What to tell your ISP about slow internet: the call script that gets a technician

Most calls to an internet provider go the same way. You say the internet is slow. They ask you to reboot. They run a remote check, tell you everything "looks fine on our end," and offer to sell you a faster plan. Twenty minutes later nothing has changed.

The call goes differently when you arrive with three things: a wired test result with a packet-loss number, the signal readings from your own modem, and the right vocabulary. Support agents work from a flowchart. Specific technical language moves you to the branch of that flowchart where a technician gets scheduled.

This page gives you the evidence routine, the exact words to say, what the numbers mean for each kind of connection, and what to do if the first call fails.

Run the speed test first. The script depends on your numbers. Run the test on a device plugged in with an Ethernet cable if you possibly can, then come back. Run the speed test →

What this fixes

• You pay for one speed and consistently get far less, even on a wired connection.
• Your connection drops for a few seconds at a time, several times a day.
• Video calls freeze or games lag even though your speed test number looks fine.
• You have already called once and were told everything is within spec.

Step 1: Build your evidence (20 minutes, spread over a day)

The goal is to prove the problem is on the provider's side of the wall, not inside your house. Providers can charge for a technician visit if the fault turns out to be your equipment or your Wi-Fi, so this step also protects your wallet.

• Go wired. Plug a laptop or desktop directly into your modem or gateway with an Ethernet cable. Wi-Fi results are the first thing an agent will dismiss, and often rightly so.
• Run the test three times at different hours. One in the morning, one mid-afternoon, one between 7 and 10 p.m. when your neighborhood is busiest. For each run, write down download, upload, idle latency, latency under load, the bufferbloat grade, and packet loss.
• Note the pattern. Slow all day with packet loss means a line or signal problem. Fine by day and slow every evening with clean signal levels means congestion on your provider's node, which is a different conversation (see Advanced).
• Find your plan's "typical" speeds. Since 2024, US providers must publish a broadband label for every plan showing typical download speed, typical upload speed, and typical latency. It is on your provider's plan page or in your online account. Write those numbers next to yours. "Typical" is a stronger word than "up to," and it is the provider's own claim.
• Pull your modem's signal readings. See Step 2. If you rent a locked gateway and cannot see them, that is fine; you will ask the agent to read them to you.
• Save it. Screenshot each result or use the share link. Dated evidence is what turns a complaint into a ticket.

Step 2: Find your line numbers

Each connection type has a status page with a few readings that tell a technician more than any speed test.

Cable (Spectrum, Xfinity, Cox, Optimum, Mediacom, and most "high-speed internet" over coax)

Open a browser and go to 192.168.100.1. Most cable modems show a Status or Connection page without a password. Look for Downstream Power, Downstream SNR (sometimes MER), Upstream Power, and a column of Corrected and Uncorrectable errors (sometimes Correctables and Uncorrectables, or Codewords). Also open the Event Log and look for lines containing "T3" or "T4."

DSL (CenturyLink/Lumen, Frontier, Windstream, AT&T Internet over copper)

Sign in to your modem (the address is usually on a sticker, often 192.168.0.1 or 192.168.1.254). Look for a DSL Status or Line Status page showing SNR Margin (also called Noise Margin), Line Attenuation, and error counters labeled CRC, FEC, ES, or SES.

Fiber (AT&T Fiber, Verizon Fios, Google Fiber, Frontier Fiber, local fiber co-ops)

Look at the box where the fiber enters the house (the ONT). Note which lights are on, especially anything labeled LOS, Alarm, or PON. Some ONTs and gateways show an Optical or Light Level page with a reading like "-21.3 dBm." If yours does, write it down.

5G or fixed wireless home internet (T-Mobile Home Internet, Verizon 5G Home, AT&T Internet Air, local wireless ISPs)

Open the provider's app or the gateway's status page. Look for signal readings labeled RSRP, SINR, or RSRQ, and note the band or frequency. These tell you whether the gateway is in a good spot in your home.

Starlink

Open the Starlink app and check Statistics for Obstructions and Outages, and note how often the connection drops per hour.

