Wi-Fi Complaints: Coverage, Congestion and the Device

Thursday, 14:10, at Stonebridge Insurance. Three tickets have arrived in forty minutes, all from the second floor, all saying some version of "the Wi-Fi is terrible today". Hamid takes them in order. The first is from Claire in the claims team, whose desk is in the far corner by the fire escape. Her laptop shows one bar and a page that will not load. The second is from the large meeting room. The quarterly review has twenty-two people, twenty-two laptops, roughly twenty-two phones, and a video call that keeps freezing. The third is from Ravi, who sits in the middle of the floor, ten paces from an access point. His laptop is crawling while the colleague beside him streams a training video without complaint. Three tickets, one floor, one word, and not one of them is the same fault.

A Wi-Fi complaint is a slow or dropping connection reported by someone on the wireless network. Nearly all of them come down to one of three things. Coverage: the signal where the person is sitting is too weak. Congestion: too many devices are sharing one access point's airtime. Or the device: the laptop itself is connecting badly, to the wrong access point, or with an elderly driver. The person's word for all three is "the Wi-Fi", and the desk's job is to tell them apart before anyone touches the network. This article is part of our The Internet Is Slow series. The four-way isolation of any slow complaint, user, device, network or expectation, is in the first article. This one picks up the ticket once "who else" has said "this room" or "this laptop" and the branch is wireless.

It starts with what Wi-Fi actually is, in plain words, the dead corner, and the full meeting room. Then come the laptop that clings to the far access point, the old wireless driver and the device side, and the microwave and other neighbours. Next are the walk-test with a known-good device, and what to tell the user and what to tell the network owner. A short war story and the Wi-Fi complaint checklist to copy finish the article.

What Wi-Fi Actually Is, in Plain Words

Wi-Fi is radio, and almost everything confusing about it stops being confusing once you hold on to that. An access point is a small radio transmitter, usually on the ceiling, that turns the wired network into a signal that laptops and phones can hear. Each access point covers a rough bubble of space. Walls, floors, metal cabinets, large aquariums and crowds of people shrink the bubble. The signal gets weaker the further you are from the centre. A laptop at the edge of the bubble hears the access point faintly and talks back faintly. Faint radio is slow radio: the two ends keep slowing down until they can understand each other.

The second thing to hold on to is that the airtime is shared. One access point can only talk to one device at a time, taking turns very quickly. So every device connected to it is sharing the same slice of radio time. Twenty devices doing nothing cost almost nothing; twenty devices each on a video call cost everything, and the access point spends its day handing out tiny turns to everyone. A fast connection to the building is no help if the access point in the room cannot get the bits into the air. Most access points broadcast on more than one radio band. The older, lower band reaches further through walls while being more crowded. The newer, higher bands are faster and die sooner. Which band a device lands on is largely the device's choice.

The third thing is that the device decides. The laptop chooses which access point to connect to, when to move to a better one, and which band to use. The network can only offer hints. This is why two people at adjacent desks have different experiences on the same network. One laptop made good choices. The other is still loyally attached to the access point it found when its owner walked in from the lift. Nobody designed it to be this way on purpose. It simply grew up that way, like most things with a standards committee.

The Dead Corner

Coverage is the first of the three causes, and the dead corner is its usual shape. It is a desk, a row or a small room where the signal is weak for everyone, all day, whatever device they bring. Claire's corner by the fire escape is a dead corner, and the tell is in the pattern. It is slow for Claire and it was slow for whoever sat there before her. A phone shows one bar standing at her desk. Three paces toward the middle of the floor it is fine. Coverage problems do not move, do not come and go with the time of day, and do not care whose laptop it is.

The cause is nearly always distance or obstruction. The nearest access point is too far away, or there is something between it and the desk that radio dislikes. A lift shaft, a stairwell of concrete, a kitchen with a metal-fronted fridge. A server room cage, or a wall that turned out to be solid where the plan said partition. Buildings also change after the access points are placed. The new filing cabinets, the glass-walled office with the metallic film, the plant on the windowsill that grew: each moves the edge of the bubble a little. The desk only hears about it when someone sits down on the wrong side of it.

The fix belongs to the network owner, and it is one of three things. Move the desk, move or add an access point, or run a cable to the corner and let the person plug in. The desk's contribution is the evidence: where the signal is weak, measured the same way at three or four spots, with the known-good device. Then the owner can decide between "add an access point" and "give Claire a cable" without visiting. A dead corner is also a mark on the coverage map, which is the first tool in reducing slow-internet tickets. Three tickets from the same corner in a year are the argument for the access point. The person in the corner cannot be fixed by restarting anything. They can be fixed by a cable in about four minutes.

