Welcome to the inaugural edition of Cherry Picked 👋

Each month, we’ll be cutting through the noise of Streaming TV and commerce media news to bring you the relevant and timely updates you actually need to know. Expect expert opinions, hot tips, and answers to the burning technical AMC questions we know you must have 😏

This edition, we’re recapping the hottest event to happen in the Amazon Ads world — unBoxed, exploring why frequency metrics could change how you approach your ads strategy, and taking a first dive into the differences between AMC SQL and SQL, and why this impacts how you use AMC to extract insights 👇

The Gigi team headed to Austin for Amazon Ads unBoxed. We hosted customers and partners at our cherry-filled Gigi House, and sat down with some experts from Slalom, Cartograph, and Code3 to get their opinions on data collaboration, STV budget allocation, and incremental audiences on Amazon Publisher Cloud.

But that’s not all! A big part of attending unBoxed is getting a first look at new product launches coming to Amazon Ads, and we have some hot takes on what marketers should be keeping an eye on:

Brands can now onboard their first-party data into Amazon’s Ads Data Manager. A privacy-safe interface that simplifies and streamlines the process of first-party data management across Amazon Ads ad tech and can be reused across Amazon’s ad products. This is designed to help marketers measure conversions, engage relevant audiences, and optimize campaigns while retaining visibility and control of their data.

Our Take: 📈📈📈📈/5

This product is the aftermath of different Amazon teams building various places for brands to upload or collaborate their first-party data within the Amazon ecosystem. Is it shiny and exciting? Not particularly. But it does provide two important actions for a brand. First, being able to integrate their first-party data with Amazon directly via API from their CDP/CRM and secondly, allowing the brand to set permissions and controls in order to understand how their first-party data is being used across agencies and other stakeholders.

There is no AWS Clean Room or AMC integration for the Ads Data Manager yet. If you’re a brand who wants to build audience segments within their CDP or CRM and send them directly to the Amazon DSP, this is something to adopt.

Amazon Ads announced that Prime Video Ads (PVA) will be coming to Brazil, India, Japan, the Netherlands, and New Zealand in 2025. Coming off the success of PVA in 2024, Amazon states that this will help brands further connect with new audiences in new countries, achieving full-funnel business results via Amazon’s differentiated value across premium content, reach, first-party signals, and innovative ad tech.

Our Take: 🌎🌎🌎🌎🌎/5

The launch of ads on Prime Video in 2024 has been a smashing success, and Amazon’s playbook for introducing an ad tier has been the envy of every other streamer. Plans to expand internationally is Amazon doubling down on its success. Whether or not international expansion is successful in each market will be predicated on matching the rising reach of Prime Video in each regional market with demand (i.e. brands and agencies). We’re bullish, but of course we are 😉

A new multi-touch attribution beta is coming to Sponsored Ads and the Amazon DSP. Amazon’s new attribution model will divide credit for purchase conversions across touchpoints in proportion to their value, and their likely contribution to shopping decisions. Historically Amazon Ads has only offered first-touch, last-touch, or linear attribution models which don’t accurately convey the impact of awareness campaigns on downstream outcomes—like sales.

Our Take: 🎯🎯🎯/5

We are true advocates of using different attribution models at Gigi…we have 4 of them just on our home screen. But, we believe there is no one-size fits all attribution model as each brand typically has a different view of how they want to weight different touch points or campaigns. Amazon has “launched” (still testing it this year and releasing it in 2025) a science driven multi-touch attribution model that sounds fancy, but the brand has no control or knowledge of how it was built. Once this is publicly released, we would urge brands to understand what is being run behind the scenes if they are going to rely on this model to make big buying decisions. P.S. Gigi is happy to build a custom attribution model with and for you.

You can now tap into incremental audiences and inventory beyond In-Market and Lifestyle in Amazon Publisher Cloud (APC) with AMC Audiences. This update from Amazon Ads allows marketers to curate programmatic deals in APC and identify higher indexing audiences by overlapping their proprietary signals with Amazon insights and publishers’ first-party data.

Our Take: 👤👤👤/5

How this actually works (based on our understanding): the Amazon supply side deal desk team can manually create custom deals with 3rd party publisher signals. These deals represent audiences at premium CPMs that you can layer on existing campaigns. They appear in the inventory hub within the Amazon DSP, and it’s unclear how AMC is incorporated. Many cite “expanded reach” as the primary benefit of doing this. We’re unsure how, exactly, that’s the case and if the benefits of reach outweigh the added costs. It’ll be interesting to see how this develops.

AMC rule-based and lookalike audiences can now be used in Sponsored Display targeting or in bid boosts for Sponsored Products and Sponsored Brands. Marketers can leverage insights from AMC to increase ad relevance across Sponsored Ads (previously only available to Amazon DSP) and build full-funnel strategies that focus their investment on customers most likely to convert.

