Attribution is the next best thing in digital media. Using real-time performance data on digital campaigns, attribution technology can assess which ads on which sites, formats and channels (smartphone, tablet or desktop) are most effective in driving conversions.
“Consumers have no less than four devices on average that they use to interact with any given brand or product,” said Stephen Yap, Google‘s head of data & measurement platforms. “The consumer no longer follows a linear path to engage with a brand. The path they’re on is very chaotic — it’s more like a squiggly line with lots of blind spots for advertisers.”
In the always-connected world, understanding the consumer’s path to purchase is even more important.
“It becomes very confusing very quickly as to what actually gets credit for the consumer engagement or ultimate credit,” said Yap. “Would it be the store where I ultimately bought [the product]? Would it be the review I looked up on my phone? Would it be my desktop, where I started to look for locations to buy the product?”
In steps attribution measurement. Attribution technology can tell you your 30-second spot on broadcaster X’s mobile site drives more sales at a more cost-efficient rate than, say, that 300×250 banner on publisher Y’s desktop page. Getting that information in real-time allows you to shift a larger share of spend the broadcaster’s mobile channel, while the campaign is still in market.
Most attribution relies on so-called “last-touch” methodology, which gives all the credit for each conversion to the last ad the consumer saw or clicked on. But we all know that’s not really how advertising works — consumers may see a pile of different messages before they buy. Say the consumer watched that broadcaster’s video and then the next day was served a banner on another site — the banner would get the credit, even if the video did the convincing.
Recently, attribution providers have begun offering “multi-touch” attribution, which doles out credit to all the ads the consumer sees on their path to purchase. A simple multi-touch attribution scheme will break down credit evenly, or given higher weighting to the first and last ads the consumer sees.
The problem, said Yap, is that even in those models, the technology is making an assumption about how each ad should score — rather than measuring how much influence it actually has.
“There’s really no mathematical foundation for that. It’s more of a hunch,” he said. “You’re saying, ‘Mmm, I think I did a really good in job in video, so I’m going to weight it a little bit more.'”
In May, Google acquired Adometry, a company focused on developing advanced statistical models for attribution. Like other advanced attribution specialist vendors, Adometry measures credit by comparing the likelihood a consumer will purchase if they’ve seen a series of ads to the likelihood they’ll purchase if they’ve seen the same series of ads, minus one. The difference between those probabilities yields the relative credit that the missing ad should receive.
By iterating this process, Google can measure the influence that each ad has on the consumer, relative to the rest of their journey. The system still isn’t perfect — there will be some aspects that attribution will never be able to model, Yap said, but it is as close as we’ve come to understanding how ads actually lead to sales.
The one challenge with advanced analytics, versus last-touch or multi-touch, is that it requires a large number of conversion and therefore a very large number of impressions. That means it’s hard to score attribution in real-time, since ads have to have been in the field a long-time before the stats will be significant. Real-time attribution is key to media mix modeling, where marketers can use algorithms to optimize their spend based on performance while the campaign’s in market.
But Yap said it’s a little too early for marketers to put their eggs in the media mix basket anyway. Right now there’s still a lot more work to do on easier elements of digital advertising, like dynamic creative. Until we can serve a digital ad that actually engages users, shifting minute shares of spend between sites will have a less-than-significant effect.
