40 kilobytes? Are you kidding?
Years ago, when bandwidth was a tiny fraction of what it is today, the ad industry codified a set of standards for banner ad file sizes. A common limit was 40kb (sometimes even 30kb) including all images, fonts, animations and scripts which Flash compressed into a single amazingly small swf file. Technically each publisher determines its own file size policies, but almost everyone looks to the IAB (Interactive Advertising Bureau) as a standards-setting body, like the W3C for web browsers. The IAB exists to help guide the industry but they don't mandate or enforce anything. When Flash ruled the banner ad landscape, certain file size specs were recommended by the IAB and the system worked well. However, the technology landscape has changed drastically.
Since 2008, average bandwidth has grown by a factor of 5.6 which is remarkably on-pace with the growth of the average web page size (5.7), but the IAB has been cautious about declaring HTML5 specs due to all the complexities involved. They released a set of HTML5 guidelines in 2013, but omitted any file size specs, saying only that HTML5 ads weigh "more" than swf ads. Without specs, many publishers clung to the safe limits of yesteryear. The gatekeepers who impose the 40kb budgets often do not have the authority or wherewithal to allow more than what the latest IAB spec dictates. Consequently, developers are forced to shoehorn HTML5 banners into archaic Flash specs which isn't what the IAB intended. This must change. From our vantage point, fear is driving the industry. Publishers and networks are afraid to raise the file size limits without IAB approval. Some do it anyway, but disagree on exactly how much, leading to wild variations. Developers have no choice but to build for the least common denominator in their ad campaign which is either totally unclear or ends up being the dreaded creativity-crushing 40kb. (UPDATE: The IAB released a draft of its new HTML5 specs.)
HTML5 is fundamentally different...embrace that
HTML5 banners often weigh 3-5 times as much as a Flash swf but far too many people myopically focus on the aggregate total file size. They miss the unique strengths of HTML5 technology that we should be exploiting - shared resources and browser caching. These have a tremendous impact on loading time and overall performance which is the whole point of the file size limits anyway! Flash compiled all assets into a single swf meaning that if 10 different banners on a site all used a certain library, it got baked into each and every swf. End users paid the file size price 10 times. Multiply that by millions of ads and it gets pretty crazy. In HTML5, however, a library can be dropped onto a CDN (content delivery network) and shared among all banners, thus end users only load it once and it’s completely "free" thereafter...for all ads pointing at that CDN...on all sites. This is a BIG deal. It means that common animation chores like the requestAnimationFrame loop, timing, sequencing, intelligent GPU layerizing, lag smoothing, compatibility workarounds, performance optimization, etc. can be extracted and shared among them all (much like what the Flash Player did for swf files). The unique banner-specific code can be much more concise, reducing overall load times and improving performance. File size limitations should be applied to the banner-specific assets, excluding the shared resources that drive common functionality. Imagine how silly it would have been if the 17MB Flash Player download was included in the aggregate file size for each swf banner. Ad networks and publishers can put a certain subset of tested-and-approved libraries onto their CDNs and exempt them from file size calculations. We're thrilled to see industry leaders like Advertising.com/AOL, Google DoubleClick, Flashtalking, and Sizmek already taking this approach with GSAP. This strategy allows developers to avoid burning hours manually cooking up their own proprietary libraries to fit within the ad specs. Ad networks and publishers win because load times (and costs) are lowered and it's easier to troubleshoot problems when a common toolset is used. They reap the benefits of all the compatibility and performance optimizations in tools like GSAP. End users get ads that perform better, load faster, and look more appealing.
