Why WiFi Is Slow in Just One Room

The router is in the living room, the WiFi bars look fine everywhere, and the back bedroom still buffers. This is one of the most common WiFi complaints there is, and it usually comes down to one thing: signal bars and actual speed are not the same measurement, and a wall most people don’t think twice about can be costing far more than they’d guess.

Signal is not the same thing as speed

Your phone’s signal bars are a coarse, roughly-logarithmic summary of one number — received signal strength, measured in dBm — and they don’t tell you what that number actually buys you in throughput. A connection can show “full bars” while still being far too weak for its band’s higher data rates, because the bars UI is deliberately forgiving; it’s built to avoid alarming you, not to report precisely. The number that actually matters is the dBm reading itself, and the throughput it implies changes in real, discrete steps as that number gets weaker — this is the whole reason a room can “have signal” and still stream badly.

What one wall costs

Every wall a signal passes through subtracts a real, measurable amount of strength, in decibels — and different materials subtract wildly different amounts. This site’s own simulator uses these per-wall figures, in the spirit of the COST-231 multi-wall model used in real RF engineering: glass costs about 2 dB, drywall about 3 dB, wood about 4 dB, brick about 6 dB, and concrete about 10 dB, each for a 10cm-thick wall at 2.4 GHz. One interior drywall partition is nearly nothing. Two or three of them add up. And decibels are logarithmic, not linear — every additional 3 dB or so roughly halves the signal power reaching the other side, so three drywall walls in a row don’t just add inconvenience, they can genuinely cut a connection from strong to unusable.

What the wrong wall costs

Metal is the outlier, and it’s a big one: roughly 26 dB for the same 10cm reference thickness — an order of magnitude worse than drywall. A metal wall stud pattern, a mirror-backed closet door, a refrigerator, an HVAC duct run inside a wall, even a lot of foil-backed insulation, can each behave like this. This is why a room that’s the same physical distance from the router as another, equally-clean-looking room, can perform completely differently: it’s not really about distance, it’s about what happens to be between the router and the receiver, and one metal obstacle can undo what several ordinary walls wouldn’t.

The 5 GHz trade

5 GHz doesn’t travel through the same obstacles as well as 2.4 GHz — real-world measurements consistently show it losing more per wall, and this site’s simulator applies a 1.4× multiplier on the same base per-material loss (1.8× on 6 GHz), a deliberately conservative estimate rather than a lab measurement. That’s a real cost. But it’s paired with a real benefit: at the same signal strength, 5 GHz’s throughput ceiling in this site’s own model runs three to six times higher than 2.4 GHz’s, and the multiple grows as the signal gets stronger — this site’s throughput ladder puts a strong 5 GHz signal at up to 600 Mbps where 2.4 GHz tops out around 100. The trade is real: 5 GHz is faster where it reaches, and it reaches less far through the same obstacles. Which band wins in a given room depends entirely on what’s actually between the router and that room — which is exactly the question a floor plan answers and a signal-bars icon cannot.

It compounds, and it compounds differently by band

None of this happens in isolation — a signal on its way to a back bedroom might cross a hallway wall, a closet wall, and a bathroom wall in sequence, and each one subtracts its own loss on top of the last. Three ordinary drywall partitions in a row cost roughly the same as one brick wall; a drywall partition plus a brick wall plus a metal-backed appliance can very plausibly cost more than the entire rest of the distance to the router combined. And because adding decibels is the same thing as multiplying the power loss, the order and count of obstacles matters as much as the raw distance — which is exactly why two rooms the same number of steps from the router can perform completely differently, and why “just move closer” is often the wrong advice when “go around, not through, that wall” would have been the right one.

People and large pieces of furniture absorb some signal too, though far less dramatically than a wall — a body sitting directly between a laptop and the router measurably shifts a live signal reading, which is part of why real-world signal strength always has some jitter even when nothing about the floor plan has changed. It’s a second-order effect next to the walls themselves, but it’s part of why two readings taken minutes apart from the same spot are rarely identical, and why one bad measurement is worth double-checking before you conclude a fix didn’t work.

Diagnosing it without walking around with a phone

Walking a phone around a house, watching the bars, is slow, imprecise, and tells you almost nothing about why one spot is weak — you get one number per point, and no way to see what’s causing it. A room-by-room simulation does the opposite: draw your actual walls, mark what they’re made of, place your router where it really sits, and see the full coverage picture at once — including the parts you’d never have thought to walk over and check, and including why a given spot is weak, not just that it is. The two floors this site’s estimates are built around are worth knowing regardless of which tool you use to find them: −67 dBm is the usable floor, below which throughput drops sharply; −82 dBm is a dead zone, where there’s effectively no usable signal left. Everything in between those two numbers is where a wall count, a material choice, or a router repositioning can make the difference between a room that works and a room that doesn’t — and it’s worth saying plainly that any tool doing this from a floor plan, this one included, is running an estimate from a physics model, not a measurement of your actual walls, which vary in moisture, thickness and construction in ways no simulator sees.