Speed tests are run on good networks. Product demos are run on good networks. Almost every number a connection product publishes was produced on a line where nothing goes wrong.
Nobody reads the news on that line.
People read on the subway, in a queue, in the last row of a stadium, on the road out of town. That is where the network loses data on the way, and where an ordinary connection does what it has always done: wait, try again, wait longer, give up.
What we measure instead
We put the same phone through the same network with controlled losses and short blackouts, always with and without Lagbuster, against the same site. Then we go outside with real phones on real mobile networks. Every number on this site carries the identifier of the run that produced it.
What that changes
It changes what we can say. We cannot say “faster”, because on a good network Lagbuster is not, and we publish that toll. We can say “keeps working”, because that is what the runs show: pages that load with six of every seven packets lost, a download that finishes on its own through two six-second blackouts, a speed test that completes with four of every five packets lost.
It also changes what we build. When the benchmark is the subway, whatever only shines in the lab does not make it into the app.