Raghavan was speaking at Advertising Week Europe in his first public appearance since taking over the company’s main source of revenue. His words to the industry: Google is listening.
Ultimately, the promise is that with advances in AI, which will touch several elements of Google’s advertising offer, greater sophistication will allow the company to focus on the little bits of important data that can help to make advertising more effective without compromising privacy.
“For years, Google has been waging a comprehensive global campaign against fraud through a combination of technology, policy, and people,” Raghavan told the audience, before pointing to Google’s part in dismantling a multi-million dollar ad fraud scheme. On its own platform, Google has also worked to remove bad actors, having taken down 2.3 billion bad ads in 2018, in addition to removing a million bad ad accounts.
The impact of digital advertising on the election landscape is also spurring change at the search giant. “We made changes to how we verify advertisers, especially in the critical arena of elections. It’s absolutely essential that you protect the fundamental institutions of democracy,” said Raghavan.
In response, Google is implementing a new policy for the upcoming European Parliament elections where “advertisers must be verified as EU-based”, he said, adding that the company would provide disclosures to make it absolutely clear who is paying for each ad. The policy will come into force in the next few days.
YouTube has been a particular problem for the company in recent years, first with brand safety fears surrounding the appearance of reputable brands’ ads on inappropriate YouTube content. It has also faced ongoing problems, beginning in 2017, and resurfacing earlier this year, over concerns that the comment section on videos was attracting paedophiles.
To these concerns, Raghavan explained that Google’s machine learning capabilities were helping to remove inappropriate content. “Take-down times have come down dramatically,” he said. “Today, over 83% of videos taken down for extremism are taken down before a single human has flagged them.”
The teams are manually reviewing videos in order to provide more complete training data for its machine learning classifier that removes violent content. To tackle inappropriate content around minors, YouTube adopted an immediate strategy to tackle this content, including deleting accounts and channels, as well as reporting illegal activities to authorities. It has removed the comments on tens of millions of videos featuring minors.
On a more positive note, Raghavan also talked about the application of machine learning to enable advertisers to manage their advertising across search, display, and video advertising.
But the real leap forward is in measurement, and bringing simplicity to this advertiser and agency headache. “Over the past year, we’ve significantly advanced management integration with trusted third parties. We now have a community of measurement partners that meet a very high bar in the implementation of privacy-safe technologies.”
Some of these partners, for which the company is trying to broaden and deepen its geographical reach, have allowed Google to offer impressive new capabilities. With greater investment on YouTube, advertisers are keen to measure sales lift off the back of campaigns there. Working with Nielsen Catalina Solutions in the US, Google was able to offer insight in this area. It is planning to expand this offer to Europe.
Raghavan also had a message for consumers, addressing some of the common assumptions about the ways in which big tech companies use customer data in their advertising. People feel that advertising is creepy, and tend to overestimate the quantity and intrusiveness of data that Google takes, Raghavan claimed.
Ultimately, he said, Google handles some of its customers’ most private pieces of information, from emails and files to photos. “That data is sacrosanct,” he stated. “It’s private data: secure it, lock it up … we keep it separate from advertising.
“Users also trust us with data from their activity, and we use that to improve their experience. We do not pass the data on to advertisers,” he said. “In fact, we use virtually none of it for monetization.” How much comes under “virtually none” was not made clear.
“I constantly challenge my team with this question,” he continued: “what’s the least amount of data with which you can provide magical and trustworthy advertising experiences.
“We always want to do more for users, we want to do it with less data, we want users to be informed and in control. We want to be very upfront with what we do not do: we do not sell personal information to anyone ever.”
Neither, he said, does Google use device fingerprinting to identify people by device – a technique that was made more complex by the introduction of GDPR legislation last year. In a dig at Facebook, which has come under fire for allowing advertisers to target by race or ethnicity, Raghavan said Google does not target “sensitive categories” or serve personalised ads to those under 13.
Sourced from WARC, Engadget, EFF, ProPublica