The Amazon Attribution beta program has given a select few agencies and brands a window into which ad campaigns drive consumers to fill their digital shopping carts. The trial, which opened late last year, is closed to newcomers for now, but when it opens widely, brands need to know what it can tell them about where to spend their ad dollars.
"What they're trying to do is give insights into what media off Amazon drives the most sales on Amazon," says Mike Seiler, director of search and shopper marketing at AKQA.
Attribution is an obsession for digital marketers, who need to understand where their ads run, which ones were viewed or clicked, and at what point a consumer took an action.
Amazon is the largest online retailer by revenue, and although most online attribution models can tell the advertiser when a sale happens on their own websites, Amazon wanted the ability to tell brands when ads lead to sales on Amazon's site. Without Amazon's new ad technology, the logic goes, marketers are advertising blind: They don't know if a person who just bought their health shakes on Amazon learned about those products in a Google search or from a display ad on a website.
"There has been an arms race for attribution for quite some time," says George Manas, president of Resolution Media, which is owned by Omnicom. "Google has bought up quite a few attribution solutions. Facebook bought Atlas, which had an attribution tool, and now Amazon has gotten in the game."
For the attribution beta trial, Amazon invited a small group of brands and their agencies—representing consumer goods, apparel, food and other categories—to participate. Those brands and outside agencies were able to apply a pixel to ads that direct consumers to Amazon. When a consumer views a video ad or clicks on a display ad leading them to the product on Amazon, the pixel records where the consumer saw the ad. The pixel can suggest which ads were most effective and provide basic data about what type of consumers responded to the ads.
Amazon has shared one case study so far, from Premier Nutrition, which used the pixel on search, display and video ads. The brand found that running ads on health-related websites was most effective, targeting consumers interested in sports and fitness.
Amazon provides the analysis to brands in real time, so they can see what channels are performing best and adjust ad campaigns accordingly. The reports disclose when a person exposed to an ad viewed a product detail page, put an item in a digital shopping cart and when they bought a product.
Brands are managing an increasingly large number of pixels, tags and cookies to track their campaigns online, on mobile devices and laptops and in apps. They need a disparate network of tracking tools to get a clear understanding of how effective their marketing is because walled gardens such as Facebook, Amazon and Google don't make all their data available to each other. What's more, cookies only work when people are on their desktop web browsers; they don't work well with mobile devices.
"Tagging itself becomes a bit of an operational burden," Manas says. "The world has gotten fragmented, and you have to implement multiple tags in multiple places for multiple parts of a campaign."
And brands are tracking more than just attribution. They measure ad viewability, ad frequency, email open rates and more. To manage all this tagging activity, advertisers use the programs Google Tag Manager and Signal to organize and deploy all their tracking technology.
Amazon's attribution program does have shortcomings—it cannot operate fully when ads run on Facebook, according to a digital ad executive at a major agency, who participated in the beta tests and spoke on condition of anonymity.
More than 90 percent of Facebook's ad business is on its mobile app. However, Amazon, a clear rival, is blocked from Facebook's mobile software developer kit, which is the platform developers use to build products and services connected to the social network.
A consumer scrolling through the Facebook app might see an ad, but if they buy the product on Amazon, the advertiser can't tell if that person ever saw the ad on Facebook.
"We are recommending that brands don't even tag social campaigns," the anonymous agency exec says. "There are too many limitations to not being able to attribute mobile in-app sales."
Amazon's attribution pixel is also handcuffed on Facebook and Google in that it only detects when a person clicked on an ad in those ecosystems; it can't detect when a consumer simply sees the ad. It is limited to "click-based" attribution and does not perform "impression-based" attribution in those environments. If a person searches Google for a product, for instance, but doesn't click an ad, Amazon can't see how that search could have influenced a purchase on its site.
Brands and agencies use software services like Conversion Logic and Visual IQ, owned by Nielsen, to sift through the strands of data from their multitude of tags and pixels. The programs look for patterns in ad campaigns that run through Google, Facebook and Amazon. Taken all together, the data might present a complete picture of what strategy works best.
"When you have blind spots, you just have to model around them," says Manas. "You build models that fill in the gaps in attribution."