Many of us complained about our ad coverage getting lower and lower across 2018, causing a major drop in revenue.
I began using a different management system on November 7. They show an LLR on each ad unit (lazy-load-request) rather than Coverage, but they are similar. And I've discovered a direct correlation between a low LLR and responsive banners.
I had a theory that the responsive banners are taking longer to load, possibly because it has up to 5 times as many bidding options before finding the highest bid (and if just one of those take a long time to respond then the entire system breaks). By taking longer to load, I suspect that many of the users simply aren't seeing any banner, so there's no value.
So I began running an A/B/C test on my below-the-fold location (BTF), using a vertical responsive for "A", a fixed 300x600 for "B", and a horizontal responsive banner for "C". On the same page I have a "control" of a fixed 300x250 banner in my above-the-fold location (ATF).
Before the test I had a fixed 300x250 in the BTF location, and the fixed ATF banner and BTF always had a virtually identical LLR, within 1% of another (eg, on Nov 28 the ATF had 25,494 and the BTF 300x250 was 25,445).
For the A/B/C test in PHP, I used:
$btf_ad = { code for vertical responsive banner }
$rand = rand(0, 2);
// if $rand == 0
// automatically show the vertical responsive banner
if ($rand == 1)
$btf_ad = { code for fixed 300x600 }
else if ($rand == 2)
$btf_ad = { code for horizontal responsive banner }
What I discovered with both responsive banners is that the LLR is consistently 27-35% lower than the LLR of the fixed ATF, while the fixed 300x600 BTF banner is exactly what I'm expecting.
So far for today (roughly half the day), the ATF banner (control) has an LLR of 16,745.
With 3 banners in rotation, I would expect each to have an LLR of roughly 33% of the ATF (5,525). But in reality, the results for the test are:
A (vertical responsive): 3,298
B (fixed): 4,854
C (horizontal responsive): 3,322
As you can see, the fixed banner is loading much closer to the projected 5,525, and almost 50% more than the two responsive banners. It's still 700 less than the projected 5,525, but that's much better than how the responsive banners are 2,200 lower than projected!
So my conclusion is that my theory is at least in part correct, and that the responsive banners are directly responsible for a lower LLR. Which in Adsense would translate to a lower coverage.
And if each of those unloaded impressions count towards a pageview but have $0 RPM, then they would lower your Page RPM average proportionately.
I haven't been able to prove my theory that the time to find a high bid is the reason for the delay, though.
So if you're having a problem with low coverage and use responsive banners, I would suggest running a similar test of using your preferred fixed size against a responsive, and see if your coverage and Page RPM improve.
