SmugMug Statistics update?

redcrown@mchsi.com

Can anyone point me to a comprehensive, current, and accurate description of Smugmug Statistics?

I gave up on Smugmug stats back when the "new" Smugmug was implemented because they were a mess and unuseable. There were lots of issues with owner views being counted. Something about "pre-fetch" causing tons of false hits. Confusion over whether gallery thumbnails were being counted. More controversy about some types of custom pages or slide shows causing unwanted hits.

I just spent some time Googling and searching Dgrin for updates. Looks like the "owner view" issue was resolved, but I can't find anything about the "pre-fetch" or thumbnails. So I did a quick test. Looks to me like thumbnails are still being counted as "small" hits, thus inflating totals. Other aspects of my quick and probably sloppy test left me confused.

So with emphasis on "comprehensive, current, and accurate" - how do Smugmug stats work? What's fixed, what's still broke?

Ferguson

Just as an aside, the answer might vary by style, e.g. collage landscape vs classic. But I'd like to understand that better as well. I THINK that in collage scrolling through the whole gallery shows one hit per image, but in thumbnails only one per larger image displayed (e.g. if it's showing 20 thumbs and 1 image and you go to next screen)?

leftquark

With the exception of mobile apps, all owner views shouldn't count at all. There's a bug in which mobile app views from the owner are being counted that we're aware and have in the queue to fix.

Pre-Fetching refers to the act of sending images to you (or your visitors) before they ask for the image, so that when they do ask for the image, it loads immediately. For example, lets say you're looking at an image in the Lightbox (though this applies to all galleries as well), we load the image you're on, and the next several images. You may pause for a few seconds to view the image ... while you're paused, that "Pre-Fetching" is occurring, grabbing the next several images. When you go to the next image, it's already been retrieved, so it loads immediately. If this didn't happen, it could take some time to load, of which you're just sitting there waiting, with a bad experience. There's some more technical discussion I could get into but that's the gist of what "Pre-Fetching" is.

Pre-Fetched images shouldn't count until the image is actually viewed, but if you are noticing anything different, we can investigate further. [Edit: It's possible I'm wrong about this. I'll talk to Engineering in the morning]

redcrown@mchsi.com

Thanks, leftquark.

Your reply did not answer the thumbnails, but my tests seem to confirm that the small sized thumbnails are not counted in stats when a normal gallery is viewed. I was suspicious because stats on my active galleries always show lots of hits on the small size. I guess that's due to the increasing number of smart phone browsers whose small screens force small image displays in lightbox.

This SmugMug help page:
http://help.smugmug.com/customer/portal/articles/93330-does-smugmug-track-views-of-my-photos
Contains the following:

"We also do our best to predict which images you're most likely to click, so we deliver those to your browser, too. They aren't visible to you until you click on a small image to get a bigger one, but it's an image we sent to your browser so it's counted as a view."

That seems to describe "pre-fetch" and also implies that pre-fetched images are counted in stats even if they are not viewed.

So I ran a more careful test. It was difficult because I discovered that SmugMug does not update the statistics in any predictable way. In some initial tests, results would show up in the statistics within a few minutes. In later tests, the stats did not show up for hours.

But in one test, I created a new "collage landscape" gallery and loaded 100 images. I used the organizer to create and load the gallery, so it was never viewed. Then I used Firefox to view the gallery while not logged in. I scrolled thru the gallery thumbs several times, but never viewed any images in lightbox.

Then I viewed the 1st, the 24th, and the last image in lightbox, closing lightbox each time and using the thumbs to navigate to the next image. StatCounter and Google Analytics both showed the gallery page hits and the 3 individual lightbox page hits immediately. No other hits shown.

SmugMug Statistics showed no records for several hours. Gave up and went to bed. The next day, SmugMug Stats finally showed 4 hits of the "X3" size. These 4 hits included the 3 images I viewed, plus the SECOND image in the gallery, which I did not view. That implies that the second image was pre-fetched and counted. But it does not explain why viewing the 24th image did not cause the 25th image to be pre-fetched and counted.

So I have a theory. Pre-fetched images are counted in statistics, but the pre-fetch logic tries to be smart. If you view images in sequence, it pre-fetches. If you jump randomly (1st image to 24th image) it stops pre-fetching because it has no idea where you will go next.

That would mean the statistics are falsely inflated by pre-fetching, but not that much. If someone bounces around randomly, only sequential views cause a pre-fetch count.

I'll test some more to confirm, but that may take days given the random nature of updating statistics. Plus, there is the question of "mobile" users. SmugMug confirms that owner views are still counted when the mobile app is used. But what about when a mobile device uses a normal browser (Safari, Chrome) instead of the SmugMug mobile app?

BTW, the SmugMug help page for statistics says, "Take a look at our stats feature to see real-time information on how often your photos and videos were viewed." The term "real-time" in that statement is a big stretch of imagination.

redcrown@mchsi.com

Another test done, bizzare results, seems to prove SmugMug Statistics are very inaccurate.

Created a new gallery (using Chrome, logged in)
Loaded 100 images. Again, collage landscape format.

Using Firefox, not logged in:
Viewed the gallery thumbs, scrolled up and down a few times.
Backed out, cleared the Firefox cache, viewed and scrolled the gallery again.

