Unlocking Growth on Amazon DSP with Human-Centered AI

Navigating the Amazon Ecosystem with AI.

Many of us know Amazon for its ecommerce prowess and popular media platforms like Prime Video and Twitch. But Amazon boasts over 40 different owned properties, and this extensive network creates a rich tapestry of first-party data, offering a significant understanding of consumer behavior. However, truly harnessing this data and understanding how shoppers move across properties—from watching a show on Fire TV to shopping on the Amazon website—can be complex.

Beyond Amazon's owned properties, Amazon DSP also offers access to a massive network of third-party inventory through Amazon Publisher Direct and various ad exchanges. This opens up even more avenues to reach your audience wherever they are online. The challenge, then, becomes identifying the most efficient path to deliver your impressions. This is where Amazon's human-centered AI truly shines.

Introducing Brand+ and Performance+

Amazon's AI models, Brand+ and Performance+, are built on the foundation of Ad Relevance, an AI machine learning tool developed to increase addressability in cookieless environments. These models are not about replacing the human media buyer; instead, they are designed to collaborate, offering transparency and control. You can see exactly what the AI is doing, why it's doing it, and crucially, you can intervene, nudge, and redirect to ensure it aligns with your strategic goals.

While both models utilize the same underlying logic, their primary differentiator lies in the outcomes they are designed to generate:

• Brand+: Focuses on the top of the funnel, prioritizing awareness, reach and frequency, with a secondary goal of future conversions within a 12-month conversion horizon. This model is ideal for building long-term brand salience and memory structures. We've seen Brand+ campaigns perform exceptionally well with video creative and streaming inventory.
• Performance+: Centers on consideration and conversion within a shorter 30-day conversion horizon, aiming for immediate conversions and measurable ROI. This model is highly effective for driving actions like purchases. While it can utilize streaming and online video, we often recommend focusing on online video (OLV) and display inventory for optimal results.

How do these models work?

The intelligence behind Brand+ and Performance+ can be broken down into four key components:

1. Advertiser Event (Human-Defined): The process begins with you! You define the specific conversion events you're tracking. This could be anything from website purchases tracked via the Amazon ad tag, offline conversions integrated through a Conversion API, purchases of specific products on Amazon via ASINs, or even mobile app installs and actions reported by an MMP. This human input is crucial, ensuring the AI aligns with your unique business objectives.
2. Addressability Graph: Based on your defined advertiser events, the AI creates a comprehensive "addressability graph." This map provides a multi-property overview of how your converting customers interact across the vast Amazon ecosystem and beyond. It helps understand their journeys, their paths and their engagement patterns.
3. AI Prediction (Lookalike Audiences and Predictive Scores): From the addressability graph, the AI intelligently creates audiences similar to your existing converters—essentially, a lookalike audience. It then generates predictive scores for individuals within this lookalike audience, identifying those most likely to convert based on the creative being served and the desired outcome within a specific timeframe. This ensures your ads are delivered to the most receptive audience.
4. Campaign Delivery and Continuous Optimization: The AI then dynamically delivers impressions through the most efficient channels, whether it's Amazon Prime Video, Magnite or Amazon Publisher Direct. What's truly powerful is the continuous learning loop. The AI constantly tracks post-impression conversions, updates its model, and refines its targeting as more data becomes available, ensuring your campaigns are always optimizing for better outcomes. This process requires a consistent feed of advertiser events, as static data alone isn't enough.

Easily set up your campaigns.

Setting up campaigns with these AI models is incredibly straightforward, often taking less than two minutes for Performance+ and even quicker for Brand+.

For Brand+, once you've defined your conversion event and set your budget, you're ready to launch.

For Performance+, you have the flexibility to choose from three sub-models:

• Customer Acquisition: Targets net new shoppers who haven't interacted with your brand, excluding those who have converted in the past 90 days.
• Remarketing: Focuses on users who have engaged with your brand or product but haven't converted in the last 90 days.
• Retention: Aims to re-engage customers who have interacted with your brand and have not converted in the last seven days, fostering loyalty.

We often recommend starting with all three sub-models for Performance+ and then reviewing their performance to optimize your strategy.

Empowering marketers beyond automation.

Even with AI doing the heavy lifting, you retain full access to all the familiar optimization tools and reporting capabilities you'd expect from Amazon DSP. This includes granular inventory reports, creative performance analysis and insights into geographic and audience metrics. You can still leverage pacing controls, day-parting, frequency settings and the powerful insights from Amazon Marketing Cloud (AMC).

For first-time users, we always suggest running Business as Usual (BAU) campaigns alongside Brand+ and Performance+. This allows you to truly appreciate the incremental value these AI models deliver. We also recommend a minimum flight of two months for both models to allow the AI sufficient time to learn and optimize.

Demonstrating real-world impact.

We've seen firsthand the significant impact these AI models can have. In a recent link-out campaign (tracking conversions outside Amazon's ecosystem) for an advertiser, our manual interventions—specifically refining sub-models and optimizing frequency caps based on AMC insights—led to an incredible 400% increase in ROAS for the Performance+ campaign. Ultimately, this campaign delivered 1.6 times more ROAS compared to the BAU approach.

Similarly, for a link-in campaign (tracking purchases on the Amazon website), the Performance+ campaign achieved a 50% higher ROAS than BAU. While the CPM for Performance+ was 13% higher, the CPA was 66% lower, underscoring the AI's ability to drive more efficient outcomes by focusing on valuable conversions rather than just cheap impressions. This highlights the ongoing industry conversation about prioritizing effective inventory over merely the cheapest.

A partner for success.

At Monks, we're proud to be an AWS Partner, with a 13-year partnership with Amazon. Our expertise in AI has been recognized with awards like the AI Pioneer by The One Show and AI Agency of the Year from AdWeek. We are also a certified AdTech Reseller and Activation Partner for Amazon DSP.

If you're looking to simplify your media buying, unlock the full potential of Amazon's ecosystem, or just want to explore how these human-centered AI models can transform your advertising strategy, we're here to help. Reach out to us for a tailored session or a casual conversation about achieving your goals with Amazon. We'd love to partner with you on your journey to growth and innovation!