Measuring Wi-Fi Signal Where the Equipment Lives
Why this matters
"The Wi-Fi is fine, I have full bars" is the sentence that starts most connected-device arguments, and it is almost always said in the living room. The bars on a phone are a crude summary of the band the phone happens to be on, in the spot where the phone happens to be. A number taken where the equipment actually sits, on the band the equipment actually uses, ends the argument and decides the job: commission it, fix coverage first, or tell the owner the device cannot be supported there as the network stands. It takes three minutes and a free app, and it is the reading every other connected-device decision on this shelf leans on.
What you are measuring
Signal strength is reported as RSSI (received signal strength indicator, a term from the IEEE 802.11 standards) in dBm, decibels relative to one milliwatt. The numbers are negative, and closer to zero is stronger: -50 is excellent, -80 is poor. The scale is logarithmic, so every 3 dB is a halving of power and 10 dB is a tenth. A drop from -65 to -75 is not "a bit worse", it is a tenth of the power.
The thresholds this shelf uses, as rough rules for low-bandwidth equipment such as thermostats, controllers and sensors, measured on the band the device uses:
| Reading on the device's band | What to expect | What to do |
|---|---|---|
| -67 dBm or stronger | Solid. Joins and stays joined. | Proceed. |
| -68 to -75 dBm | Marginal. Usually joins; drops with doors, weather, a parked car. | Proceed only with the owner told, and recorded as marginal. |
| -76 to -80 dBm | Poor. Joins unreliably, drops often. | Fix coverage before commissioning. |
| Weaker than -80 dBm | Unreliable to absent. | Do not commission onto this network as it stands. |
These are working thresholds, not standards, and two things move them. A camera or anything streaming video needs more margin, so read the solid line as nearer -60. And a reading can look fine and still fail on a crowded channel, which is why step 5 records the neighbors too.
Signal is not the only number. Noise (the background radio energy) and the gap between signal and noise, the signal-to-noise ratio, matter as much in crowded places. Few phone apps show noise directly; the neighbor count stands in for it.
Before you start
You need a phone with a signal app that reports dBm per network and per band, and shows nearby networks and their channels. Platform rules on what apps may read change: some phone platforms expose full readings to apps, others restrict them and show only bars, in which case a different phone or a cheap dedicated Wi-Fi analyser is needed. That is current as of 2026; check your platform's current behavior before relying on it.
You also need the device's band (2.4 GHz only, or dual band) from its spec sheet. A 5 GHz reading taken for a 2.4 GHz device is a measurement of something irrelevant.
The steps
1. Read at the router first, as your control
Stand within a few feet of the customer's router or main access point and record the reading on each band. This proves the app works and the router is broadcasting, and it sets the top of the scale for this house. Expect something in the -30s or -40s.
What breaks if skipped: a bad reading at the equipment could be a weak router, a phone with a poor antenna, or the app reading the wrong network, and you will not know which.
2. Go to the device and get the phone where the device's antenna is
Hold the phone at the device's antenna location: at the height of the module, on the side of the cabinet where the module sits. If the module is inside a sheet-metal cabinet, hold the phone against the closed cabinet face nearest the module, with any service panel installed as it will be in service. Do not reach into a powered cabinet to put a phone next to a module. If the module is behind a high-voltage panel, read from outside the closed panel; that reading is optimistic, and step 4 corrects for it.
Equipment rooms bring their own hazards. In an attic, stand on boards or joists, never on bare drywall, and take the reading before the heat of the day. In a crawlspace, check for standing water near wiring before you crawl in, and if water is in contact with any electrical equipment, do not enter: have the circuit shut off at the panel first.
What breaks if skipped: a reading taken three feet away in open air reads several decibels better than the module will ever see.
3. Take three readings, on the device's band, and record the weakest
Read the device's band, then wait about ten seconds and read again, three times. Signal fluctuates, and the weakest of the three is the honest number because the weakest moment is when a device drops. Note whether the network you are reading is the main network the device will join, not a guest network or a neighbor with a similar name.
What breaks if skipped: a single reading catches whatever moment you happened to take it, and a lucky -66 hides a typical -72.