Animation technologies and approaches
For those tasked with building HTML5 banners, the choices are perplexing. Is it best to use a visual IDE like Adobe Edge Animate, Google Web Designer, or Tumult Hype? Even Flash is capable of outputting HTML5 content. These tools can make building ads easier (especially for designers who don’t want to write code), but a common complaint is that the resulting output is bloated and slow, making them ill-suited for banner ads. Some networks explicitly state that they won't accept ads built with these tools. We'd love to see the visual tools mature and export concise, performant, ad-friendly code because plenty of designers aren't comfortable hand-coding banners yet. Ideally, they'd tap into GSAP under the hood so that designers and developers could collaborate on the same files without worrying about runtime redundancies. There are also network-specific banner-building tools but their proprietary nature makes them impractical for many campaigns. If an agency uses one network’s proprietary tool and then their client asks to run the ad on another network too, it must be rebuilt. Learning how to use each network's proprietary tool can be cumbersome. Hand-coded animations are usually much lighter-weight, performant, and universally accepted, but building them requires a particular skill set. And which underlying technologies should be used? CSS animations? jQuery? GSAP? CreateJS? Once again, answers vary wildly among ad networks and publishers. The goal of this article isn't to provide an in-depth review or comparison of the various tools. Each has its own strengths and weaknesses, but let's briefly touch on some of the major runtime animation technologies:
- jQuery - it was never intended to be a robust animation tool, so jQuery suffers from poor performance and workflow issues. Most ad networks strongly advise against using it. GSAP is up to 20x faster. Additional reading: http://greensock.com/jquery/
- Zepto - like a lightweight version of jQuery that uses CSS transitions under the hood for animations. Zepto is better than jQuery for banners, but it suffers from similar workflow issues as well as the inconsistencies/bugs inherent in CSS transitions/animations (like with SVG transforms). Active development seems to have stalled. Additional reading: http://zeptojs.com
- Web Animations - a new spec being worked on that has a lot of promise, but it just isn't a realistic contender at this point because it is in flux and several browser vendors remain noncommittal about ever supporting it. The polyfill has performance problems. Additional reading: http://w3c.github.io/web-animations
- GSAP - Widely recognized as the performance leader, GSAP solves all kinds of real-world animation problems from browser inconsistencies to workflow headaches (far too many to go into here). The Flash banner ad community is full of designers and developers who use GSAP daily, making it much easier to transition to HTML5; no new syntax to learn. Ongoing development and support have a solid track record for over 7 years. Additional reading: http://greensock.com/why-gsap/
Based on our experience and the results from our survey, we suggest the following:
Ideally, the IAB would equip the community with a short list of recommended libraries that get CDN-ified and exempted from file size calculations. Historically, the IAB has been extremely reluctant to officially endorse any third party tools. That's understandable - it could be seen as playing favorites or unfairly excluding someone's favorite library. However, without specific recommendations, the HTML5 landscape is so fractured and complex that it will result in a free-for-all (which is basically what it is now). The IAB can set the tone and move the focus away from aggregate total file sizes and into the modern era that leverages shared resources and browser caching to deliver excellent performance. It is imperative that this list of "recommended" libraries be very short, otherwise the caching impact will be diluted. The IAB can run their own independent tests and look at performance, features, compatibility, support, workflow benefits, and overall industry demand to determine which libraries get recommended. Of course we feel strongly that GSAP belongs on that list because:
- It is the top performer.
- It has widespread industry acceptance, both in Flash and HTML5. It's recommended by Google, used by the biggest brands in the world, etc.
- It is framework-agnostic, super flexible and robust, able to animate anything.
- It is professionally supported, yet free to use in banner ads.
Modernize file size specs
Given the 5.6x growth factor of bandwidth and page size since 2008, it seems entirely reasonable to adjust the old 40kb limit to 200kb (5x) for the modern HTML5 era. This is entirely consistent with some in-depth testing that has been done recently aimed at identifying the file size threshold at which real-world users perceive a dip in performance. The results showed that the threshold was upwards of 250kb. Combined file size isn't the only issue that contributes to slow load times; the number of server requests can have a significant impact. A single 300kb file can often load faster than 200kb split among 20 files. HTML5 banners can't realistically mash everything into one file, though. Doing so would kill the benefits of caching and resource sharing. So a reasonable compromise seems to be a 10-file maximum. Sprite sheets can be used to combine images. Given all the factors, we'd recommend the following for standard (non-rich media) ads:
- 200kb combined total (gzipped)
- Maximum of 10 files. Any additional must be loaded "politely" (after the parent page finishes loading)
- Shared CDN resources like GSAP don't count toward these totals.
