The worst version of this job is the one where you find out on the ridge. Ladder set, homeowner already watching the loading spinner on their phone at work, and now you are troubleshooting a network problem while standing on a slope. Nobody makes good decisions in that position.
So I check cell signal before an inspection every time, and it takes under a minute. Not because I am careful by nature. Because I got burned enough times that the minute became cheaper than the alternative.
The problem with the check most people do
Most people open a speed test in the driveway, see a big number, and climb. That check is close to worthless, for four reasons.
It reports download. You are uploading. Those are different numbers on a cell network and the second one is usually a fraction of the first.
It reports a burst. A speed test grabs everything the network will give it for a few seconds against a nearby well-provisioned server. A stream needs steady throughput for twenty minutes. Peak and sustained are not the same measurement.
It does not report loss or jitter, which is what actually breaks live video. A clean 1.2 Mbps link beats a ragged 3 Mbps link with 8 percent packet loss every time, and the speed test will tell you the ragged one is better.
And it happened in the driveway. Twenty-five feet of elevation and forty feet of horizontal put you on a different sector, with different obstruction, possibly above the downtilt of the tower you were happily attached to at ground level.
The 60-second version
Here is the actual pre-flight. It fits in the time it takes to get tools out of the truck.
Turn Wi-Fi off. Do this first, every time. Your phone latched onto the homeowner's network the moment you pulled up, and it will hold that connection at the far edge of range on the roof rather than falling back to cellular. This one habit fixes more bad streams than everything else on this list combined.
Read the real signal numbers, not the bars. On Android that is Settings, About phone, SIM status or network status. Received power in the negative 80s or 90s is fine. Around negative 105 I get cautious. Negative 115 and I am planning a different day. If your phone shows signal-to-noise, that is the better number: healthy is above about 15 dB, single digits is rough, negative means bars without throughput. iPhone's field test screen has moved around across iOS versions and is not reliable to count on, which is why I keep an Android in the truck.
Run an upload-only test, twice. Not download. Upload. Run it, wait fifteen seconds, run it again. Two similar numbers means a stable cell. A 4 Mbps result followed by a 900 kbps result means a congested one, and the second number is the one you should believe.
Test from where you will work, not where you parked. If you can walk the perimeter to the side of the house you will spend the most time on, do it. If there is a second story deck or a walkout, test from up there. Elevation changes the answer, sometimes for the better and sometimes badly.
Check the homeowner's end. This gets forgotten. It does not matter what your uplink does if the person watching is on hotel Wi-Fi, in a parking garage, or driving through a dead stretch. Ask where they will be. Thirty seconds of asking has saved me entire appointments.
What the numbers mean in practice
I keep three buckets in my head and I decide which one I am in before I climb.
Above about 3 Mbps sustained upload, clean and repeatable, I run a normal live walk at 720p and do not think about it again.
Between roughly 1 and 3 Mbps I still go live, but I change how I move. Slower pans, longer holds on anything I want seen clearly, more verbal narration to carry the parts the picture cannot. I tell the homeowner up front that the video will be soft and that I am recording a clean copy either way. Managing the expectation before it happens is a completely different conversation than explaining it after the picture turns to mush.
Below roughly 800 kbps sustained, or any upload test with visible loss, I do not go live. I record the walk and send it. That is not a failure, it is a different tool. The mechanics of why a link behaves that way are in the main piece on streaming at one bar.
The exact cutoffs drift depending on your gear and resolution. The buckets are the useful part. Decide which one you are in on the ground.
The stuff that changes the answer
A few things reliably move the result, and it is worth knowing them so you are not confused when the numbers surprise you.
Time of day matters more than most people expect. Cell capacity is shared. The same address at 8 a.m. and 4 p.m. can be two different networks, especially near a school or a commuter route. If you book a marginal address, book it early.
Weather matters less than people think for the radio itself, and a lot for you. Rain does not do much to low and mid band cellular at these distances. Wet clothing and a wet body between the phone and the tower does more. So does standing on the far side of a metal roof from the tower.
Being high can help or hurt. If the problem was trees or a hill, height fixes it. If you are close to the tower, height can put you above the antenna's downtilt and make it worse. There is no rule, which is exactly why you check cell signal before an inspection instead of assuming.
And carrier plans are not equal. A lot of unlimited contractor plans deprioritize your traffic after a data threshold, which you will never notice on an empty tower and will feel immediately on a busy one. Tethering a tablet through your phone is usually deprioritized harder still, so stream from the phone directly if you can.
Make it a habit, not a decision
The value here is not the specific thresholds. It is that you find out on the ground, with a ladder still on the truck and a homeowner who has not yet been promised anything.
Sixty seconds. Wi-Fi off, look at the real numbers, upload test twice, ask where they will be watching. Then decide.
If you want to see what a marginal link looks like before you meet one on a customer's roof, run a test session on your own house from the worst corner of your yard. Watching a stream degrade when nothing is riding on it teaches you more than reading about it. The session flow is the same either way.
InspectStream records what an inspector observed. It is not a public adjusting service and does not prepare, negotiate, or advise on insurance claims.