The installer runs the speed test twice before he coils up his cable spool, and both times the numbers land almost on top of each other — 935 down, 925 up, printed on a slip he hands me on his way out past the standing desk, the shelf where the router and modem now share space with a box I leave running around the clock just to log this stuff, and the whiteboard where I track every provider I've tried, sorted by time of day. The only light down here comes off two monitors and an LED strip behind that shelf — no windows to speak of, which is fine, because none of what matters in this room happens outside it.
That evenness on the slip — not the download number printed on my old plan's box — is the whole reason a video call holds together or falls apart. More download speed does not fix a bad Zoom call. Jitter does: the variation in how consistently your data packets show up, one after another, in order. Most internet performance advice for remote work skips straight past it and goes back to arguing about gigabit tiers.
The Myth: More Mbps Won't Save a Bad Call
Picture a driveway that's four lanes wide coming in and a single lane going back out — that's roughly the shape of a lot of residential cable plans. You might be sold 500 Mbps down and only 20 Mbps up, and on a call, upload is the lane doing all the work: your camera, your microphone, your screen share, all pushed out through that narrow exit while the wide entrance sits mostly idle. Zoom's own published requirement is about 3.8 Mbps of steady upload for a clean 1080p picture (Zoom), and "steady" is doing more work in that sentence than the number itself. A connection that occasionally spikes above 3.8 Mbps but keeps dipping under it will still stutter — the average looks fine, the moment-to-moment reality doesn't.
What Jitter Actually Measures
Jitter is the variation in how long each little packet of your voice or video takes to arrive, relative to the one before it. On a bad connection, packets don't get lost so much as they get delivered out of rhythm — like a phone call where every third word lands half a beat behind the rest, close enough to follow but never quite natural. For video conferencing specifically, jitter decides call quality more directly than raw download speed does. A 50 Mbps connection holding jitter to 5 ms will consistently outperform a 200 Mbps connection that swings up to 40 ms once real people start talking over each other — the fast connection just can't keep its packets in line. The commonly cited threshold for VoIP and video quality sits around 30 ms of jitter; cross it regularly and voices start to clip and flatten into that telltale robotic stutter.
Does Fiber Really Handle Jitter Better?
Copper lines — cable and DSL both — pick up electrical noise and react to weather in ways that fiber mostly doesn't, and that shows up as jitter long before it shows up as a slower download number. I know this one the hard way: when I called AT&T to upgrade my old DSL line to fiber, the rep told me it wasn't provisioned at my address yet and put me on a waitlist that ran fourteen months before a crew finally showed up with the printout I mentioned earlier. Fourteen months is a long time to keep telling clients the choppy audio is on their end, not mine. The comparison I eventually ran between AT&T Fiber vs Spectrum Cable for Kansas City Remote Workers came down to exactly that jitter floor, not the gigabit ceiling printed on either brochure. None of that means cable is unusable for calls, and the fuller case for choosing fiber over cable as an IT professional working from home is really its own conversation. It also isn't the whole story: a fixed-wireless 5G plan and a satellite connection can each post an eye-catching download number on a speed test, and each has its own separate jitter behavior worth weighing on its own terms before you sign anything.
Testing After Dinner Changes Everything
A speed test run late at night, when nobody else on your street is using their connection, tells you almost nothing about how your calls hold up during the workday. What actually matters is loaded latency — how your connection behaves once it's carrying real traffic instead of sitting idle. This is bufferbloat: your router or your provider's equipment gets a pile of requests at once and starts queuing packets in a buffer rather than dropping them, which is harmless for a paused video stream and fatal for a live conversation. Widely cited guidance puts the ceiling for acceptable one-way latency at around 150 ms; cross that regularly and you start talking over people, because the delay is long enough to wreck the natural rhythm of a conversation. A reader named Mercy Elrod, a freelance video designer, once wrote to me describing an afternoon spent watching a large render creep up a shared cable connection while a client waited on the other end — her plan's advertised speed was never the issue, the queue sitting in front of her upload was. The deeper argument for treating upload as the priority metric over download is a bigger topic than I can do justice to here, but her situation is a clean example of why it matters.
Where the Fine Print Starts to Add Up
Fixing the technical side only gets you halfway, because the same providers that fumble jitter also tend to bury the terms that make switching painful. My colleague Lindiwe Odden, who works remotely out of Olathe, treats every call to a retention desk like a negotiation with a floor price already fixed in her head before she even dials — she won't hang up until the rep either meets it or admits they can't. That instinct is worth borrowing, because promotional pricing resets, equipment rental lines, and installation fees are exactly the kind of thing that shows up quietly a few months in. I went into more detail on what one newer fiber provider actually itemized on my bill after switching in a separate piece on Quantum Fiber Billing Transparency and What I Noticed After Switching — worth a read before you assume every provider's fine print behaves the same way.
Troubleshooting the Connection Like a Technician Would
None of this is useful if you're testing wrong, so here's the actual process worth running before you blame your provider for a bad call. Start by plugging directly into the router with an ethernet cable — testing over Wi-Fi measures your walls and your microwave as much as it measures your internet. From there, look for a testing tool that reports loaded latency specifically, meaning latency measured while the test is actively pushing data in both directions; if your idle number is a clean 20 ms but jumps past 200 ms the moment the download or upload starts, that gap is your bufferbloat problem laid out in plain numbers. Check the jitter reading next and expect it to sit consistently under 30 ms — wild swings point to a line-quality issue rather than a capacity one. Then ignore the peak number the test flashes at the top of the screen and look instead at the floor, the lowest reading during the run, because that floor is what a live Zoom call actually has to work with the whole time. For anyone stuck on a cable contract while better infrastructure makes its way down the street, I laid out what a no-contract stopgap plan looks like in practice in Spectrum Internet Review for Households Avoiding Long Term Contracts.
The rule worth carrying out of all this is simple: pull up the jitter and loaded-latency numbers before you even glance at the download figure on the flyer. A quiet, boring five milliseconds of jitter will carry you through a client presentation further than any advertised gigabit ceiling ever will — the download number sells the plan, but the jitter number is what actually shows up on the call.