Some have suggested slicing the 200kb standard limit into two parts - a 50kb initial load, and then the rest "politely" loads. However, we advise against this for standard (non-rich media) ads because it unnecessarily complicates the design and production process as well as QA and enforcement. Rich media ads will likely require more files and kb than the limits mentioned above, and those should be polite-loaded. By "rich media", we mean ads that contain video or expand or perform API calls (like feeding the viewer's zip code to a backend script), etc.
Update documentation and guidelines
It is surprisingly difficult to get answers to some of the most basic questions when preparing a banner ad campaign for even the biggest networks and publishers. What are the file size limits? Which libraries can be used? Do CDN resources count against the total file size? Is there a network-specific CDN link for common libraries? Online docs either have outdated information or none at all related to HTML5.
Drop support for IE8
Legacy IE support is not just painful for developers, it's exceedingly expensive for advertisers. Certain effects are outright impossible, so creatives must learn about the IE8 pitfalls and adjust their designs. Developers are forced to rebuild entire portions, implement workarounds and perform extra testing, all to accommodate a tiny fraction of the web audience who probably don't represent the demographic that advertisers are targeting anyway. This was never an issue for Flash, but it's a HUGE issue for HTML5 because it relies on native browser technologies that are absent from older browsers like IE8. Our recommendation is to draw a line in the sand and drop support for IE8 for sure, and potentially even IE9.
Consider SVG instead of iframes
Displaying ads inside an iframe is nice for security, but it forces ads into a strict rectangular space (ruling out fancy overlays with transparency/mask effects that show the main web page behind) and there's a performance price too. SVG is widely supported and it has some excellent transparency/masking capabilities, plus it can serve as a single container for an entire ad (see Chris Gannon's blog post and video)! Further testing needs to be done to better understand the performance and security implications, but it certainly seems like a worthwhile contender.
Create a gallery of sample banners and templates
Rather than pouring over specs and instructions and then building something from scratch, most developers prefer to analyze banners that already conform to the standards and use one as a template for their own project. Each network has different API's and ways you must track clicks, etc., so it would be lovely if each one provided a gallery of demos at each standard size. Codepen.io is a great place to host a collection because it's so easy to see (and edit) the HTML, CSS, and JS as well as the result all in one place. Developers can simply click the "fork" button and start producing their own version of that banner immediately in the browser. Codepen even integrates nicely with crossbrowsertesting.com for easy QA.
Adjust client expectations
As the industry transitions from Flash to HTML5, clients must be made aware of the design, budget, and schedule implications. HTML5 banners take more time to produce and test, therefore they will be more expensive. Plus there are certain effects that were easy in Flash but are virtually impossible in HTML5, so creative expectations need to be adjusted as well.
Common GreenSock Questions
With the broader discussion out of the way, let's narrow our focus to GreenSock for a moment and address some of the most frequently asked questions:
Which networks support GSAP?
All networks that we're aware of allow GSAP, and most even exempt its file size from the ads and host it on their CDNs. Google DoubleClick recommends GSAP for complex animations. Here's a breakdown of how some of the major players stack up:
|Allows GSAP||Excludes GSAP from file size calculation*||Hosts GSAP on CDN|
TweenMax is too big! Where's TweenNano?
Let's face it: TweenMax (the most robust tool in the GSAP suite) is overkill for many banners that are only doing simple fades and movement. Wouldn't it be smart for GreenSock to create a super-small animation engine that's targeted at banners and only has the basic features? In the Flash days, we did exactly that and called it "TweenNano". It weighed about 2kb. On the surface, it sounds like a great idea but there are several reasons we avoided TweenNano in the HTML5 toolset:
- Performance - GSAP has been engineered with a huge priority on performance which sometimes comes with a file size tradeoff. We could accomplish the same tasks with less code in places, but runtime performance would suffer. We feel strongly that when it comes to animation, it's wiser to pay a small up-front kb tax (only a fraction of a second in most cases) in order to get maximum runtime performance. Animations must look smooth and conserve battery power. Think of it this way: would you rather buy a computer that boots up 2 seconds faster or one that's 30% faster all the time (after it boots)?
