There's also this misconception that speed is the only thing that matters, so "higher speed = better connection" and the vast majority of consumer-grade tools only ever test for this. This is where it will mislead you, as an unstable connection with short bursts of high speed will appear "better" (despite being fragile and completely unusable for anything real-time) than a slower but more reliable connection. In-browser speed tests are extremely bad for this because all the buffering & various layers could even fool the test code itself, making it believe it's getting a steady stream while in reality it's getting merely short bursts of data (in between tons of underlying TCP retransmissions). Iperf & ping are the tools of choice if you actually want to look into it - they are closer to the metal and will give you faster feedback (you will be able to actually see the dropouts due to packet loss).
Mesh systems are a step up and can work but due to their cost (and also the fact it's usually hard to tell how well it'll perform without actually buying it and trying it out in the field) I would very much recommend just biting the bullet and doing it once and well with wired access points. A stop-gap solution would be to use powerline adapters to provide the Ethernet backhaul to the access points; you'll be limited in terms of bandwidth (mine top out at ~100Mbps in my current property, but went to ~300 in my previous one) but latency and packet loss-wise they're rock-solid and won't hog the precious Wi-Fi spectrum.
Forget consumer-grade crap as well, go for enterprise-grade equipment or at the very least "prosumer" grade such as Ubiquiti.