Ava Hinds: 0:00

Today we're getting radically candid with Aqua Bailey, our Operational Support Lead here at Cognition. In this episode, we take a deep dive into the Cognition Platform's Data Studio. Aqua shares insights into her work on the development support side, explaining how she ensures your campaign data is thoroughly managed behind the scenes. If you've ever wondered what a data clean room actually is or how we process large data sets, this is an episode you're not going to want to miss. Enjoy. Hello everyone. Today I'm here with Aqua. Aqua, do you want to tell everyone what you do at Cognition and your title?

Aqua Bailey: 0:38

Hi, hi. I'm Aqua. My title is Operation Support Lead. I do all of the troubleshooting or any type of inventory issues, data reporting issues, just anything related to troubleshooting, anything that needs to be sent to our development team for bigger bugs like anything that I cannot fix on the front end and they have to do on the back end. So I'm a liaison between ops and dev and also ops in the product team for any type of feature enhancements or new features that ops operations or account managers would like for their clients to make anything in our platform easier.

Ava Hinds: 1:18

So my first question is can you explain how Data Studio integrates with other platforms to provide insights for clients?

Aqua Bailey: 1:25

Yes. So our Data Studio actually can integrate into multiple CDPs, which is customer demand platforms. We have successfully linked our reporting to multiple outside sources, just a bunch of our clients. We also can integrate it with Google Analytics API as well. So that's one of the selling points that we have was being able to tell our customers, like, hey, you guys don't have to use our platform. You can use your own, and we can integrate our reporting into you guys' dashboards.

Ava Hinds: 2:00

And then what types of data are most impactful for clients?

Aqua Bailey: 2:04

So the data that is most impactful for clients are, of course, impressions and conversions, but not only impressions and conversions, but also being able to see what customers or what conversions from their ads actually went in to buy a car. And that's where our first-party data comes in. So if a dealership was to send us their sales data, we can actually aggregate it into our platform with AMC, and we can give them matchbacks to see exactly what customers from that dealership saw their ad and bought a car.

Ava Hinds: 2:39

And then how does Data Studio help in identifying KPIs or key performance indicators for campaigns?

Aqua Bailey: 2:46

Oh, so we actually get our KPIs from the DSV. We integrate it into our platform. So we look at anything that's clicks, impressions, conversions off Amazon conversions, on Amazon conversions. We also look at our what we call Headless Analytics API. We also call it HAD. That actually will give us the entire story of what our customer is clicking on into an ad. By that, we can actually use that those KPIs to then do retargeting campaigns to make sure, like, hey, yeah, I saw this ad, you haven't bought a car yet, but we can save you another ad that can push you to buy that new car.

Ava Hinds: 3:26

And speaking of key performance indicators and everything that clients are asking for, can you share an example of when Data Studio uncovered a critical insight that led to a campaign optimization? So kind of like a problem you guys identified.

Aqua Bailey: 3:42

Yes. So with our conversions, we have actually been able to de-duplicate a lot of convergence because we have clients who run multiple campaigns. But you don't want to see those same numbers each time you're looking at your metrics. So what we have done and identified was like, hey, we're seeing different numbers in the DSP versus in AMC. And we're like, hmm, I wonder why. So then they led us to investigate why our numbers in AMC were lower than once in the DSP, which is because AMC takes out those duplicated numbers, which actually helps a lot because you want to know your true conversion number, not just the customers who have seen your ads multiple times. Or let's say you have a display ad and a video ad. You want to know which customer saw the display ad. Okay. But that customer could have also seen the video ad. You don't want to count them as a conversion for both of those ads for just one dealer. You want to target them that one time. So that way they're not getting tired of seeing your ads and like, oh, let me skip this. Let me just bypass it. You don't want to do that. You want them to actually look at it and actually interact with it. So we get them de-duplicated numbers. And then what is a deduplicated number again? So a deduplicated number is a conversion number that is from the DSP, but because our clients run multiple campaigns, that one customer could have saw each advertisement or creative from each campaign. So we're going to deduplicate that. They also could have saw it more than once. So we know we like to have our ads out there multiple times on like Hulu or when you're web browsing. You only want to count that customer one time because that will give you a true metric of how many conversions you have of who clicked that ad. Instead of me being able to count Ava, click this ad twice, and she saw the video four times, and then she clicked the video twice. We want to count you just one time so we get the true number. Because duplicated numbers, they just show you, like, oh, it's working. You know, you have impressions, you have conversions. But of those conversions, how many of them are the exact same person? You don't want to sit here and have the same person clicking on it and counting them every single time because then they're just going to make your numbers look good. However, you are going to get a true ROI of a return from having a match back because she saw or Ava saw the campaign multiple times and clicked on it.

Ava Hinds: 6:16

And is that one of the ways you ensure data accuracy and relevance, or what are some other ways that you do that?