If you cannot find any of this, skip it. The script below still works; you will just ask the agent to read the values to you.

Step 3: Fill in your card

Keep this in front of you during the call. If you ran the speed test on this site, your latest numbers are filled in below. Tap Copy my numbers to paste them anywhere, or Print to keep a paper copy by the phone.

Step 4: The call

Call during business hours on a weekday if you can. Tier 1 is often better staffed and more likely to have a technician slot the same week. Have your numbers ready first — run the test if you have not.

Opening

Say this before they start their script.

"Hi. I've already done the basic troubleshooting. I've power-cycled the modem and router, I'm testing on a wired Ethernet connection, and I have my test results and my modem's signal readings in front of me. I need you to open a trouble ticket and run a line test from your side."

That sentence does four things: it skips the reboot step, it rules out Wi-Fi, it signals you can read the same numbers they can, and it names the two things you want.

When they ask you to reboot anyway

"I've done that twice today. I'm happy to do it once more while we talk, but I'd like you to pull up my modem's signal levels first so we can see what they look like before and after."

Give your evidence, in this order

"On a wired connection I'm measuring [download] down and [upload] up against a plan with a typical speed of [label download] and [label upload]. More importantly, I'm seeing [X] percent packet loss and my latency goes from [idle] milliseconds to [under load] milliseconds when the line is busy. That's a line-quality problem, not a speed-tier problem."

Lead with packet loss, not speed. Any packet loss above zero on a wired connection is a legitimate fault. Speed numbers get argued about; loss does not.

Ask for the line test and the specific readings

For cable:

"Can you run a signal check on my modem and read me the downstream power, downstream SNR, upstream power, and the uncorrectable error counts? And can you check whether my modem is logging T3 or T4 timeouts?"

For DSL:

"Can you run a line test and read me the SNR margin, the line attenuation, and the CRC or errored-seconds counts since the last sync? And can you tell me if my line has been put on a reduced profile by line management?"

For fiber:

"Can you read me the optical receive power at my ONT, and tell me whether you're seeing any LOS or signal alarms on my account in the last 30 days?"

For 5G / fixed wireless:

"Can you check the signal quality on my gateway, specifically the SINR, and tell me how loaded the sector I'm connected to is during the evening?"

When they say the numbers are "within range"

"Within range of what? I'd like the actual values, please."

Then compare against the cheat sheet in Step 5. If a value is outside the healthy band, say so:

"Upstream power at [value] is above the normal ceiling, and the uncorrectables are climbing while we talk. That points to the drop cable or a connector outside, not my equipment. I need a technician to check the line from the tap to the house."

If the values are genuinely healthy but you still have packet loss, say:

"Then the loss is happening past my modem, on your side of the network. I'd like that escalated to your network or line team as a ticket, with a ticket number I can reference."

The dispatch ask

"Please schedule a technician. Since the readings indicate the problem is on your side of the demarcation point, I'd like it noted on the ticket that there should be no service charge if the fault is found outside my home or in the drop."

The demarcation point is the line between what the provider owns (everything from the street to the box on your wall) and what you own (everything inside). Using the word tells the agent you know who pays for what.

Before you hang up

"Can I have the ticket number, your name, and the date and window for the technician? And can you note on the ticket the readings you gave me today so the technician has a baseline?"

Write all of it on your card.

Step 5: Cheat sheet for the numbers

Values differ slightly by modem and provider. These are the bands used by the equipment makers and standards bodies, with sources at the bottom of the page.

Cable (DOCSIS 3.0 and 3.1)

DSL (ADSL and VDSL)

Fiber (GPON and XGS-PON)

5G / fixed wireless

Fixed wireless and satellite give you less leverage because there is no physical line to repair. Your options are gateway placement, an external antenna where the provider supports one, or asking for a different plan or a credit.

