Upcoming change that will affect stats
Baldy
Registered Users, Super Moderators Posts: 2,853 moderator
Hey everyone,
Possibly as soon as the first of July, we’re going to make a major change in the way images and videos are delivered, and we thought we’d give you some advanced notice because it affects how stats are reported.
Your images and videos are stored at Amazon, but we have used Akamai for years to deliver photos and videos from Amazon to you. Akamai distributes your images via tens of thousands of servers around the world, keeping frequently requested images on a server near you for faster delivery.
But now we’re moving away from Akamai to another provider. We’re doing it for several reasons, chief among them being much faster and smoother delivery of video, and more advanced security. We're hopeful, and our testing seems to show, that image delivery should be somewhat faster for most people depending on where they live. It's possible that we may see somewhat slower delivery for some people because Akamai had great worldwide coverage.
But in a world of trade-offs, we’re losing the ability to discern between user views of your photos, and visitor views. The way we had been working with Akamai, we could detect when you were browsing your own photos and not count them in the stats.
We’re very sorry we won’t be able to do that anymore when we make this switch, and for some people subtracting out owner views was a great feature, but this is a tradeoff we felt we had to make for the speed and security gains we’re seeing in our testing.
Let us know if you have questions!
Thanks,
Baldy
Possibly as soon as the first of July, we’re going to make a major change in the way images and videos are delivered, and we thought we’d give you some advanced notice because it affects how stats are reported.
Your images and videos are stored at Amazon, but we have used Akamai for years to deliver photos and videos from Amazon to you. Akamai distributes your images via tens of thousands of servers around the world, keeping frequently requested images on a server near you for faster delivery.
But now we’re moving away from Akamai to another provider. We’re doing it for several reasons, chief among them being much faster and smoother delivery of video, and more advanced security. We're hopeful, and our testing seems to show, that image delivery should be somewhat faster for most people depending on where they live. It's possible that we may see somewhat slower delivery for some people because Akamai had great worldwide coverage.
But in a world of trade-offs, we’re losing the ability to discern between user views of your photos, and visitor views. The way we had been working with Akamai, we could detect when you were browsing your own photos and not count them in the stats.
We’re very sorry we won’t be able to do that anymore when we make this switch, and for some people subtracting out owner views was a great feature, but this is a tradeoff we felt we had to make for the speed and security gains we’re seeing in our testing.
Let us know if you have questions!
Thanks,
Baldy
0
Comments
My Website index | My Blog
The use case I'm familiar with is a photographer puts boudoir photos in a gallery, then sees there was a view or two and emails the heroes to find out who it was, but it was them. We'd love to solve for that use case like we did before, but tradeoffs.
Thanks for the heads up.
--- Denise
Musings & ramblings at https://denisegoldberg.blogspot.com
If i run the slideshow on my own computer... will the photos counted as views?
But it went so well and our Euro heroes reported such great results for both images and video (especially video) that we haven't yet stopped the test. So as of 9:15 Pacific Time, it's live with your images and videos. We can switch back in an instant if we find an issue.
Let us know if you see any problems but so far, the help desk has been silent about it.
Yes, the first view of each photo should count as a view. But if you left it running for days straight, your browser would (in theory) have cached the photos and only that first view would matter. So no matter how many times the slideshow cycled, you would have only generated a single view on each photo.
Does that make sense?
be loading much faster.
http://www.photosbyat.com/MyKeywords/Bird-Videos
The only nit is some (gallery view) run pixelized, not the highest resolution. In the caption I have the
the lightbox view link which shows a much larger video. But it seems I do not get a good resolution
without manually changing the size. Any thoughts? I'm afraid visitors will never see this size changing.
One other thing, it seemed like none of videos where cached and had to reload from start each re-run.
My Website index | My Blog
Yeah, videos in particular should get much faster. We put a LOT of work into the video streaming architecture in this switch, though it is a tough problem and we are definitely keeping an eye on performance. I really appreciate the feedback.
The auto-sizing should work a lot better, but we have a bunch of a improvements we are working on for it. For one, we have improved the way we encode videos for our new architecture, so newer videos will likely snap to a higher resolution quicker/better. But even for older videos we can do a lot better. We have improvements in the works for the video player to do a better job of getting you to a higher resolution quicker. We had a deadline with the CDN switch, so the video player improvements weren't a requirement for this change. They are coming though.
Regardless of other improvements coming, right now videos should be significantly faster and more reliable than they were yesterday. The resolution changing should also be quicker and more reliable, we just want to make it even MORE quick and reliable.
As for caching, I know our engineers/ops were chatting with CloudFlare late into the night last night on making sure videos get cached properly at local nodes. I don't know the status of that unfortunately, but it is also actively being looked at.
Thanks again, I really appreciate the feedback on this. It was not a trivial change.
I honestly don't know anything more than that, but I will update you guys if I hear anything good or bad.
Don't worry, we *definitely* understand how crucial stats are to you guys.
width. Changing size would vary video size. I have not seen this for a few years? though. Looks like all here
now open in lightbox at ~930x525 no mater what resolution is selected so missing the large HD size.
24" monitor FF. Selecting 480 > 1080p HD all same size.
Edit: lightbox is still much larger then gallery view so really miss the click to lightbox for videos.
Adding /A to the link for me helps but visitors don't have a clue.
My Website index | My Blog
What are you using to get to lightbox? The only way I'm aware of from a video gallery view is with keyboard shortcuts: http://help.smugmug.com/customer/portal/articles/84540-are-there-keyboard-shortcuts-for-smugmug-
If you type "a" you'll get the screen-filling resolution (the same /A you get for lightboxes when clicking on a gallery image)
My hand is on the mouse and the other is usually away from the keyboard so hitting "a" is an extra
movement. But why has Smug stolen the cntl and alt "a" makes no sense? I used to use alt-a to
select everything to check for hidden page items. Good for troubleshootin'.
