UPDATED 09:00 EDT / MAY 23 2017

BIG DATA

Google applies machine learning to help marketers better track ad impact

Are my ads working?

That’s the essential question that every company faces with its marketing. Today, Google Inc. is aiming to help them find out how their ads and other marketing are working with a new service unveiled today at its annual Marketing Next conference in San Francisco.

Google Attribution, as the company has named the new service with its usual blandness, promises to provide a better way to track the results of marketing across devices from personal computers to smartphones and across channels from search to store visits.

The idea behind Attribution, which is offered for free, is that customers looking to buy something engage with companies in a dozen or more distinct interactions today — search, social networks, apps, websites and display and video ads, not to mention television and in-store visits. Current methods of tracking how people are influenced are limited, often losing the plot between devices, for instance. That’s why most companies have simply attributed a “conversion” or sale to the last time someone clicked, thus missing many of the contact points that actually led to the sale.

“Existing attribution tools… just aren’t cutting it,” Bill Kee, Google’s group product manager for attribution, told some 1,000 marketers at the conference today.

Although details were sparse, Attribution integrates information from several Google ad services — its AdWords ad serving system, Google Analytics and DoubleClick Search system for search ads — to provide a more complete picture of how they move the needle with target customers. In particular, Attribution allows marketers to use what it calls data-driven attribution, which uses machine learning to analyze sales or conversion data and calculate the actual contribution of each step the consumer takes, from the first brand impression to the last click before a purchase.

Naturally, Google said this works best when used with other Google ad tools such as AdWords and DoubleClick Search. It’s not clear to what extent Attribution can track or use data with other tools. However, an already available paid version called Attribution 360 that grew out of its 2014 acquisition of Adometry Inc., starting at $150,000 a year, includes several other features for larger or more sophisticated marketers and ad agencies: integration with DoubleClick Campaign Manager, custom data integrations so marketers can use their own data and attribution to television ads.

Google also previewed a store sales management product. Kee told SiliconANGLE that Google uses machine learning to build models based on analyzing groups of people who, for example, clicked on various types of ads and then bought something — or didn’t. The model can analyze the effects of clicking on a search ad, seeing a display ad, watching a TV ad and various combinations of those actions and, with data from retailers, their impact on store visits and ultimate purchases, thus determining which types of ads lead to what types of purchases. That way, the relative effectiveness of each ad in the stream that consumers see can be weighted on their effectiveness, allowing marketers to shift spending to ads that are more effective at the right point in time.

It’s far from a perfect measure, Kee conceded, but better than many current approaches. “It’s a model and all models are wrong, but they can be useful,” he said.

The matching of marketing touchpoint and purchase data naturally raises privacy concerns, which Google acknowledged in a short press conference. Jerry Dischler, vice president of product management for Google search ads, said the company doesn’t use actual names of people in the matching. Google has what he called an “opaque box” of data on its own users’ actions, and marketers have their own opaque box of purchase, location and other customer data provided through third-party data brokers that Google’s system can match anonymously to determine cause-and-effect. But he said Google can’t see the retailers’ data — and even then, can see only the amount spent, not details on individual transactions — and the retailers can’t see Google’s data.

Privacy advocates such as the Electronic Frontier Foundation remain skeptical about how anonymous the data actually is, especially since Google isn’t revealing much about how its system works.

Attribution, which is now in beta test mode, will roll out to “more advertisers” in the next few months, Google said.

Image: Google

Image: Google

Given Google Chief Executive Sundar Pichai’s single-minded emphasis on making the search giant into an “AI-first” company ever since he took over the top job in 2015, it should be no surprise that artificial intelligence and machine learning are central to Attribution and several other new services being introduced today.

“This technology is critical to helping marketers analyze countless signals in real time and reach consumers with more useful ads at the right moments,” Sridhar Ramaswamy (pictured), Google’s senior vice president of ads and commerce, said in a blog post before the conference. “Machine learning is also key to measuring the consumer journeys that now span multiple devices and channels across both the digital and physical worlds.”

Google isn’t the only one using machine learning for marketing. Nearly every company today is touting its AI chops. “Google’s renewed advocacy for upper-funnel marketing will make more revenue if more advertisers focus less on the last click before purchase, and start focusing more on top-funnel activity, namely display ads, Chaitanya Chandrasekar, co-founder and CEO of the predictive ad management firm QuanticMind, told SiliconANGLE in an email. “But what’s really fascinating to me is the company’s increasing push for machine learning to navigate an increasingly fragmented customer journey.”

Google also emphasized the role of machine learning in improving its store visits measurement system introduced in 2014. Since then, it has sought to convince retailers that it can help them tie ads to store visits, but it has been a challenging process.

“Our recent upgrade to deep learning models enables us to train on larger data sets and measure more store visits in challenging scenarios with greater confidence,” Ramaswamy said. That includes, he added, visits in multi-story malls or dense cities such as Tokyo and São Paulo.

Store visits soon will be available for YouTube video campaigns in addition to current availability in search, Google Shopping and display ad campaigns. Moreover, in the next few months, Google will roll out store sales measurement by device and ad campaign, enabling retailers to track in-store revenue from the ads.

Finally, Google is bringing the ability to find people who are believed to be in the market for a particular product, based on products they’re researching, a method it calls in-market audiences, to search ads. So a car dealership, for instance, could reach people who have searched for “SUVs with best gas mileage” and “spacious SUVs.”

This feature, once again, uses machine learning to divine purchase intent more accurately by analyzing trillions of search queries and activity across millions of websites.

The full keynote address with more details is now up on YouTube:

Photo: Robert Hof

A message from John Furrier, co-founder of SiliconANGLE:

Your vote of support is important to us and it helps us keep the content FREE.

One click below supports our mission to provide free, deep, and relevant content.  

Join our community on YouTube

Join the community that includes more than 15,000 #CubeAlumni experts, including Amazon.com CEO Andy Jassy, Dell Technologies founder and CEO Michael Dell, Intel CEO Pat Gelsinger, and many more luminaries and experts.

“TheCUBE is an important partner to the industry. You guys really are a part of our events and we really appreciate you coming and I know people appreciate the content you create as well” – Andy Jassy

THANK YOU