- Flexibility/Creativity - what if you want to animate a non-essential CSS property like boxShadow or slide along a curve or scrub through a timeline? Even if there's just one part of your banner that needs a more advanced feature, it presents a dilemma. Creativity is hampered. Again, the fraction of a second one-time cost difference for TweenMax seems well worth it for the added flexibility and peace of mind.
- API confusion - years ago, Adobe created a lightweight version of the Flash Player dubbed "Flash Lite" with similar aspirations (bake only the essentials into a lighter weight flavor), but it was a complete failure. One of the problems was that developers couldn't remember which features were available in the regular Flash Player versus Flash Lite. Likewise, TweenNano's feature disparity would create some confusion/frustration.
What about creating a tool that lets users select only the features they need, and then it spits out a customized stripped-down version of TweenMax? Again, this sounds appealing, but it would likely lead to worse load times because instead of having one common TweenMax that gets shared and cached, every banner would have its own different (and partially redundant) flavor to load. Ultimately, we're committed to delivering the tools that are needed most, so if the broader industry decides not to leverage shared resources and publishers insist on sticking to all-inclusive aggregate file size totals, we're open to creating TweenNano. Luckily, it looks like there's excellent momentum behind TweenMax getting CDN-ified and exempted from file size limits. In our opinion, that's definitely the smartest approach.
What's so special about GSAP?
It's beyond the scope of this article to explain all the benefits of using GSAP; see http://greensock.com/why-gsap/ for a summary. If you're still wondering what the big deal is, we'd encourage you to find someone who is proficient with it and ask about their experience. Usually people who take the time to learn it have a "light bulb" moment pretty quickly and never want to go back to using other libraries or CSS. It's difficult to explain to the uninitiated - lists of features don't really do it justice. It's not merely about performance (although that's a biggie) - it's about feeling empowered to animate almost anything you can imagine with minimal code.
Do I need a commercial license to use GSAP in banner ads?
GreenSock's standard "no charge" license covers usage in banner ads even if you get paid a one-time fee to produce the banners. We fully encourage the use of GSAP in banner ads and beyond. You may want to check out Club GreenSock for some bonus plugins that allow you to easily achieve advanced effects.
Is anyone building a GUI for GSAP?
A visual tool for building GSAP-based animations is a popular request, and we have been approached by several large and small companies about the possibilities, but there's nothing rock solid to report yet. We hope that companies like Adobe and Google will offer export options from their tools that leverage GSAP as the runtime engine and produce well-formatted, concise code. There's a pretty neat tool called Animachine that's in alpha and can be installed as a Chrome extension. It shows promise, but isn’t entirely stable at this point. There are also several online GSAP-based banner builders: http://html5maker.com/, https://tweenui.com/, and http://www.loxiastudio.com.
Where can I get GSAP training?
You can have GreenSock come directly to your organization and sit with your team to get them up to speed quickly. We can even convert one of your Flash banners and then teach you how we did it which is an excellent way to learn banner-specific tricks. The Q&A sessions are invaluable. We have limited slots available, though, so contact us as soon as possible to get your event scheduled. There are plenty of other learning resources available:
- GreenSock's getting started video/article
- GreenSock's learning resources
- New GreenSock eBook (published by Noble Desktop)
- Lynda.com course
- ihatetomatoes.net course (intermediate/advanced)
- Noble Desktop class in NYC
- 02Geek course
The GreenSock forums are a fantastic place to not only ask your question(s), but also poke around and see what others are saying. It's one of the best places to learn even if you never ask a question. There are plenty of demos on codepen.io as well. For inspiration, we'd suggest following these people: