EXTRA EXTRA 🚨 AWS Cleans Rooms are the latest, greatest data collaboration tool! So great, in fact, that the Gigi team has officially deprecated AMC uploader for all data collaboration 😱

But why is it so great, you ask? In this issue of Cherry Picked, we’re breaking down the top 4 ways AWS Clean Rooms enhance data collaboration, our potentially controversial take on Path to Purchase reports, and a sneak peek at the inner workings of how AWS Clean Rooms actually work.

We’re not telling you to keep your room clean, but we think you should.

Let’s dive in 👇

Data collaboration is foundational to the future of our business. That goes for us here at Gigi and across broader ad tech. When building products to enable data collaboration for our customers, we seek to address three central tenets: 1) maximizing privacy and security, 2) minimizing technical and operational friction, and 3) capturing as many rich signals as possible to enable more effective targeting and measurement.

Our first iteration of the Gigi 1P platform sought to do this for brands by allowing them to authenticate their DTC storefront and collaborate all their transactional data with Amazon ads via the AMC Uploader. This went well. Over the past 6 months, we’ve helped dozens of brands build AMC lookalike audiences using 1P and Amazon signals to deterministically measure outcomes after being exposed to an STV ad across both Amazon and 1P sources. This week, we launched a new set of tools for data collaboration built on AWS Clean Rooms. By doing so, we’re betting our core data collaboration infrastructure on AWS Clean Rooms. Here are the top reasons why we made this choice and why brands should care:

We are confident that building our core data collaboration infrastructure for AWS Clean Rooms will improve our customers' advertising outcomes. Please contact us if you’d like to begin using AWS Clean Rooms to collaborate data with Amazon Ads.

In theory, Path to Purchase reports offer exactly what marketers are looking for. A way to see all of the Amazon ad units users were exposed to, and purchased from taking a specific path. Whenever anyone mentions STV ads and AMC, path to purchase reports are immediately mentioned. This can be via the instructional query library or through most AMC tool providers. Guess what? We don’t like them.

Path to purchase reports are not actionable, and they don’t provide an accurate picture of total conversions due to AMC’s aggregation thresholds.

Firstly, when you get a Path to Purchase report you’re inundated with thousands upon thousands of rows of different paths users take to purchase. Brands that run both Sponsored and Amazon DSP ads have so many placements throughout a buyer’s journey, that users could see up to 200 different ads before buying. So you can see that a user that viewed an STV ad, 2 Amazon DSP display ads, 1 Sponsored Brand ad, and 2 Sponsored Product ads has the highest likelihood to purchase. What are you supposed to do with that insight? Spend more on Amazon ads? You can’t control which ads get shown to users when, and how to change a user’s paths in the midst of their buying journey. Path to Purchase reports are essentially a proverbial pat on the back for participation: congrats on spending money on Amazon ads that lead to purchases.

We’ve also observed data discrepancies in Path to Purchase reports. Over the past several years, we’ve seen inconsistencies between summed conversions not adding up to the number of conversions seen in the Amazon Ads campaign manager. We think this is because the Path to Purchase report must have 2 or more users take the exact same path to show up in the report. To which we have to ask, with that many ad placements and possible purchase paths, what is the likelihood that 2 or more users are seeing the exact same ads along the same path within the same time period? Unless the path is simple, the odds are not in our favour. Hence, the discrepancies.

At Gigi, we intentionally omitted Path to Purchase reports in our measurement suite. However, we do believe it is important for our customers to understand how their STV campaign, ADSP Display campaigns, and Sponsored Ads campaign work in harmony amongst each other to drive purchasers. So, how have we addressed this?

We’ve created a Paths tab dedicated to surfacing visualizations that actually help action a path to purchase. For example, we’ve created a Venn diagram that breaks down how many users saw different types of ads (Sponsored, STV, and/or other Amazon DSP) separately or together and if they purchased or not. This data is the same type of data received in the standard Path to Purchase reports, but queried at a higher level. This enables us to show all of the conversions in a digestible way that can be used to make quick actions like extending budget if you see users that view STV, Sponsored and other Amazon DSP ads have a 3x higher likely purchase rate versus only seeing Sponsored and other Amazon DSP ads.

While the detailed Path to Purchase reports can be filled with insights data, its ultimately overwhelming for brands to action and is too granular for AMC to extract insights given aggregation thresholds. Instead, Path to Purchase insights should give you easily actionable insights. You don’t need to know that a single user saw 3 STV ads, 4 Amazon DSP, and 2 Sponsored ads before purchasing. But you should know if multiple users seeing those three ad types incrementally drive higher purchase rates so you can adjust your strategy accordingly.

To collaborate your first-party data in as secure a way as possible, limiting the movement of your data is crucial. While it can be a powerful tool for measuring deterministic outcomes across first-party channels and Amazon ads, data collaboration requires rigorous data-handling practices to protect sensitive information. This is ultimately why we chose to use AMC on AWS Clean Rooms as our preferred means of collaboration over AMC Uploader.

AMC Uploader streamlines the data preparation and upload process into AMC, automating tasks like normalization and hashing to meet AMC's data requirements. However, AMC on AWS Clean Rooms brings data security and privacy to higher level as it acts as a bridge between data sources, providing a collaborative environment where brands can securely combine their first-party data with Amazon Ads' signals within their AWS account, enabling deeper analysis and audience insights without moving data outside their AWS environment.

Let’s dive into how AMC on AWS Clean Rooms enables a more secure means of collaboration:

Data hashing or encryption is critical to ensuring Personally Identifiable Information (PII) is anonymized. With AMC Uploader, hashing occurs before uploading data into AMC and it is the brand’s responsibility to ensure data is properly prepared and secure before being transferred.

With AMC on AWS Clean Rooms, hashing occurs automatically within AWS Clean Room as part of the collaboration process, ensuring that no PII data leaves its origin environment. Additionally, there are standardized hashing processes in place to ensure data can be matched across anonymized records without revealing the PII.

Brands using AMC on AWS Clean Rooms for data collaboration not only have to adhere to AMC’s aggregation thresholds, where 2 or more (anonymous) users have to do the SAME action, but can also layer on their own data usage and aggregation rules as to what can be queried.

For example, a brand could say that you can only see aggregated results where 5 or more (anonymous) users have done the SAME action. This enables brands who may have more stringent compliance requirements to determine how their first-party data is collaborated with.

During the AMC on AWS Clean Rooms setup process, we build the Gigi data model within the AWS S3 bucket using Apache Iceberg table format. This allows us to structure data across multiple sources in a standardized manner and ensure insights are extracted without discrepancies or inconsistencies.

Additionally, this allows brands to collaborate their first-party data with other cloud environments or multi-party sources that also use AWS and Apache Iceberg table formatting. And provides cross-channel insights across a vast set of data sources without any raw data exchange needed.

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