The Full Meeting Room

Congestion is the second cause, and the full meeting room is where it lives. The quarterly review has twenty-two laptops and as many phones on the access point in the ceiling, plus the room's own screen. The presenter's video call is sharing the airtime with twenty-one people checking mail, three downloading the deck, and every phone quietly synchronising something. The signal is strong, every bar is showing, and the call freezes anyway. Congestion looks like weak signal to the person and nothing like it to the measurement, which is the main reason to measure rather than guess.

The tells are time and headcount. A congested room is fine at eight in the morning with two people in it and unusable at ten with twenty. It recovers the moment the meeting ends. The same room with the same access point and the same laptop is a different network depending on who else is in it. A desk that keeps a note of the room, the headcount and the time on each complaint will see the pattern inside a month. The pattern is the whole diagnosis. The question is not "is the Wi-Fi broken" but "how many people is this room's wireless actually sized for". That is a question for the network owner with a number attached. The sixth article in this series returns to it under the meeting-room capacity question.

The fixes are capacity and discipline, and only one of them is the desk's. Capacity is a second access point in the room, or a better one. Or moving the room's own screen and call equipment onto a cable so that the heaviest traffic in the room is not in the air at all. That is the network owner's decision and the room owner's budget. Discipline is the desk's: the presenter's laptop on the room's wired port, and the deck shared before the meeting rather than downloaded during it. Phones that do not need to be on the office wireless are left on their own service for the hour. The hybrid meeting, where the remote half of the room has its own complaints, is the subject of Hybrid Meetings and the Remote Half. The wireless in the room is half of what makes it hard. (The other half is that the remote half cannot see who is talking, and that is not a radio problem.)

The Laptop That Clings to the Far Access Point

The third cause begins with the device, and the laptop that clings is its commonest shape. Ravi walked in from the lift lobby this morning, and his laptop connected to the access point nearest the lift. It has stayed connected to that access point ever since, across the floor, through two walls, to a desk that sits almost underneath a different one. The signal from the lift-lobby access point is faint at his desk. The laptop, which decides these things, has judged it good enough and has not looked for anything better. The colleague beside him, whose laptop arrived from the stairs, is on the near access point and perfectly happy.

The tell is the mismatch: a weak or slow connection on one device in a place where another device is fine. The reading shows the laptop attached to an access point it has no business being attached to. Most operating systems will show which access point a device is on, or at least the signal strength, somewhere in the wireless status. The network owner's controller shows it plainly. A laptop at one bar sitting ten paces from an access point is clinging. The fix in the moment is almost absurdly cheap: turn the wireless adapter off and on. The laptop, forced to choose again, chooses the near one. The person reports that the Wi-Fi "has come back", and in a sense it has.

The lasting fix is partly the network's and partly the device's. Some wireless controllers can nudge a clinging device toward a nearer access point or decline to serve one that is too far away. Some cannot. The network owner knows which theirs is, and a note saying "clinging devices on the second floor, three this month" is the prompt. On the device side, a current wireless driver chooses better than an old one, and a laptop that clings every day is usually a laptop that is overdue for one. The desk's job is to recognise the pattern, apply the adapter off-and-on, write it down, and stop the next technician from blaming the access point that was never involved.

The Old Wireless Driver and the Device Side

A driver is the small piece of software that lets the operating system talk to a particular piece of hardware. The wireless adapter's driver decides how well the laptop hears, chooses and roams. An elderly one makes bad choices quietly: it clings, it prefers the crowded lower band, it drops and reconnects. It does all of this on a machine that is otherwise fine, in rooms where every other device is fine. The person reports the room. The room is innocent. One slow laptop in a happy room is a laptop problem until proved otherwise, and the proof is the known-good device beside it.

Four device-side checks earn their place on every Wi-Fi ticket, in order of cost. First, the known-good device on the same desk doing the same task; if it is fine, stop blaming the room. Second, the adapter off and on, which forces a fresh choice of access point and band. Third, the state of the wireless driver and the operating system's pending updates, which the desk applies through whatever process it already uses for updates. Fourth, the power-saving setting that many laptops apply to the wireless adapter on battery. This lets the adapter doze to save power and wake up slow. A laptop that is only slow when unplugged has usually told you the answer. The general order of checks for a machine that is slow at everything, not just the wireless, is in The Diagnostic Order for a Sick Laptop or Desktop.