Step 6: When the technician arrives

Have your card ready and show them the readings you collected. Then ask for specific checks:

• Cable: "Can you check the levels at the tap, at the ground block, and at the modem, and tell me the difference? Can you look at the drop for damage and replace the fittings? Are there any splitters I don't need?"
• DSL: "Can you test at the NID and compare it to the modem? Can you check for a bridge tap and confirm the filters are in the right place?"
• Fiber: "Can you read the light level at the ONT and at the outside terminal, and clean the connectors?"

Before they leave: "What were the levels before and after? Can you put them on the ticket?" If the tech says the problem is upstream of your house (the tap, the node, the cabinet), ask them to escalate it to maintenance and give you that ticket number too. Line technicians and maintenance crews are different teams.

Step 7: If the first call did not fix it

Work up this ladder. Each step references the ticket number from the one before.

1. Call back and ask for Tier 2. "I have an open ticket, number [X], that wasn't resolved. I'd like to speak with advanced technical support or have this escalated to the network team." Repeat the evidence; add the technician's findings.
2. Ask for retention or the escalation desk. These teams have more authority to dispatch, credit, and schedule maintenance. "I'd like a credit for the days of degraded service on ticket [X], and I'd like to know what the plan is to fix it."
3. Write to executive customer relations. Most large providers have an executive escalation email or form. Send a short, dated summary: plan, typical speeds from the label, your wired results, the readings, the ticket numbers, and what you want.
4. File an informal FCC complaint. Go to consumercomplaints.fcc.gov and choose Internet, then Speed. It is free, takes about ten minutes, and the FCC serves it on your provider, which is then required to respond to you in writing within 30 days. Attach your screenshots. Providers route FCC complaints to a dedicated team that can do things front-line support cannot.
5. File with your state. Your state public utilities commission or attorney general's consumer protection office also takes internet service complaints. If your plan's broadband label promises typical speeds you never get on a wired connection, say so explicitly.

How to check it worked

After the fix, run the speed test again on a wired connection, at the same busy evening hour as your worst result. You want packet loss at zero, latency under load close to idle, and speeds near the label's typical numbers. Check your modem's signal page again too; the levels should now sit in the healthy columns above. Keep the before-and-after screenshots for a few weeks in case it slips back.

If your result was already good

If your wired test shows no packet loss, speeds near your plan, and the signal readings are healthy, do not call your provider. The problem is inside your home, and a technician visit may cost you money and fix nothing. The two usual suspects are Wi-Fi (slow in one room, fine next to the router) and bufferbloat (fast speed test, but calls freeze when someone else is streaming). Read the bufferbloat guide for the router fix, and if your router is more than five or six years old, see the Wi-Fi 6 router guide.

What not to say

• "The internet is slow." Say what you measured, on a wire, at what time.
• A Wi-Fi speed test number. It will be dismissed, and the agent is right to dismiss it.
• "Fix it or I'll cancel," on the first call. Retention threats work with the retention team, not Tier 1.
• "Everything is fine, I just want it faster." That is a sales call, and you will get a sales answer.
• Anything angry. The agent did not break your line, and a friendly, precise caller gets more done than a loud one.

Want to go further? Switch to the Advanced version above for what each reading actually measures, how to tell a line fault from node congestion, and how to prove where packet loss starts with a traceroute.

This is the technical version: what each reading actually measures, how to tell a physical line fault from shared-node congestion, and how to prove where packet loss begins. If you just want the words to say on the phone, switch back to Simple above.

Why the demarcation point decides everything

The demarcation point (demarc) is the physical boundary between provider-owned plant and customer-owned wiring. For cable it is typically the ground block or the first fitting where the drop enters the home; for DSL it is the network interface device (NID); for fiber it is the ONT or the outside fiber terminal. Providers can bill a service call when the fault is on the customer side. Every part of the script is designed to establish, with readings, that the fault is on the provider side. Agents hear that framing and know the dispatch will not be charged back to them.