Our Take: 🤨/5

Perhaps our spiciest take of them all…AMC Audiences for Sponsored Ads will be the most over hyped product launch of unBoxed 2024. Every agency and tool provider will say its a “game changer” that you need to act upon quickly. But really, this is an elaborate ploy by Amazon to combat flatlining CPCs on Sponsored Ads by selling the same inventory at a higher price. Not only will the price of the ad be higher, but this will come with the added operational cost of buying CPC inventory with AMC audiences. Hold off on adding this as part of your strategy until it’s demonstrably proven this can be a profitable tactic (both in ad and operational cost). We’re open to being proven wrong here though.

There are some new metrics in town that you should be paying attention to…

Streaming TV is an awareness investment and you’re likely looking to prove its value. But it’s important to set timeline expectations for measuring success. Understanding the time to convert for your Streaming TV campaigns can give you a clearer picture of when you should begin looking at results of your campaign. For example, if we’re going off of our benchmark above, to see an accurate picture of performance, you shouldn’t begin measuring the results of your campaign until at least one month post-campaign launch, and up to two months beyond when your campaign ends.

We recently spoke about why you should be rethinking ad frequency. With concepts like the “rule of 7” that suggest it takes 7 to 10 interactions with a brand before a person is ready to make a purchase, frequency is so much more than just a defense mechanism. Its a strategy.

Taking into consideration it takes on average 28 days for a user to convert, finding the optimal number of exposures will help strengthen your brand narrative and guide customers to that eventual purchase.

But how do you measure that ideal exposure?

Unique Average Frequency

This is the average frequency of unique users who have never seen an ad before. It’s important to callout the inclusion of “unique” here as this is different than just saying “average frequency”. By looking at unique frequency, you’re able to understand how many touchpoints a true new-to-brand shopper has versus bucketing them in with users who may have seen an ad previously.

Unique Median Frequency

This will show you the true middle point of frequency that you’re serving to unique users. Unique Median Frequency should be looked at in conjunction with Unique Average Frequency. Why? There can often be frequency outliers that may potentially skew your average frequency. For example if you’re hoping to serve users with frequency of 5, but are seeing your average frequency much higher or lower, there could be outliers that are being served significantly more (or less) ads than desired.

We recommend looking at the Unique Average Frequency when there are few outliers, but Unique Median Frequency to find the central tendency when dealing with skewed number distributions.

Why is frequency important to measure?

Using these metrics you can see how well your campaign is doing based on frequency KPIs along with detail page view rate and purchase rate to understand what strategic levers you need to adjust to reach that ideal frequency (i.e. increase spend, inventory, etc).

p.s. stay tuned because these metrics may be coming to a Gigi dashboard near you very soon 👀

AMC SQL is not like regular SQL, it’s the cool SQL.

But in actuality, AMC SQL can be more difficult to use and harder to pull insights from if not done correctly. Because Amazon Marketing Cloud is a clean room, there are higher privacy thresholds put in place to protect a user’s information. And as a result, the SQL used to query the clean room needs to be adjusted to respect these privacy thresholds.

If you (or your AMC analyst) wanted to query AMC to analyze which users purchased an ASIN from which campaign and for how much, it isn’t as simple as one would imagine.

In a normal non-AMC world, the user would be able to SELECT the columns “user_id”, “campaign”, and “total_product_sales” from the table and the output would be very detailed (see regular SQL columns in the image). You would then see that 2 users converted on the STV - Gigi - High LTV Lookalike campaign for $29.99 and 1 user converted on the Prime Video - Gigi - NTB Shopify Lookalike campaign. Great!

Well, in the AMC world, the user is not able to SELECT “user_id”. Why? The user_id field is a high aggregation threshold column, meaning it would be against privacy policies to “follow” one user in their journey and access details such as the zip code or state they purchased from. You’ll never be able to see user_id’s in the final output.

To adhere to this aggregation policy, we will need to adjust our SQL to count the user_id’s versus extract the actual user_id value. If you know SQL well, because you’re counting one of the values, you’ll also need to aggregate the other two columns you’re extracting, resulting in summing total_product_sales and grouping by campaign.

That’s not all. Once you’ve run your query, there is a rule where 2 or more (anonymous) users have to do the SAME action for you to get detailed results. In our example use case, 2 users did purchase on the STV - Gigi - High LTV Lookalike campaign therefore we get to view that information at an aggregated level (again, no user_id details but we counted the user_id’s thus it would be 2). We can also see the sum of the total_product_sales which is $29.99 + $29.99 = $59.98!

But…for the ONE user who purchased on Prime Video - Gigi - NTB Shopify Lookalike, unfortunately in the AMC SQL world, we cannot extract that detailed information. We will only see that 1 user purchased for $29.99 but no idea which campaign they did that on.

Because of this, AMC SQL needs to be crafted very carefully to extract the information wanted within these guardrails. But that’s what makes Gigi experts, and what makes it fun. Who doesn’t love a good challenge?

Thanks for reading Cherry Picked, and we’ll see you next month!