Those studies all come from companies trying to sell you something. They seemed to come from when mobile phones were much slower when there was a valid use for

Those studies all come from companies trying to sell you something. They seemed to come from when mobile phones were much slower when there was a valid use for AMP pages. A/B test data is very easy to cherry pick from and you can see a 10% increase in conversation based on random chance. I can see my marketing team misinterpret data like that and use it for promotion. I am skeptical. I improved the initial loading speed of a website so it was 30% faster on an old mobile phone and it made no difference in conversation based on the analytics. Most people used faster phones with faster internet that it really didn't matter. After the first page load, most of the bloated assets were cached so even those with slow connections were relatively fast navigating further. There could be some types of websites where speed matters more like if you promoted click bait articles with high bounce rates. But if you have high quality content, it isn't going to be that important unless you have really terrible bloat.