Aqua Bailey: 6:23

Yes, that is one of our major ways to ensure accuracy. Another is, of course, our first party data, which is our clients giving us their sales data, and then we upload it to AMC, which is Amazon Marketing Cloud. They look at the data, run it against the customer information that is collected from who saw a campaign, as long as their Google is like signed in, not private settings or anything like that, then we can aggregate those numbers. That also gives us a true read of how many people truly went and bought a car and saw our ad. So we can aggregate that and then count it towards that. So I feel like that's definitely one of our biggest things to be able to accuracy when it comes to conversion.

Ava Hinds: 7:07

And then along with that, once the data is processed within our data studio, how do you push those segments into DSP audiences? Kind of like what you're talking about with first party.

Aqua Bailey: 7:16

So we can actually build an audience straight off of any type of sales data, or we can even build it off of conversions. So what we do is we will take the convergence that we have in our platform and we'll send it to AMC. As long as we have a big enough audience, like if we have over 205,000 impressions, we can take those impressions or convergence and we can make them into an audience so that way we can target those.

Ava Hinds: 7:47

And then I have some dev talk questions, just so we can get a basic understanding of some things in the development world. But another big question that a lot of people are curious about is I didn't kind of talk a little about Amazon Marketing Cloud, but what is a DCR or a data clean room?

Aqua Bailey: 8:05

So a data clean room is basically where we store all of our data. So you would think of it as like a filing cabinet. And inside that filing cabinet, we have different sections and different sectors that will hold certain information that we would need to keep, um, whether this data or client information, we want to make sure that our client information definitely stays private for PII reasons. But it's literally just a like a huge filing cabinet where we can store all of our data, all of our information. And we're able to go and grab that at any time if we're needed. We do have what we call a backfill, but Amazon only allows us to backfill data for up to 95 days, which I believe is amazing because other GSPs and platforms don't give you that same room build-to-backfill data from your clean room. It holds everything, like your inventory that we use, it holds your impressions data, your sales data, and it holds it all in the data table. So those are the things that you would think of in a data clean room that would be literally like a huge private cabin. Then why would you say they're really important to have? Well, they're really important to have because you don't want to lose data. The moment that you lose data, you can lose a client, you know, because what if they're like, oh, you know, we want to know all of our data for the past year so we can compare it and be able to keep our clients on, keep them happy so that way we can show them, like, hey, this is actually working. Month over month, you're getting more impressions, you're getting more convergence, you're getting more off Amazon events. You want to be able to keep all of that data to be able to track and prove what we do here at Cognition is actually working for that client.

Ava Hinds: 9:48

It's really cool. And I know a lot about data cleaners about privacy and keeping all of that with for clients as well. But going along with more data and everything, how do you leverage first party data to maintain attribution and targeting accuracy? I know you talked about the duplicates, but are there any other ways?

Aqua Bailey: 10:06

Yes. So actually, when it comes to our attribution and reported sales data, so there's a little bit of a difference. So with first party data, we actually get that data strictly from the um dealership. Now for reported sales data, the attribution, we actually pull that from Pulk. And so with that, our customers don't have to give us their sales data if they don't want to. And most of the time it's for PII reasons, which is completely understandable because you do want to keep all of your customers' information private. So with that, they don't have to give us sales data. We can just go and pull it for them and tell them, like, hey, this is how many cars they sold. This is how many we can aggregate back to cognition because these people saw your ad. That's one of the biggest things that we use for Polk. And a lot of our clients actually use it, love it because they don't have to give us data and worry about, oh, what if there's a data leakage or you know, exposing their clients' data? So that is one of our biggest other points as well, because they don't have to do anything. All they have to do is just push it up.

Ava Hinds: 11:12

That's really awesome as well. And I know you talked a little bit about Polk. Can you go more further into like what Polk is and just explain?

Aqua Bailey: 11:19

Yes. So Polk data is reported sales data from different dealerships. And what they do is they send over their sales data, and whatever customers or sales they get goes to then the DMV. And so we didn't get those numbers from the DMV of those registered buyers import into our platform, aggregate it, and then that way they know, oh, so this dealership sold this many cars, and of those people who bought a car, they saw our ad. So with that, it does take a little bit of time for pulp to settle in because again, the data is coming straight from the DMV. And we all know, and we all hate to go to the DMV, so they are a little slow. If you're like looking for the month of let's say December, you would wait until like February to let those numbers settle in so that way the DMV can actually get all of those customer registrations put into their system and then report back, and then we'll pull those numbers.

Ava Hinds: 12:22

That is really cool. Well, those are all my questions, but thank you so much, Aqua. Do you have anything else you want to share or any last remarks?

Aqua Bailey: 12:33

Um no, I had a good time with you, Abe. Let me know if you have any other questions.

Ava Hinds: 12:39

Bye, guys. See you on the next one.