random access memory lane

In the last century, this is how we did things. I was working for the predecessor of Scoota, Tangozebra. In those days, most people connected to the Internet by connecting a piece of string to a baked bean tin and connecting the other end of the string to an RJ11 cable, then chucking the whole lot over a telephone line and hoping for the best. Something like that – the details are hazy, it was a hundred years ago. The dizzying 28k download speeds that could be achieved were unable to cope with much more than a heavily-pixellated 8-bit jpeg, or a 4-bit animated gif that would sit there, flashing “ME! ME! ME!” in a sort of monochrome existentialist primal scream.

We were pretty sure we could do better than that, and had developed a java-based system that could deliver very small audio and image files, which were knitted together on the client side into something greater than the sum of its parts.  A tiny configuration file described how they should be played back, and how they should be dealt with in the event that one of the constituent parts should be held up en route – a Java applet did the rest. This was pretty nifty and won lots of awards and a good chunk of funding, but the winds of digital change were blowing and Java was hoofed out of the client-side like yesterday’s chip-wrapper. So we turned our attention to the new kid on the block – Javascript. This allowed us to do what we cool kids called DHTML.

Rich Media didn’t exist, at least, not in the way it’s understood now. We started to propose ways of building creatives that would be more imaginative and, we hoped, more impactful, without resorting to a forced intrusion into the end-user’s private digital world. I remember, in around 1999, building a small physics engine in Javascript for the sole purpose of making some gifs of a Loch Ness Monster swim convincingly around the screen, and a golden eagle take off from a page element, fly about a bit, squawk, then land on another page element, revealing a call-to-action. These ads had extraordinary engagements rates – and because they were hand-crafted to each page they were served to, they never got in the way of the user’s browsing. The eventual call-to-action was revealed in a standard ad unit on completion of the animation, which reduced the likelihood of accidental clicks, so we were fairly confident that people were actively engaged – no other, similar ad slot was giving anything like these response rates, accidental or not. We served the ad into five or six sites, each receiving a big chunk of detailed attention – I wrote bespoke paths for Nessie and the eagle for every site. By the end of it, I felt like I could enter them both into Crufts.

Such were the early days of rich media – soon afterwards, Flash cracked transparency and became the favoured way of delivering overlaid formats, albeit at the cost of gargantuan memory-gobbling. Formats sprouted from every element and the world went rich-media crazy. But it wasn’t all plain sailing. There were still barriers that made things tricky – the vast majority of users had narrow bandwidth connections, or hopelessly old browsers, or antediluvian operating systems, or any combination thereof. Websites were still thrown together with a sort of cavalier insouciance that meant you never knew what to expect when you delivered something into them, and quite often the DOM was completely broken and the browser would just throw up its hands and guess. And, from the advertisers point of view, the whole model was fundamentally broken. Ads were charged per impression – a nebulous term that was really much closer to page view than ad view. This was quite enough for standard formats, which were usually a direct image call anyway – an ad tag was just an anchor tag wrapped round an image tag. Rich media necessarily separated the tag impression from the view – you can’t have something that, on mouseover, sends a dazzling aerial penguin display looping the loop from one ad slot to the other without some sort of up-front payload. Inevitably, agencies and brands cottoned on to the fact that a substantial proportion of the ads they were paying for were never displayed, so the model switched from page view to ad view, and CPMs were shifted accordingly to protect margins.

By about 2005, operating systems, browsers and hardware had all caught up and were at a general level of sophistication and widespread adoption for the landscape to stabilise for a bit – you could be pretty sure that the vast majority of end users could receive the optimum experience for every format in your arsenal. This meant that volumes and throughput could increase steadily, and the creative juices really started to flow. Designers started to push the formats as far as they could, and some really innovative and exciting work started to emerge. This was a happy time – things chuntered along cheerfully and all we had to worry about was whether we could outdo each other in the creative application of technology. Then Steve Jobs ballsed it all up.

In 2007, the iPhone was released, and everyone decided the Internet was something they had to have readily to hand 24/7 and nothing else would do. Previous attempts to bring the World Wide Web to mobile devices had been laughably inadequate (WAP! HAHAHA!), and almost universally ignored. But suddenly phones were where it was at. Phones that looked gorgeous, but had hardware, software and connectivity that threw us back ten years. There were some good points – universally strong Javascript support for example, and some that were a mix of good and bad, depending on who you were, like the hot-potato dropping of Flash. Overall, though, it meant everything had to be reinvented – how sites are built and displayed, how ads work, how users are tracked and understood. This process is still ongoing and will probably continue to cause headaches for developers and designers the world over. It was around then that Tangozebra was absorbed into the bosom of the great behemoth that is Google. We could stand back and watch the unfolding circling and snorting of the tech giants as they jostled for the world.

The battle of the phones ensured that advances in their capabilities were pushed hard, and it wasn’t long before they could handle surprisingly demanding tasks – although the obsession with the single-page app did rather too often define the boundary of client-side capability (and still does, by God).

Meanwhile, data was being gathered by the exabyte, and the wherewithal emerged to process that data within a meaningful timeframe and have it available for rapid retrieval. So now we could target to the user and not the channel, and programmatic was born. But, of course, it was built around standard formats, like ad systems of yore, and rich media just wasn’t going to fit. The problem with rich media is that it’s complicated. Really, really complicated. It used to be relatively straightforward, when 98% of the world was on the same version of Windows and used one of two browsers. But now there are more platforms and devices than you can shake a stick at, a bazillion versions of everything (damn you, Continuous Release), and everything is shifting, dynamic and hilariously overcrowded. With programmatic, you’ve at best a limited idea of where your tag might end up, and when it gets there, what it’s going to have to deal with. A lot of rich media formats require full access to the DOM, but you’re serving from a different domain, quite possibly into at least one level’s depth of iframes, which makes for all manner of security hoop-jumping. Sophisticated inter-communicating systems have to be created to overcome these significant challenges, and that usually means close collaboration with individual publishers. Our Sentinel system has emerged after many years’ work to link all the variables together and make sense of them.

We’ve designed Sentinel to greatly extend the standard functionality of the tag. When we deliver any creative, we’re actively gaining as much environmental information as we can and feeding it back to the Rig, constantly updating our understanding of the programmatic landscape, and enabling the system to make decisions about where to focus attention for optimal performance. We’re continually building format-by-format profiles of supply so we know what to target where, thereby minimising waste and making sure the users we’re targeting have the best possible experience. Using Sentinel, we’re ironing out the remaining glitches that make serving programmatic rich media tricky. Smooth, reliable, scalable rich media will give a stable platform upon which the creatives can go to town – like they could ten years ago. The combination of innovative, tailored creative and precise, self-managing targeting has the possibility to transform brand engagement and the user experience. The current crop of rich formats is just the beginning. A programmatic approach to delivering content that is beyond the standard issue formats, that adapts to any number of environments, and that gives huge creative freedom, could really usher in an exciting new era where digital advertising is more than just a picture in a rectangle that people occasionally accidentally click on.