The dock deserves a sentence, because a docked laptop may not be using its own wireless at all. Some docks carry a wired connection; some carry nothing. A laptop that was fine undocked and slow docked is pointing at the dock, the cable behind it, or the port on the wall it is plugged into. None of those is radio. The things that break on and around the desk, docks and cables among them, have their own article in Docks, Cables and the Things That Break. Here it is enough to ask whether the laptop is docked, and to try it undocked before anyone says the word "Wi-Fi" again.

The Microwave and Other Neighbours

Interference is coverage's awkward cousin: the signal is strong enough, but something else is shouting on the same frequencies. The access point and the device keep having to repeat themselves. The microwave oven is the classic, because it leaks a little energy on the same frequencies the older, longer-reaching wireless band uses. A kitchen between an access point and a row of desks will produce a slow patch at twelve-thirty every day that is fine at every other hour. Cordless phones, some wireless cameras and badly shielded equipment all do versions of the same thing. So do a wireless printer broadcasting its own network to nobody and the neighbouring tenant's access points on the other side of the wall.

The tell is a pattern in time or place that does not match headcount. Slow at lunch and fine at two is the kitchen. Slow along one wall and fine in the middle, with no cabinets or concrete to blame, is often the building next door. Slow near one cupboard is whatever is in the cupboard. The desk's evidence is the same as for coverage, the walk-test with times. The fix is the network owner's: a channel change on the access point, a move of the equipment, or a move of the desks. Nobody is going to move the kitchen. (Someone will propose moving the kitchen. It will be on the agenda for a year.)

The Walk-Test With a Known-Good Device

The walk-test is the one method that settles coverage, congestion and the device in a single pass. It needs one trusted laptop, a route and ten minutes. Take the known-good device, the spare the desk keeps clean and updated for comparisons, and walk the complaint's route. Start at the person's desk, then a point halfway to the nearest access point, then directly under the access point. Then the next room, and a wired port if there is one. At each point, note three things. The signal strength as the device shows it. Which access point it is on, if the device or the controller will tell you. And the result of one repeatable task, which can be a speed check or a fixed large file opened from the same server each time. Write the time beside each line.

Read the results as a shape. Weak everywhere near the desk and strong under the access point, with the task slow where the signal is weak: coverage. The map shows the owner where the bubble ends. Strong signal everywhere but the task slow in the room and fine next door, at a time the room is full: congestion. Strong signal and a fine task on the known-good device at the person's own desk, while their laptop crawls beside it: the device. The four device-side checks follow. A slow patch at a set time of day along a set wall: interference, and the time is the clue. The walk-test is a comparison, not a verdict. The full method of taking numbers that the network owner or the provider will accept is in Measuring Slow, which this article leans on rather than repeats.

Do the walk-test at the time of the complaint, not the next morning, because congestion and interference keep office hours. A walk-test at eight that finds nothing proves only that the room is fine at eight. If the complaint says ten o'clock on review day, that is when the known-good device goes to the room. If the desk cannot be there at ten, it should ask the person to run the self-check from the sixth article at ten and send the numbers. The evidence is only as good as the hour it was taken.

What to Tell the User and What to Tell the Network Owner

The user and the network owner need different messages from the same ticket, and writing one message for both serves neither. The user needs to know what was found, what it means for them today, and what happens next. Use plain words with no radio in them: "Your desk is at the edge of the wireless coverage on that floor, which is why it is slow for anyone sitting there. I have plugged a cable in so you are on the wired network from now. The network team has the measurements to decide whether to add an access point for that corner." That is the whole reply. It says the person was right, it fixes their day, and it does not ask them to understand bands.

The network owner needs the evidence and nothing else. The walk-test lines with times, the known-good device used, and which access point the slow device was on. The headcount if the room was full, and what the desk has already done. "Second floor east corner, four desks. Signal at one bar on the spare laptop, three bars eight paces in, fine from the wired port. Three tickets from the corner this quarter. Cable run to one desk as a workaround; nothing on the network touched." An owner can act on that note from their desk. It is the kind of note that gets a corner fixed. The owner does not need to hear that the Wi-Fi is "terrible". They have heard that. They have never once heard a number.

Keep the two messages in the one ticket, the user's as the public reply and the owner's as an internal note. If several people in the same corner raised separate tickets, link them to one ticket for the corner. Then the owner receives a single ask and the users each get their own answer. If you run QuerySolv, linking the three second-floor tickets to one network ticket is the ordinary way to do it. Then the owner's reply on the parent reaches the technician without three copies. In any tool, the test is that the corner's history is in one place, because the fourth ticket from that corner will come. Whoever takes it should find the first three in under a minute.

A Short War Story: The Kitchen Access Point at Redbrook Engineering

At Redbrook Engineering, the design office complained for most of a winter that the Wi-Fi along the north wall collapsed around midday. A technician called Freya restarted the access point serving that wall from the controller each time. This took the whole design office off the wireless for two minutes. It fixed nothing, because by the time it came back the lunch rush was over anyway. The pattern was in the tickets, eleven of them, all between twelve and one, all from the same six desks. Nobody had read them together until a new network contractor asked for a month's worth. The contractor noticed that the six desks sat directly behind the staff kitchen's microwave. It was on the far side of a plasterboard wall and ran almost continuously from twelve to one. The access point was moved six metres and set to a different channel in an afternoon. The complaints stopped the same week. The cost was eleven tickets and eleven restarts of an innocent access point. There were about thirty hours of a design office working on their phones at lunch for a quarter, plus the contractor's half-day. The lesson Freya kept was that a Wi-Fi complaint with a time of day in it is a clue, not a coincidence. A walk-test at twelve-fifteen would have found the microwave in a week.

The Wi-Fi Complaint Checklist

The checklist below is the ticket's internal note for any complaint where the branch is wireless. Section A captures the pattern in the requester's words with a time and a headcount. B is the device side, done first because it is cheapest. C is the walk-test with the known-good device, one line per spot. D is the reading of the shape. E is the two messages and the handoff. Fill A and B at the desk, C on the route, and D and E before anything on the network is touched. It is joke-free and the lines are short so that it survives being pasted into a plain-text note.

WI-FI COMPLAINT — CHECKLIST
Ticket: ________   Requester: ______________   Floor / room / desk: ______________
Time of complaint: ________   People in the room at the time: ____

A. THE PATTERN (requester's words, then checked)
   Slow for: this person only [ ]  this desk for anyone [ ]  this room when full [ ]  one wall / area [ ]
   Time of day it happens: ________   Fine at other times: yes [ ]  no [ ]
   Neighbouring desk checked: ______________ result: fine [ ]  also slow [ ]

B. THE DEVICE SIDE (do first; cheapest)
   [ ] Known-good device on the same desk, same task: fine [ ]  also slow [ ]
   [ ] Wireless adapter turned off and on; which access point it is on after: ______________
   [ ] Wireless driver and OS updates current:  yes [ ]  applied [ ]
   [ ] On battery with power saving on the adapter: yes [ ]  no [ ]
   [ ] Docked: yes [ ]  no [ ]   Tried undocked: fine [ ]  also slow [ ]
   [ ] Cable at the same desk: fine [ ]  also slow [ ]   (if also slow -> not a wireless ticket)

C. THE WALK-TEST (known-good device, same task at each spot, time on each line)
   Spot                     Signal     Access point     Task result     Time
   Requester's desk         ______     ______________   ___________     ____
   Halfway to access point  ______     ______________   ___________     ____
   Under the access point   ______     ______________   ___________     ____
   Next room                ______     ______________   ___________     ____
   Wired port               n/a        n/a              ___________     ____

D. THE SHAPE
   Weak at the desk, strong under the access point ............ coverage
   Strong signal, slow only when the room is full .............. congestion
   Known-good fine, requester's device slow at the same spot ... the device
   Slow at a set time along a set area ......................... interference
   Device on a distant access point ............................ clinging

E. MESSAGES AND HANDOFF
   Workaround given today (cable, seat, wired port): ______________
   Reply to requester sent at: ________ (plain words, no radio)
   Internal note to network owner with C and D: sent at ________  to ______________
   Linked to other tickets from the same area: ______________
   Nothing on the network touched: [ ]   or: ______________

Hamid's three tickets closed differently. Claire got a cable and a note to the network owner about the corner. The meeting room got the presenter on the wired port and a line in the capacity file saying twenty-two people. Ravi got his adapter turned off and on, and a driver update. He also got a sentence explaining that his laptop had been loyal to the wrong access point since the lift. Three tickets, one floor, one word, three answers, and not one of them a restart of anything in the ceiling.

Frequently Asked Questions