latency to your own router first. If that stays low under load, your Wi-Fi is clean, and whatever rose after it is on the router or the line. If the latency to the router itself climbs, move closer, switch to 5 GHz, or test once over a cable — then measure again.
- Latency to the router stays low under load — the Wi-Fi is clean; anything that rose is past it.
- Latency to the router climbs under load — the rise is on the Wi-Fi between the phone and the router, measured on that phone; other devices may not feel it. Fix the Wi-Fi first, then re-test.
- A wired test still shows the problem — now it is the router or the line, and the ticket will hold.
What to paste into the ticket
Support acts on numbers with a place and a time, not on adjectives. Keep it to what a technician can check, in plain text:
- When. The date and clock time of each run, and your time zone.
- Where the loss starts. The hop number from MTR and its loss percent, run at the bad hour.
- Latency idle vs under load. The idle figure, then the figure under download and under upload, with the direction.
- The servers you actually use. Median and jitter to two or three of them, with your own router measured first — a clean router shows the rise is not your Wi-Fi.
- The control run. One run at a quiet hour that comes back clean, so the bad-hour runs mean something.
- Everything as text, pasted, not screenshotted, so it can be forwarded to the engineer who fixes it.
One report instead of six screenshots
NetDiag+ puts the daytime and evening Health Score checks into a single document for your provider, with the two averages side by side and the raw numbers behind each one. It is in English, and it stays on your phone until you share it. It reads as a measurement, not an accusation: it names where the delay builds up, not who is at fault — which is what keeps a support desk reading.
Export it as text for a ticket or a regulator's web form, or as a PDF to attach. One file carries the whole pattern, so the first-line agent can forward it without re-typing anything.
Free · 31 tools on iPhone, 32 on Android
Questions people actually ask
What should the ticket actually contain?
Dates and times of each run. The hop where loss starts and its percentage. Idle latency versus latency under download and under upload. Median and jitter to two or three servers your household uses. Which test was the control. Everything pasted as text, not described — a technician can act on a hop list, not on "it lags".
Support says the line tests fine. Are they lying?
Usually not. Their tests check signal levels, error counters and throughput, and all of those can be perfect while a queue adds 300 ms or a congested peering link drops 2% of packets at 9 pm. You are measuring something they do not measure. Say so, and hand them the numbers.
Screenshots or text?
Text. It can be searched, quoted, pasted into an internal tool and forwarded to the engineer who actually fixes things; a screenshot cannot. Text also survives a forum's image compression and a regulator's web form. Keep the screenshots for yourself.
How many runs make a case?
Enough to show the problem is repeatable and time-bound. Three runs at the bad hour across two or three days, plus one at a good hour that comes back clean, is the minimum that turns a complaint into a report. If the good-hour run is also bad, the fault is not congestion — that is useful too.
Will this actually get something fixed?
It moves the ticket from first line to someone who can read a hop list, which is the step that usually stalls. Where a regulator exists — BTK in Turkey, the national telecom authority in most of Europe — the complaint forms ask for exactly this: dates, measurements, what the provider replied. Evidence does not guarantee a fix; it removes the reasons for saying no.
The single strongest number in any of this is latency under load. Next: which server is far, and which is unstable →