Optimizing your site’s ads
Last week, Techcrunch wrote about an ad-optimizing service called YieldBuild. I’ve been reading through YieldBuild’s blog and I think they could really be on to something. Below I’ve listed several nuggets of wisdom from their blog. But first, words from Techcrunch on why you should care:
They’ve been privately beta testing the system with some fairly well known startups and saw improvements of 50% to over 100% in effective CPM rates on user profile pages.
Despite the commonly-held perception that bigger ads are more likely to capture the attention of a viewer and entice him/her to click, our experience using YieldBuild across a number of very different sites paints a complex picture. On some pages, bigger ads perform better. On others, more, smaller ads perform better. Seemingly without rhyme or reason, different visitors respond differently to differently-sized ads on different pages.
Wow, sounds like something the Mad Hatter would say. But it’s true. Some visitors might find larger ads too aggressive and will avoid them. Others need something that stands out from the page’s text to warrant attention. This sort of guesswork is interesting, but thankfully YieldBuild doesn’t require us to really understand the whys of visitors’ behavior with respect to text ads. It just consistently knows the hows of ad formatting and optimization and what it takes to achieve higher clickthroughs and revenue.
It stands to reason that ads that naturally integrate into the general style of a page are more likely to get clicked on, since they suggest they flow with the content of the page.
And often this is the case. On many pages, we’ve been amazed to see YieldBuild slowly but surely, from among the millions of style and placement permutations possible, arrive at a format the blends beautifully into the style of the page, from color, placement and even border style.
But not always.
Quite often, we’ve actually seen the contrary – a style completely incongruous with its surroundings outperforms the blended style. On a background of pastel colors, a bright red ad will turn out to be the best performing. A dull gray ad can surprisingly do better than all other variants on a brightly-colored page.
We here at YieldBuild have seen it happen enough times to know that there is no hard-and-fast rule with respect to format blending and performance. Each page should be optimized by itself, according to its unique traffic blend, to determine whether eye-dazzling, subtle, or a blend provides the greatest receptivity among visitors.
The goal of time-consuming and laborious ad format optimization should be to find the holy grail of ad formats, one whose color, placement and format magically achieve the optimum clickthrough rate and revenue for the page. Right?
Wrong.
What must be one of the strongest testaments to the adage “the whole is greater than the sum of its parts” in this business, YieldBuild has consistently found that a set of different “good” layouts beats a single “great” layout in overall page performance. Although, head to head, any one of these “good” layouts would underperform the best-performing layout, a system where all the “good” layouts cycle on the page actually bests the supposed best layout.
What this means in practice is that once your optimization testing has yielded an array of high-performing layouts, you’ll do yourself a favor by filtering them all into your ad formatting system, rather that performing even more optimization testing to find the supposed standard-bearer.
It would make ad format optimization a lot easier if all you had to do was come up with a set of best-performing ad layouts for a page, and be done with it. If only it were that easy!
The reasons it unfortunately doesn’t work this way are two-fold:
1) A page’s traffic sources typically change over time. Each traffic source’s visitors have varying responses to different ad formats and layouts, depending on what site they had visited before. As the balance of visitors tilts in a new direction, the page’s ad layouts should respond accordingly.
2) For sites that rely on repeat visitors, ad blindness can set in, and without some variation, repeat users will unconsciously ignore ads (even those that worked like gangbusters the previous month), reducing clickthrough. Continually coming up with new, best-performing ad layouts is an undeniable need if your goal is to maintain or increase clickthrough rates.
This might sound more daunting than it is. It does require, however, continuously generating new layout variants, testing to see which perform best, and adding those to the collection of layouts cycled on the page. It also, naturally, requires putting a few former great layouts out to pasture. But the payoff is worth the effort.
If everyone responded to ads the exact same way, then you’d see remarkably little variation in the formats and approaches that sites use to expose their visitors to their advertisers’ messages. You could format your ads to look like everyone else’s and stop worrying about it.
However, it’s been clear from YieldBuild’s own testing that each site’s body of visitors has its own preferences and ad sensitivities, and that is often due to the range of different responses among individual visitors. Where visitors came from to visit a site can often give a glimpse to what ad formats they respond to, and which they ignore.
Think about it. Visitors coming from a boisterous social community site with loud, visual ads (think MySpace) might completely ignore subtly-formatted ads. Likewise, visitors from a text-heavy informational site with muted colors (like Craigslist or Digg) might be put off by obtrusive, wildly-contrasting colors. It all depends on what their eyeballs have become accustomed to, and each visitor’s browsing history paints a different story.
Ideally—and this is, forgive the shameless self-promotion, something YieldBuild already does—you’d correlate incoming traffic source with visitor response to an array of text ad layouts, and then build your ad layout engine to respond appropriately as the mix of traffic sources to a page changes.
If it works, YieldBuild will take all the hassle out of optimizing your ads. Even many large publishers haven’t the time or expertise to continually optimize their ads, so YieldBuild scratches a huge itch. I think they’re going to have fun collecting their 3% commission, but in general their business model is solid. Man, if they could sell this to or partner with someone up and coming like Facebook for instance, wowsers they would be worth a lot immediately. Even without signing a huge partner, the longtail implications for this service are huge.
Another startup in this arena is PubMatic. They have a case study of their partnership with Sportsvite where they claim that “Sportsvite realized an increase in revenue within a few days. The combination of an ad network auction and optimized text ad layout led to an 90% increase in revenue.”
In the end, these optimization services will be purchased and rolled into the big ad networks, and provided to publishers for free. Frankly, I can’t wait. In the mean time however, I’m going to try using YieldBuild’s service for Faceplanting and see what kind of eCPM increases can be had. Even a 20% increase would make their 3% cut worth it.
Be the first to comment
Jump to comment form | comments rss | trackback uri