Harnessing Deep Learning to Revitalize the Business Value of the Open Web

In the global traffic market, the open internet occupies a pivotal position but receives significantly less advertising budget compared to the self-contained "walled gardens" like Google, Meta, and Amazon.

Compared to the fierce competition found inside walled gardens, the open internet remains a vast blue ocean of traffic, with data showing that users are increasingly spending more and more time on the open internet.

So why are advertisers hesitant to allocate more of their budget to the open internet?

The primary reason lies in the persistent stereotypes surrounding the open internet within the industry. Even today, some advertisers perceive the open internet as overly complex, requiring substantial investments of manpower and budget, and carrying higher risks compared to the familiar, straightforward, and secure walled gardens. But are these concerns still valid?

The answer is no. Thanks to the continuous efforts of the advertising industry, the open internet's advertising ecosystem has undergone multiple upgrades, and MediaGo's latest five deep learning models take this a step further, positioning the open internet as a growth channel that advertisers can no longer ignore.

Is The Traffic Quality on The Open Internet Really That Poor?

First, let's address advertisers' biggest concern – traffic quality. The open internet's traffic sources are diverse and fragmented, and the quality of traffic can indeed be inconsistent. Especially in recent years, as the threat of invalid traffic (IVT) has grown, further frustrating advertisers.

On platforms like Facebook or Google, advertisers know exactly how their ads will appear and where they will be displayed. By contrast, on the open internet, a single campaign may span dozens or even hundreds of websites, each with its own unique ad formats and display positions, resulting in variations in traffic quality across different placements.

Therefore, effectively placing ads across the vast open internet requires advanced technology to ensure traffic quality. MediaGo has developed a Traffic Value Assessor based on deep learning algorithms, essentially acting as a traffic quality gatekeeper for advertisers on the open internet.

Through sampling, the Traffic Value Assessor model evaluates hundreds of billions of daily traffic requests, assessing the ad value of each source and filtering out low-quality traffic with poor performance. Only the highest-value traffic enters the bidding model, improving ad performance and ensuring advertisers' interests are well protected.

According to Pixalate's Invalid Traffic and Ad Fraud Benchmarks Report, in Q2 2024, the global open programmatic mobile online advertising market had an IVT rate of 11.6%. However, for high-quality traffic filtered by MediaGo's Traffic Value Assessor, the IVT rate is less than 1%. MediaGo has already helped thousands of advertisers achieve high-quality user growth on the open internet.

Revolutionizing the Business Value of the Open Internet with Deep Learning

On the open internet, hundreds of millions of users click, browse, and interact daily, generating trillions of ad display opportunities. At the same time, advertisers worldwide are sending out massive volumes of ad display requests.

Advertisers are presented with countless media exposure opportunities backed by real users with diverse interests and intentions. On the other hand, there is a vast demand for ad displays across different product categories and price points. It can often feel like an impossible challenge, similar to tossing hundreds of millions of screws and nuts of various sizes into the ocean, hoping to find the right match. This is where deep learning proves its value.

From the earliest days of artificial intelligence to the rise of machine learning, and with MediaGo introducing deep learning into the online advertising industry, technological innovation has continually reshaped industry standards.

When confronted with this challenge of “screws and nuts in the ocean,” traditional AI's computational power can only divide the ocean into regions, rely on human assistance to extract features, and identify suitable matches within each region. In contrast, deep learning, powered by deep neural networks (DNN) and trained on billions of data points, far surpasses traditional AI and machine learning in computational capability. In just milliseconds, it can perform billions of neural network calculations to find the best match across the entire ocean. In such a situation, the advantage provided by deep learning is clear.

Implementing Deep Learning at Every Conversion Stage of the Marketing Funnel

MediaGo is an advertising engine powered by deep learning technology. Since its inception, its primary mission has been to help advertisers improve ad performance through deep learning.

So, how exactly is this achieved? MediaGo's strategy leverages the superior computational power of deep learning to meticulously deconstruct the conversion path.

Let's start with the first step in the conversion path: getting users to notice the ads. We've all browsed the web and noticed the varying effectiveness of ads in different placements. However, these differences are not solely due to location.

Through the OpenRTB protocol, MediaGo partners with premium media to access anonymized contextual data in the real-time bidding stream. Armed with this data and the robust computational capabilities of deep learning, MediaGo can swiftly analyze the attention levels of different ad placements. This enables us to help advertisers secure the ad slots that are most likely to capture users' attention. This is the essence of MediaGo's Attention Prediction Model, which can enhance ad campaign exposure efficiency by an average of 20%.

Beyond visibility, the next critical questions are: Will users click on the ads after seeing them? And will they convert after clicking? Ads, as a form of commercial content, are fundamentally similar to regular media content. By leveraging the recommendation logic of media content, we can better align ads with users' reading interests.

To enhance ad effectiveness, we start with interactive data—clicks, reading duration, completions, and conversions. These metrics help us assess content quality, user interest, and intent, thereby measuring the alignment between ads and users. Combined with the contextual media data, ads are placed in the right spots and delivered to the right users.

MediaGo has developed Interest Prediction and Intention Prediction models using deep learning technology, with click-through rate (CTR) and conversion rate (CVR) as the core optimization metrics. These models have helped hundreds of millions of ad campaigns achieve an average increase of 15% in click-through rates (CTR) and 40% in conversion rates (CVR).

From capturing attention to driving clicks and completing conversions, each precise recommendation is executed in milliseconds. MediaGo has achieved a new balance between user experience and business value, bringing added value to the open internet.

Cost and Effectiveness Are Not a Zero-Sum Game

Filtering out high-quality traffic and accurately predicting user behavior are crucial steps that ultimately converge on the final stage: bidding.

In today's online advertising landscape, SmartBid—leveraging algorithms to optimize ad bids—has become a standard practice. Campaign goals are diverse, ranging from impressions and clicks to purchases and downloads, leading to a variety of smart bidding strategies.

Through extensive industry research, MediaGo has found that regardless of the goals, cost and effectiveness remain the central concerns for advertisers.

Consider the Max Conversion mode, which prioritizes maximizing conversion volume to drive significant uplifts in a shorter timeframe. However, akin to competing for a limited supply of sought-after goods, securing a larger share often necessitates higher bids, leading to increased costs—an outcome that advertisers strive to avoid.

Addressing this dilemma, MediaGo has innovatively introduced the Target Cost Per Action (TCPA) metric in the Max Conversion mode as an auxiliary constraint to regulate each bid. This approach ensures that ad campaigns achieve maximum conversions while maintaining stable ad costs, thereby optimizing advertisers' ROI.

Since its launch, the Max Conversion mode has been adopted by nearly half of MediaGo's advertisers, yielding remarkable results. For instance, lead-generation advertiser Peak Performance Advertising has seen a 280% increase in conversions within a month while maintaining CPA within the target range. Such success stories are common, with many campaigns using the Max Conversion mode seeing substantial conversion growth alongside stable CPA.

Beyond Max Conversion, MediaGo's SmartBid product also supports the TCPA mode, catering to diverse advertising objectives. Data indicates that campaigns leveraging SmartBid on the MediaGo platform have seen an average ROI increase of 35%.

Advertisers' needs are constantly evolving and becoming more diverse. At MediaGo, we firmly believe in the power of technology to address these needs. We remain committed to exploring and refining our technology based on advertisers' requirements, ensuring that our deep learning technology contributes to the advancement of the advertising industry.

— Peter Jinfeng Pan, Head of MediaGo