Six questions with Neal Richter of Amazon Ads on Amazon DSP and its upgraded capabilities

6 June 2023 | By George Slefo, Senior Editor

Two people chatting

Neal Richter, director of advertising science and engineering at Amazon Ads, has been busy spearheading a large collection of Amazon DSP upgrades that improve areas such as campaign performance, cost efficiency and advertising addressability.

The new capabilities include improved proprietary algorithms that are powered by Amazon SageMaker, which continuously test and optimize ad campaigns for performance and help advertisers reach previously unaddressable audiences. “As the industry moves away from third-party cookies, these new enhancements make Amazon DSP more capable of advertising anywhere on the internet,” says Richter. “They’re also helping drive performance in areas such as bid pricing, pacing, addressability and performance for our customers.”

Richter explains how the new upgrades are helping advertisers meet their campaign objectives by bolstering areas such as performance and cost-efficiency. Amazon DSP customers, for instance, saw a 34.1% increase in return on ad spend (ROAS) and a 12.6% upswing in click-through rates (CTR). Additionally, advertisers experienced a 24.7% decrease in cost per click.

Notably, Richter says the latest enhancements work “behind the scenes”, helping advertisers better manage their budget, reach larger audiences and improve campaign performance – without additional expenses or investments.

We caught up with Richter, an ad tech veteran and the current chairperson of IAB Tech Lab, to ask how these new capabilities work, what made him join Amazon Ads, and why advertisers should leverage Amazon DSP beyond Amazon’s storefront, with solutions like Streaming TV.

People having coffee

Given your extensive experience in the ad tech industry, what inspired you to join Amazon Ads two years ago?

Amazon Ads is unique in our industry given its deep understanding of our customers. And as a data scientist who interprets large sets of information through computational methods, you want to go to the companies that not only have the most compelling insights, but are also in a position to turn that information into a business.

How does that tie into the new enhancements?

We started by asking what we want Amazon DSP to become. And our solution was to develop a more comprehensive DSP that leverages our valuable audience signals and measurement capabilities for our customers’ ad campaigns on any inventory on the internet, so beyond just

Additionally, we have also increased the quantity of information analyzed by our machine learning models, including a broader range of audience signals and outcomes, to continuously enhance campaign performance.

And as advertisers continue to seek ways to improve campaign performance, our latest enhancements have improved delivery accuracy, pacing and ad performance prediction. These upgrades have resulted in a 12.6% increase in CTR for our customers, providing them with greater confidence that their campaign goals will be met, no matter the audience or supply environment.

Can you provide more details about how the new Amazon DSP upgrades work “behind the scenes”, as mentioned in your blog post?

When we say “behind the scenes”, we mean Amazon DSP customers will not need to make changes to benefit. We’re essentially taking the information our customers have already given us, and we’re using advanced technology and a wider range of signals to help them reach their performance goals. The best is that they need to do little to nothing, and don’t even have to pay extra. In other words, if you’re running a campaign through Amazon DSP, you should see improvements in areas like performance and addressability.

Can you explain how the upgrades are driving performance for brands that aren’t commonly associated with advertising on

Our new algorithm upgrades have improved Amazon DSP capabilities to buy anywhere on the internet, where we have direct access to tens of thousands of sites and apps, as well as third-party exchanges using Amazon Audiences, measurement and signal-based marketing to make it a full-featured, enterprise DSP.

This means a hotel or car rental brand, for example, can reach customers who recently purchased luggage more efficiently through both traditional and emerging channels.

We’ve also enhanced tools our customers are already familiar with to better connect with hard-to-reach audiences. Our key differentiator, however, is that we’re combining those products with all the unique capabilities that Amazon Ads can offer. Customers who used contextual and modelled ways to reach audiences saw an increase in incremental addressability by 20% to 30% on previously unaddressable inventory such as Safari, Firefox and iOS.

How are these enhancements helping advertisers overcome some of the challenges seen in TV buying?

Streaming TV has especially benefited from our latest upgrades. The enhancements we’ve made have bolstered advertisers’ ability to reach audiences in a streaming TV environment, which includes Fire TV and all its streaming apps, as well as third-party streaming inventory.

All of this in turn is helping our customers solve for some of the largest challenges in TV buying, such as addressability, purchasing video inventory more efficiently, while also driving better campaign performance.

Can you explain how Amazon DSP has improved its algorithm and optimization capabilities?

We recently released a new generation of our core bidder, a multiyear build that includes upgraded core prediction systems that are powered by deep learning. This includes improvements to pacing and budgeting algorithms, as well as bid price calculation and optimization toward customer objectives. Over the past year, Amazon DSP has also refined the core algorithms to optimize for areas such as CPA, ROAS, reach and frequency.

1 Source: Amazon internal data, US, 2022; 140K campaigns across verticals