2021 in review: Ads safety in Microsoft Advertising

2021 continued to witness significant growth in digital footprint. We saw an increase in the number of businesses setting up online services and in the adoption of digital products and services by consumers to stay safe and maintain social distance as the pandemic continued. This explosive growth in digital transactions also caught the attention of bad actors, and online attacks and digital scams rose across all platforms.

Microsoft Advertising witnessed an increased number of threats and varied attack patterns in 2021; in our commitment towards offering our users safe and trusted experiences, we invested in product, algorithm, and process-based enhancement and improvements.

Looking back

In the past year, some of our main actions included:

• Taking down more than 3 billion ads for various policy violations, almost twice as many as in 2020. We suspended nearly 270,000 accounts and banned nearly 400,000 websites from our network.
• Making use of significant advancements in artificial intelligence such as zero-shot learning to quickly adapt to new patterns and methods used by bad actors.
• Ensuring that our protection mechanism involved coverage for all types of content such as text, images, and videos to quickly detect malicious activity in our system.
• Making investments in multilingual models to detect non-compliant content in more than 30 languages.
• Making advancements in our human moderation workflows to capture more insights from reviews, continuously improving our systems.
• Leveraging intelligent tools to allow our human reviewers to establish linkages between various accounts and discover fraud rings quickly and efficiently.

Key focus areas

Advertiser identity verification. In our continued efforts to make digital advertising safer, we introduced the advertiser verification program. The selected advertisers are required to establish their identity as a business or as an individual by submitting all necessary information and documents. This helps us weed out bad actors from our advertising ecosystem and stop the spread of misleading and egregious ads at the root. This program is currently available in seven countries, and we are working towards expanding its scope. Learn more about the advertiser identity verification.

Extending trust and safety checks for new markets. Microsoft Advertising has recently launched in 29 countries spanning Europe and Africa, along with extending support for 19 new languages. To ensure a safe and trusted user experience, we have extended our policy coverage and scaled all our existing checks and processes in context of these new markets and languages.

Misleading ads. We observed continued malicious advertising behaviors in 2021. While there were many new patterns, fake celebrity-endorsed investment ads saw an uptick, and we took specific measures to avoid promoting them in a nearly real-time manner. With the help of our advanced machine learning, anomaly detection models, and fraud investigation experts’ insights, we were able to suspend nearly 10,000 accounts and take down more than 200,000 ads. We will continue to invest more in this area to further deter scam investment schemes through ads.

Advertiser safety. Account takeover attempts in 2021 became more sophisticated, and different techniques were used to get past the multi-factor authentication (MFA) challenge. MFA is a security measure that requires users to verify their identity in two different ways. For example, entering a password along with a code sent to the phone. Therefore, we made use of advertisers’ behavioral identity signals to detect such scenarios and protect our customers. We then shared best practices with our advertisers to help them not fall prey to new tactics.

Our approach

Advertising fraud is fast-moving, and we continue to see new patterns surface globally. We take an all-hands-on-deck approach to deliver the highest quality content possible, constantly updating and refining our policies to ensure we meet evolving needs. Our fraud detection technology uses a wide variety of signals and the latest machine-learning algorithms to find fraud patterns that can otherwise be difficult to detect. Our geographically distributed team of experts works round-the-clock using smart and scalable tools to help us conduct detailed investigations into any new patterns we see. Detecting fraud before it has a chance to reach customers is an important part of our approach.

We also address escalations and complaints from customers to quickly remove low-quality ads. In 2021, we received ~70,000 complaints related to ads not being compliant with our advertising policies. We investigated each of them and found ~60% of the reported ads to be in violation of Microsoft Advertising policies. Most of them were related to trademark infringements. A few complaints related to unlicensed gambling sites, phishing, unauthorized government service provider websites, and other user safety issues. Our highly responsive operations team works 24/7 to promptly address concerns relating to our ads. In response to complaints, they took down nearly 650,000 violating ads from our network.

Here is a tabular representation of the addressed complaints:

Going forward

As always, the start of a new year presents opportunities for taking our quality efforts even further. We'll continue to leverage the interdisciplinary efforts of experts in machine learning and system design to scale our technology, alongside the domain expertise of policy and review specialists to deter bad actors. Despite all our efforts, scammers might still evade our checks. If you see a bad or suspicious ad, please report it through our Microsoft Advertising ad quality escalation form.

We value your feedback and partnership, and we'll continue to strive to keep the Microsoft Advertising ecosystem safe.