Google Opens Kimono (Slightly) On AdWords Screening Measures

Followers of news about Google AdWords know that problematic errors — instances in which ads that violate the company’s own advertising standards slip by — come up quite regularly.

Most recently, Google, along with advertiser ProFlowers, was roundly and publicly criticized for AdWords that seemed to indicate that local florists were sold out in advance of Valentine’s Day — when they actually were not. Back in January, a BBC report indicated that Google had admitted to running, and collecting money from, illegal Olympics ticket sellers in the United Kingdom.

The company also recently paid a $500 million fine to the U.S. Justice Department, after admitting to taking illegal pharmaceutical ads over several years. And the list of potentially problematic issues goes on and on: mortgage scams, illegal gambling ads, trademark issues (which differ from geography to geography), and even consumer advisory policies (from the Federal Trade Commission and similar agencies around the world).

How To Police Such A Vast And Complex System?

Each time one of these scandals has surfaced, the questions arise: How in the world does Google even begin to police the multi-millions of AdWords being placed each day all around the world? Do these public screw-ups lead to internal hand-wringing and staffing up, in hopes of avoiding such situations in the future?

“User quality and user trust in our ads is something that we take very very seriously as a company,” Sridhar Ramaswamy, SVP of engineering, told me. “Our success and our livelihood is really dependent on that trust. This is something that is taken very very seriously.”

Behind The Scenes With The Ads Quality Team

Though Google has always been fairly tight-lipped about its methods, believing that (similar to click-fraud investigations) bad actors would simply learn from whatever it revealed, the company today posted a blog entry and video designed to open the kimono somewhat on this mysterious aspect of Google’s operations.

The company wouldn’t reveal just how many staffers it has working on ad quality, but said “hundreds” of people — engineers, folks who craft policies, and people who manually review ads — are on the front lines of keeping bad ads away. Most of those folks are based at the company’s Mountain View headquarters, but others are in Pittsburgh, Dublin and in India, allowing the company to respond more quickly — in different time zones — to problems when they crop up. Personnel in different locales also helps with crafting appropriate policies that comply with laws in different countries.

“There is a very local element to policy formulation,” Ramaswamy said.

The engineers are responsible for developing the algorithms that either reject ads out of hand or flag them for manual review. Ads are rejected, or accounts closed entirely, when the account has been opened with a stolen credit card, or when past account activity has been identified as malicious. Activitiy coming from reputable advertisers’ accounts that raise no red flags are usually automatically approved, unless the content somehow seems (to the algorithms at least) to violate policies around counterfeiting, illegal activities, etc. Those ads get reviewed by human editors, who decide whether the ads pass muster, or should be rejected. Others review escalated cases — when an advertiser says an ad has been incorrectly rejected.

Recent Changes To Beef Up The Safeguards

Google today also said it has “recently” implemented additional measures to better police AdWords.

The company has improved its “query watch” for ads for counterfeit goods, widening proactive monitoring of sensitive keywords and queries related to these items. The goal, of course, is to catch these ads before they appear on Google.

Additionally, a new “risk model,” is in place to technologically screen ads and determine the probability of whether a particular ad violates policies. The company says this new system is more precise.

The systems and processes for manual reviews has also been sped up, Google says. The company says it is now committed to reviewing any ads that receive complaints within 24 hours.

The company says its improvement of systems is reaping rewards, reducing the percentage of bad ads by more than 50% from 2010 to 2011. More than 130 million ads (out of “billions” submitted) were disabled last year, and 150,000 accounts were shut down for attempting to advertise counterfeit goods.

“We see our job as one that is ever changing. We track metrics for things like counterfeiting very very very closely and we are proud of the work that we’ve done there,” Ramaswamy said. “There is a tremendous amount of money riding on this. This is ongoing work and we are committed to making our ad system better.”

What’s The Incentive For Google?

A difficult challenge for Google  is the perception that — because it collects money from these policy- or law-violating advertisers — the company has little incentive to shut them down. For example, in the recent settlement over pharmaceutical ads, the $500 million fine was arrived at by calculating what Google and the illegal pharmacies had gained.

Ramaswamy argues that Google isn’t profiting as much as some may think, saying many bad actors use stolen credit cards, meaning Google doesn’t actually reap any reward. More importantly, he says, allowing bad ads to get through jeopardizes the company’s relationship with users and with legitimate advertisers.

“Google benefits enormously in the long term by our users loving and trusting our ads and our good name,” he said.  “That’s the culture that permeates. Short-term gains that lead to the erosion of trust are not sustainable. “