Viewed a total of 14 images in lightbox. A few were random, a few in sequence, a few viewed twice.

Checked Google Analytics, verified the 14 hits were there and no other hits.

SmugMug Statistics showed no records 30 minutes later, but 4 hours later they appeared.

SmugMug showed a total of 357 hits! Of those, 328 were on "small" images and 29 were on "X3" images.

I never viewed any small images, so the 328 hits must have come from thumbnail views. They were spread across all 100 images, but strangely some had 2 hits, some 3, and some 4.

Of the 29 hits on "X3" images, some were on the actual images I viewed, some were images immediately before or after the images I viewed. In one case, SmugMug appeared to pre-fetch 2 images away from the image I viewed. Some of the images I viewed only once showed 2 hits. Looks like SmugMug counted the pre-fetch and the actual view.

Conclusion: SmugMug Statistics are still a false representation of what users actually select to look at. Their only use would be to report inflated counts to audiences.

The statistics could be made useful by eliminating counts on thumbnails and pre-fetched images. Seems a simple thing to do, but who knows. Maybe SmugMug coders have programmed themselves into a corner from which there is no way out. It does seem strange, however, that Google Analytics can get it right from code embedded in SMugMug. Too bad that Google does not show their results in a easy to understand way (no thumbs, just meaningless URL file names).

Ferguson

But Redcrown, in collage landscape I don't think the images are considered thumbnails, but are small images? I.e. scrolling there would be (should be I think?) like viewing each small image?

leftquark

Ferguson wrote: »

But Redcrown, in collage landscape I don't think the images are considered thumbnails, but are small images? I.e. scrolling there would be (should be I think?) like viewing each small image?

Yes, collage landscape should be pulling in smalls, mediums, or larges to fit the gallery display.

redcrown@mchsi.com

OK, more tests. I changed gallery style to "Thumbnails" (squares, which I dislike), browsed the gallery page, viewed 5 random images in lightbox. Still got "small" hits and false pre-fetch hits. Changed to "Collage Portrait" expecting the same and got the same. Didn't try "SmugMug", "Journal", or "Slideshow" styles because I don't ever want to use them. Easy to guess what they will do.

So, conclusion stands. SmugMug statistics are inflated, false, and misleading. The lack of attention given by SMugMug management is understandable. The number of people using the feature appears very low, judging by the few discussions in this and other forums. Thus, the low priority.

I just wonder how many people use statistics but are unaware of the issue. Also, imagine how disappointed people would be if SmugMug fixed the issues and their stats suddenly went way down.

leftquark

redcrown@mchsi.com wrote: »

The lack of attention given by SMugMug management is understandable. The number of people using the feature appears very low, judging by the few discussions in this and other forums.

We do have an intention to update this in the future but at this time, you're correct, we are working on some other higher priority items that we're excited to share with you in the future!

Ferguson

redcrown@mchsi.com wrote: »

So, conclusion stands. SmugMug statistics are inflated, false, and misleading. The lack of attention given by SMugMug management is understandable. The number of people using the feature appears very low, judging by the few discussions in this and other forums. Thus, the low priority.

I pay occasional attention primarily to see what galleries get most hits; a proportional sense not so much because the number itself has meaning.

Here's the issue for me -- I like collage landscape. Depending on your monitor size, you might hit a gallery and decide not to look at any of it, and just recorded 40 images displayed. Someone else may look on an iphone and page through one by one and record 40 images. Clearly one was more engaged than another, but there's no sense of that in the statistics, since a lot or even all the images come up whether you really looked at them or not.

I also haven't looked, but there's a lot of cases were automation may give misleading views. What happens if they hit a slideshow page (my home page is one) and just let it run over night -- if it looks 1000 times through 40 shots, is that 40, or 1000? Prefetches has already been mentioned.

But perhaps most importantly there's no clue given how many PEOPLE viewed your site (unique IP's or cookie tracked or whatever). I'm more interested in knowing if something was widely viewed, especially since I do not attribute much to the actual image view count.

The good news is a lot of additional and very useful information is available on Google Analytics, so I tend to pay more attention to that.

Now that brings up something... and both for current and planned -- is Google Analytics smart enough to accumulate data from your CDN?

redcrown@mchsi.com

Ha! I just found this old thread, now 7+ years old Had forgot about it. Smugmug statistics are still a mess, unfortunately. Notice that above, leftquark (apparently speaking for Adobe) said, "We do have an intention to update this in the future..." Makes you wonder what their vision of the future is.

Ferguson

@redcrown@mchsi.com said:
Ha! I just found this old thread, now 7+ years old Had forgot about it. Smugmug statistics are still a mess, unfortunately. Notice that above, leftquark (apparently speaking for Adobe) said, "We do have an intention to update this in the future..." Makes you wonder what their vision of the future is.

I gave up on SmugMug fixing stuff after the flicker acquisition and some departures of key staff. It works, I still use it, but all their efforts seem focused on the cell phone crowd, frankly.

I just looked at some statistics and my gallery with the most hits -- a test one. I tracked it down and it's because I have a logo in it that I display on the top of each page. They just never put much thought into how to tell actual engagement with users as opposed to counting stuff easy to count. Try google.