Variation in no. of request count at document complete

Every time i ran the test, total number of requests in document complete vary.
Variation can be any number between 55 to 120. As you know few the parameters result also varies accordingly like speed index. I dont believe my lazy loading feature has any impact when webpagetest is executing test

What actually this mean? How this should be analysed? I am not able consolidate any logic for it.

Why do you not believe your lazy load logic is having any impact?

Do you have sample test results? If you look at the waterfalls for both of them it should be easy to tell what the additional requests are and if they are lazyload images.

Lazy loading is not having effect because waterfall is showing the footer(bottom) images are also loaded. And page scroller is large, so i ruled out the possibility of complete page is showing in visible screen. Lazy loading is implemented thru standard pagespeed configuration, so chances for error is almost zero. This all brings me conclusion that webpagespeed test is not considering lazy loading.

All the below request are taken in a day with almost same content

Document Complete request count 116
Document Complete request count 94
Document Complete request count 57 and fully loaded request count us just 87

Thanks in advance

The difference between the two is all thumbnail images so I’m fairly certain it is the lazyload thumbnail images.

Depending on what filters you have installed and enabled pagespeed will use beacons from client-side js to figure out what the visible portions of the page are and use that information to only defer images below the fold. My guess is that the variability comes from that information not always being available (not sure when it gets wiped out or how often it recalculates) but you should probably ask over in the mod_pagespeed discussion group.

You are right these are thumbnail images and most of the images are from bottom.

Below the fold images are not loading when the site open on desktop/mobile. We found this instance only in webpagespeed?

Is this the case, webpagespeed engine crawling the scroll bar? Or may be sometime not. Because they also need to fetch the complete page analysis otherwise it is not like complete analysis?

So what’s your experience, if lazy load is there, these request count should be fixed count(most of time)?

WebPagetest doesn’t crawl. It launches a real browser and just records whatever it does.

One thing you might want to try is to check the “Preserve UA string” in the advanced settings to avoid adding the PTST to the UA string in case mod_pagespeed is tying the above-the-fold logic to the full user agent string (though the IE UA strings vary wildly so that would suck).