4. Correct for the cabinet, or read the device itself
The phone is a proxy. The device's own small antenna, inside its enclosure, usually does worse than a phone held outside. If the device or its app reports its own signal (many do, in a diagnostic or about screen), record that number too, labeled as the device's report, and treat it as the more relevant one once the device is joined. If you only have the phone reading through a closed sheet-metal cabinet, treat a borderline phone reading as being in the next band down until the device's own report says otherwise: a phone reading of -72 dBm against a closed metal cabinet is plausibly -76 or worse at the module. A plastic housing costs far less and needs no such correction.
What breaks if skipped: you commission at a phone reading of -70 that is really -78 at the module.
5. Record the neighbors and the channel
From the app's nearby-network list, note the customer's channel on that band and how many other networks share it at a readable strength. On 2.4 GHz, a dozen neighbors on the same channel can make a -65 reading behave like a -75.
What breaks if skipped: the next tech sees a solid signal on record and cannot explain the drops.
6. Measure the house as it will be lived in
The reading that matters is the one the device will see on an ordinary evening, not the one you get with the service door propped open. Before your three reads, close the garage door and the door between house and garage, put the car back in the bay if it normally parks there, leave the attic insulation and hatch as they will be in service, and reinstall any panel you removed. If the customer runs a microwave or a second router near the path, note it. A garage reading taken with the overhead door open and the car out can read several decibels better than the same spot on a winter night with both doors shut and a car between the module and the house.
What breaks if skipped: you record a solid reading for a house that only exists while you are standing in it, and the first drop happens the night the car comes home.
7. Write it on the job, in the same shape every time
A reading nobody records helps nobody. Use one line format so the next tech, or you in six months, can compare like with like.
A worked example: the record line, field by field
A plumber commissions a leak-detection hub in a basement utility room. The hub is 2.4 GHz only. The record that goes on the job reads:
WIFI SIGNAL - LEAK HUB, BASEMENT UTILITY RM
date/time : Tue 09:40
network : main (not guest), 2.4 GHz
control @router: 2.4 -41 dBm / 5 -44 dBm
at device : phone against hub case, 3 reads
-69 / -73 / -71 -> record -73
device report : -74 dBm (hub diagnostics, once joined)
channel : 6, 9 neighbors visible on ch 6
verdict : MARGINAL, owner told, see note
note : path is slab + 2 floor joists + duct
trunk; node on main floor stair wall
tested at -62 at hub, owner declined
Walk the fields. The router control at -41 dBm proves the house broadcasts well. Three readings at the hub, -69, -73 and -71, give a weakest of -73, which falls in the marginal band (-68 to -75). The hub's own report, once joined, was -74, one decibel worse than the phone. The hub sits in a plastic case on the wall, which costs very little, so step 4's next-band-down correction for a closed metal cabinet did not apply; the device's own report confirmed that. Nine neighbors on channel 6 is crowded, which pushes a marginal reading toward the poor end in practice.
The plumber tested a borrowed node on the stair wall and the hub read -62, solid. The owner declined to buy one. The verdict line records the result as marginal and records that the owner was told, which matters the first night the hub misses a leak alert. Recommending and installing that node is the owner's decision; this record is the evidence for it.
What would have changed the verdict. Had the weakest reading been -78 instead of -73, the hub falls in the poor band and the right move is to stop commissioning and fix coverage first, regardless of the owner's preference, because a leak sensor that drops is worse than no sensor: the owner believes they are protected. Had the device been a camera, -73 would be poor rather than marginal.
How to verify you measured it right
Compare the device's own reported signal, once it joins, with your phone reading. They should agree within a few decibels, with the device usually slightly worse. If the device reports much stronger than your phone, you probably read the wrong network or band; if it reports far weaker, the module's antenna may be shielded or damaged, or it is connected to a different access point than you assumed. Either way the mismatch is a finding to record, not a number to average.
References
- IEEE 802.11 family of standards (RSSI and signal reporting)
- Device manufacturer documentation (radio band, minimum recommended signal, diagnostic screens)
- See related: Home Wi-Fi for the Trades; Mesh Systems, Extenders and Dead Zones; The Connected-Device Commissioning SOP
- See related: Leaving a Device Record the Next Tech Can Use