DOCSIS readings in detail

Downstream receive power. The strength of signal arriving at your modem per channel, in dBmV. The healthy target is -7 to +7 dBmV, with -10 to +10 as the outer acceptable range; equipment makers such as Arris list -15 to +15 as the absolute spec. Two things matter more than the absolute value: all channels should be close together (a spread of more than roughly 8 to 10 dB across the bonded channels, called tilt, suggests a cable or amplifier fault), and the value should be stable across the day. Low power with high upstream power usually means excess attenuation: too many splitters, a corroded fitting, or a damaged drop.

Downstream SNR or MER. Signal-to-noise ratio (DOCSIS 3.1 modems often report MER, modulation error ratio, which is the same idea). For 256-QAM, 33 dB is the floor and 35 to 42 dB is healthy. On DOCSIS 3.1 OFDM channels, below about 34 dB the CMTS downgrades the modulation profile, which cuts your throughput before anything visibly breaks. Low SNR with normal power indicates ingress noise, often from a cracked drop or a loose connector letting RF in.

Upstream transmit power. How loud your modem has to shout to reach the CMTS. Healthy is 35 to 49 dBmV with three or more bonded channels (Arris allows up to 51 for SC-QAM with three or more channels, and higher with fewer channels). When upstream power climbs toward the ceiling, the modem is compensating for loss on the line; past about 51 dBmV it starts failing to range, which appears as T3 timeouts and partial-service mode (fewer upstream channels bonded than provisioned).

Codeword errors. Correctables are errors fixed by forward error correction and are normal in moderate numbers. Uncorrectables are packets lost. A fixed count since the last reboot is nothing; a count that climbs while you watch, across several channels, is physical-layer trouble. Packetsaver's reference puts the acceptable limit at under 0.01 percent of total codewords.

T3 and T4 timeouts. T3: the modem sent a ranging request and the CMTS did not answer in time, typically an upstream signal problem. T4: the modem waited too long for a station maintenance opportunity and reset. Repeated T3 entries plus high upstream power is the classic signature of a bad drop or tap, and a technician reading that log will know where to look.

Line fault versus node congestion. If every reading above is healthy and uncorrectables are flat, but speeds collapse and latency climbs every evening, the problem is almost certainly shared-node congestion: too many homes on the same segment. No technician visit will fix this. The script changes to:

"My signal levels are clean and my modem shows no errors, but throughput drops by [X] percent every evening between 7 and 11. Can you check utilization on my node and tell me whether a node split is scheduled?"

Ask for a credit; ask for the maintenance ticket number if a split is planned.

DSL readings in detail

SNR margin. The headroom between your actual signal-to-noise ratio and the minimum needed to hold the current sync rate at the target bit error rate. Standards and equipment vendors treat 6 dB as the minimum for a stable connection, 7 to 10 dB as fair, and above 10 dB as good. A margin that sits at 6 dB is not a coincidence; many ISPs configure a 6 dB target, and the line is then at the edge of stability by design. If your margin is low and your errors are high, you have a case for a line profile change or a pair swap.

Attenuation. Loss over the copper loop, which is mostly a function of distance to the DSLAM or cabinet. Under 20 dB is excellent, 20 to 40 dB is stable, 40 to 50 dB produces intermittent trouble, and over 50 dB is a line that will not hold sync reliably. Attenuation that is much higher than the provider's own loop-length estimate suggests a fault: a bridge tap, a wet splice, or corroded terminals at the NID.

Error counters. FEC counts are errors that were corrected and are expected. CRC errors are frames FEC could not repair; errored seconds (ES) and severely errored seconds (SES) count the seconds in which those occurred. Compare counts against uptime. A few CRC errors an hour is normal; CRC errors every minute with a healthy margin point to impulse noise, which a line tech can trace. Interleaved mode trades latency for error resilience; if the ISP moved you from fast path to interleaved, your ping went up by 10 to 20 ms and that is a conversation worth having once the line is clean.

DLM / line profile. Dynamic line management automatically steps a noisy line to a slower, more stable profile. If your sync rate dropped after a period of errors and never recovered, ask for a DLM reset once the physical fault is fixed. Otherwise the repair happens and your speed stays low.