... and visitors have no clue the "a" key works. But they are use to clicking a photo.
My Website index | My Blog
My Website index | My Blog
Yeah! I agree completely. Now that the streaming video delivery is faster and way, way more consistent, getting some smart logic into picking startup resolution should be a lot easier.
Facebook
Google+
Twitter
Photo Blog
Since I often have to edit in between uploads and am uploading multiple versions of the same images I have to review my galleries on multiple platforms as I post and make changes. Everything is now skewed and pointless. Before I could share a private gallery to a client or to someone I trusted and know exactly when it has been viewed. I could email images to a potential business partner and know that they did in fact see specific images. I could tell exactly which images were the most popular in a quick glance and see what size images folks were being drawn to for a certain gallery. I could manage and garner how much traffic I am directing from my posts on social media, simply by viewing the days traffic after I make a post of a link to an unlisted gallery that had not been shared elsewhere yet. Now it's all completely useless! As a member since 2008 this is a feature I would say I'm the most upset to see go. I know I can start using external traffic tracking and have already set up an account at statscounter but for as much as I am paying a year for you to host my photos I don't think I should have to go outside my service to another website to get ACCURATE image view stats. I also don't know exactly what kind of detailed information they will provide for specific image views.
This could very well be a dealbreaker for me. Since this was just implemented and I was completely unaware of this removal until today I'm quite positive you will be getting a lot of feedback from users who like me relied heavily on the stats feature. Frustrated and saddened by this.
I apologize for the confusion regarding this on your account. Can you please send me a link to the specific gallery you are referring to? Please feel free to PM it to me.
Thanks you.
Facebook
Google+
Twitter
Photo Blog
The new way isn't a deal breaker but it is a consideration.
The old way was useful to me in the way you describe using 'stats'.
The new way adds nothing over other services.
I am trying to decide on service, now.
Cost for Smugmug isn't in annual fee. It is in percent per earnings.
Sending you one of the recent links now, but the one that had client views has had the images deleted since they were just a copy for approval. The other gallery has a similar amount of image views. I still have it password protected though, and it shouldn't have had any public views. However since Ive had an image "featured" while its been protected anyone can use the thumbnail image of that password protected image to bring up the other size images (the owner has made available) of that particular photo just by changing the html link. I'm fairly certain that no other images can be accessed though by the public.
I have the link and I am investigating. Any further galleries you experience this with, please send over to me. Thanks!
Facebook
Google+
Twitter
Photo Blog
Just an FYI, I have a conversation going with my Stats developer and hoping to get some more detailed information as to how/when stats are counted. I personally haven't tested this myself, nor do I ever look at my stats, so I can't really give a good explanation, but I am working on it for you.
Will update you all as soon as I can, most likely sometime on Monday afternoon/evening.
Facebook
Google+
Twitter
Photo Blog
I guess I am saying that as of right now and for some time in the past, the statistics are not accurate at all. In the past I had always been impressed with the rapid replies from Smugmug and their fixing problems. Currently that isn’t happening.
Jerry
They can't be compared at all. One counts image views, the other pageviews.
Please check out my gallery of customisations for the New SmugMug, more to come!
OK, thx for the reply.
First of all, I'm very sorry this caught so many people by surprise. I'm afraid I'm the culprit because the engineers warned me ahead of time that the downside of our new image delivery network is not being able to count owner views. I thought that would be okay given the performance gains, not thinking through all the ways owners could rack up big, unexpected image views and the confusion it could cause.
Here's a common scenario: you create a new gallery and make it private. You upload 300 images using our uploader. You click Done and it dumps you into the gallery. It's only natural that most people wouldn't think of the Done button as generating a lot of image views, but suddenly we just served up a lot of images to you. How many? Well that depends first on the gallery style. If it's one of the styles that puts all images on a single page, it could be a lot, depending on whether you scroll and the width and resolution of your monitor.
Adding to the count, we 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 and get a bigger one, but to us it's an image we sent to your browser and for better or worse, we count it. We do this because we want you to see it instantly and be pleased by the speed.
Sometimes we get emails or see posts showing crazy image views, wondering what kind of bugs we have. We've chased a fair number of these down and, so far, we really are reporting images delivered to the browser, it's just that they are coming in unexpected ways like I outlined above.
At this point you're probably thinking, what kind of idiots...why would they do this...now what for my workflow?
Since I'm that guy, I'll explain what I was thinking. We have millions of customers and they have customers and speed is crucial to probably all of them. And with this change we could get nice speed gains for most customers. And, sadly, we have less than 1,000 customers who've looked at their stats since we made this change. Maybe we bury them too deeply, or they're not that important to most people, I don't know. But in the land of tradeoffs, I was doing the math between making millions happy and hundreds sad (although I didn't know how sad at the time).
Now that I just pissed some of those hundreds off with the previous paragraph, what's to do about your workflows? We're looking into the possibility of excluding owner views with our new delivery network. I honestly don't know how likely it is that we'll succeed or how long it would take, only that we're working on it and serious about it.
Some of you might ask, why would we be serious about it? Millions are happy, a few hundred are mad... Is Baldy just saying that to appease us?
And the answer is we invest over a million dollars a year collecting, storing and displaying stats. It's really a shame to make them useless to a significant number of the people who really care about them, many of whom are among our best customers. Although I often wonder if we can't simplify stats like other sites do to make the problem easier (and stats more popular).
I hope this helps.
Thanks,
Baldy