What to ask the tech. Test from the NID with the inside wiring disconnected. If the readings at the NID are good and bad at the modem, the problem is inside (your cost). If they are bad at the NID, it is the provider's pair.

Fiber readings in detail

ONT receive power. The light level arriving at your ONT, in dBm (more negative is weaker). For GPON Class B+, the receiver spec runs from -8 dBm (overload) to about -27 or -28 dBm (sensitivity). Well-designed links land between -15 and -25 dBm. From -25 to -27 the link works today but has no headroom for a dirty connector or a tight bend; below -28 you get LOS and drops. Too strong is also a fault: brighter than -8 dBm saturates the receiver and shows up as CRC errors and flapping on a link that "looks great." XGS-PON specs are similar (roughly -9 to -28 dBm for class N1). Fiber faults are nearly always a connector (dirty or damaged), a macrobend (fiber pinched or coiled too tightly), or a splice. All of those are the provider's to fix outside the demarc.

What to ask the tech. Read the level at the ONT and at the outside terminal; clean and reseat connectors; inspect the drop for bends. Ask for the numbers before and after.

Fixed wireless and satellite

RSRP (signal strength), RSRQ (signal quality), and SINR (signal relative to interference and noise) describe the radio link from the gateway to the cell. SINR above about 13 dB is good; below about 5 dB expect packet loss and retransmissions. Because the link is shared air, evening collapse is sector congestion, and the provider's only real remedies are moving you to a different band or sector, adding capacity, or crediting you. For Starlink, the obstruction map in the app is your evidence; support is ticket-based, and the fix is almost always relocating the dish.

Proving where the loss starts

If you are comfortable with a terminal, a traceroute or mtr run during a bad period lets you say exactly where packets start disappearing.

• Loss at hop 1 (your router) with no loss beyond it is in-home. Do not call.
• Loss beginning at hop 2 or 3 (your provider's CMTS, BNG, or first router) and persisting to the destination is on the provider's access network. Say: "I'm seeing consistent loss starting at your second hop and continuing end to end, so it's not my equipment."
• Loss at a single middle hop that does not continue to later hops is a router that deprioritizes ICMP, not a fault. Do not report it.

Note: the packet-loss figure from the speed test on this site is measured end to end from your browser, which is the number that matters for your experience. The hop-by-hop view is what tells you who owns the problem.

The vocabulary, in one place

Line test. Signal levels. Downstream power. SNR or MER. Upstream power. Uncorrectables. T3 timeouts. Partial service. Drop cable. Tap. Ground block. Node utilization. Node split. SNR margin. Attenuation. CRC errors. Errored seconds. Bridge tap. NID. DLM reset. Optical receive power. LOS. ONT. Demarcation point. Trouble ticket. Dispatch. Escalation. Tier 2. Retention. Maintenance ticket. Broadband label. Typical speed. Informal complaint.

You do not need all of them. You need the three or four that match your readings, said calmly, with the numbers.

Sources

• Arris cable modem signal level references (downstream power, SNR by modulation, upstream power by channel count): SB8200 · S33
• Packetsaver, DOCSIS 3.1 signal level guide (ideal ranges, uncorrectables, T3/T4 thresholds): packetsaver.com
• ModemGuides, reading modem signal levels and T3/T4 timeouts: modemguides.com
• Keenetic, VDSL line parameters (SNR margin and attenuation bands): support.keenetic.com
• Kitz, interpreting DSL line stats (SNR margin definition): kitz.co.uk
• Fiber Optic Association, GPON receiver power range: foa.org
• Fiber Optic Association, Advanced GPON Fault-Finding Manual 2026 (ONT Rx by class): foa.org (PDF)
• ITU-T G.9807.1, XGS-PON ONT sensitivity and overload by class: itu.int
• FCC, Filing an Informal Complaint (free, 30-day written response requirement): fcc.gov
• FCC Consumer Complaint Center, Internet complaint categories: consumercomplaints.fcc.gov
• FCC, Broadband Consumer Labels: fcc.gov/broadbandlabels