
Episode Description
The bird puns from Google are getting to be a bit flighty.
The now-retired Federated Learning of Cohorts left advertisers with concerns about how audience targeting will be done in a world without third-party cookies. We asked for some help to understand how advertisers can keep getting the most out of their ad spend while respecting audience privacy.
This panel covers:
- What exactly is #FLoC
- How does FLoC replace cookies
- What will advertisers lose with the demise of third-party cookies
- #Remarketing with Google’s TURTLEDOVE
Episode Takeaways
What exactly is FLoC:
- FLoC stands for Federated Learning of Cohorts.
- Designed to replace third-party cookies by grouping users into cohorts based on similar browsing behaviors while maintaining individual privacy.
- Operates by aggregating data locally on the user’s device, ensuring browsing history is not shared externally.
How does FLoC replace cookies:
- FLoC enables advertisers to target groups of users (cohorts) rather than individual tracking.
- Cohorts are determined by algorithms in the user’s browser, creating a privacy-first approach to targeting.
- Advertisers receive a cohort ID but no specific user data, maintaining privacy while still allowing for effective targeting.
What will advertisers lose with the demise of third-party cookies:
- The ability to track individual user behaviors across different sites will significantly diminish.
- Advertisers will need to rely more on aggregated data and less on personalized targeting strategies.
- Loss of precise, personal data could impact the effectiveness of retargeting campaigns and personalized advertising.
Remarketing with Google’s TURTLEDOVE:
- TURTLEDOVE is designed to enable remarketing in a privacy-preserving manner.
- It proposes a method where ad auction occurs in the browser, allowing advertisers to serve ads based on the user’s previous site interactions without revealing their identity.
- This method ensures user privacy is maintained by keeping the data localized and not shared across networks.
Episode Transcript
Frederick Vallaeys: Hello and welcome to another episode of PPC Town Hall. My name is Fred Vallaeys. I’m the host of this session. I’m one of the co founders of Optmyzr as well. So today is going to be a fun one because we’re going to be talking about a topic that quite honestly, nobody knows very much about, but it’s hugely critical to the future of PPC.
As you’ve all seen, new browsers have come along. They’re starting to block third party cookies. There’s privacy laws coming left and right. The world is just fundamentally changing. And some of the things that we as PPC practitioners have been able to use for the past 20 or so years because privacy regulations were pretty lax and because third party cookies were available to us, Well, those things are soon going to go away.
So that changes the game for all of us. And again, we don’t have all the answers, but we do know a lot of the things that are changing some of the things Google is thinking about. And that’s what the session is about today. So we’re going to be talking about the end of third party cookies, the introduction of FLoC, the introduction of turtle dove.
If you don’t know what those terms are, don’t worry. We’re going to explain it. We’re going to keep it non technical and we’re going to try to make it actionable. So we’re going to tell you what you can do. to start preparing for these things. So welcome to another episode of PPC time.
And a big welcome to our guests today, Navah Hopkins and Amy. So good to, to see both of you. Thanks for coming back on the show.
Navah Hopkins: Thank you for having us. Yeah. Thank you for having us.
Frederick Vallaeys: Yeah, never. Let’s start with you. So since the last time we had you on the show, you’ve transitioned roles working for a new company.
So tell us a little bit about what you’re up to these days.
Navah Hopkins: Sure. So just, you know, is actually just, you know, got me specifically because of what we’re about to talk about. So they are a CRO suite that helps folks earn that first party data create amazing converting experiences enhance the path to conversion and My role there is to actually be a liaison between the paid media community on their tool.
They have a phenomenal suite of zero on page optimizations. But more importantly, they’re able to help that cross platform conversation. So as we think about attribution, as we think about generating those first party data pieces, I am Helping them help all of us be, be, get ahead of FLoC and ahead of that potential retargeting death.
All right. Sounds
Frederick Vallaeys: like the right time to make that sort of a move. Exciting company. So thanks for joining us. Amy Bishop stuff is new in your life as well, right? Still cultivate more cultivative marketing. But you’ve also taken on a new role, I think at SCJ. So tell us a little bit about what’s going on with you.
Amy Bishop: Yeah, that’s right. So I’m still heading up Cultivative, which is a paid search consultancy. But recently, I’ve also started working with SEJ to help with their paid media news coverage, which has been really exciting, really fun. Just kind of helps keep me on top of everything, as opposed to reading what other people have shared.
Now I’m kind of responsible for coming up with the articles and writing those myself. So it’s a great way to Make sure that I’m always staying on top of any new information that’s coming out.
Frederick Vallaeys: Nice. So you’re the new Susan Winograd or the new Ginny Marvin, I guess. Yeah,
Amy Bishop: those are really big shoes to fill.
So I don’t know if I could call myself the new one of either of those ladies, but I definitely aspire one day to maybe fill those shoes.
Frederick Vallaeys: Exactly. Starting to fill them and then onto greater and better things. Well great to have both of you on the show. Before we get into the topic of FLoC, I did want to make an announcement.
So Optmyzr is going to be hosting its first ever user conference. It’s going to be called Unlevel and we have some great speakers for it. So we’re very fortunate. We have John Lee from Microsoft. We have Nava, who’s on the show today. Many other great speakers. This is going to be a free event. It’s going to be happening on May 18th.
I may not work for everybody’s time zones, but because it’s on demand, we do want people to register and come check it out because we’re going to have some great educational content. And that’s really one of the things we do at PPC town hall and that we’re also very big on. That Optmyzr is PPC is honestly an exciting space because it changes so often, but it also means we have to do a lot of work ourselves to keep on the cutting edge.
And so we just wanted to make that easier, especially in the pandemic days when it’s very difficult to go to in person conferences, even to this day. So so we’re going to be putting on a virtual event. If you’re an existing user of Optmyzr, you’re going to get a benefit of access to a special track, which is our roadmap track, where you’re going to be able to work directly with our product managers, give feedback to them and hear what they’re working on.
If you’re not existing, an existing Optmyzr user join us as well. Again, it’s free. You’re going to hear some of the big topics in the industry and also get to explore a little bit about what Optmyzr is about. So we’ll have the landing page for that. You can start signing up and that will be happening in a few weeks.
So we’re super excited about that. All right. But yeah, and the other thing lots of people viewing live today. So that’s fantastic. So tell us where you’re calling in from, say hello, using the comment section, that is also how you can ask us questions and start contributing to the conversation here.
And steer us in the direction that you want, right? Tell us what you want to know more about or tell us if we’re losing you, and then you got some specific questions. We’ll show that on screen with your picture. So let’s make it interactive and engaging. But with that, I figured we start maybe Amy, could you talk us through a little bit about like, Why are we having a session on FLOC?
What’s changing? What is FLOC all about and why is it important?
Amy Bishop: Yeah, absolutely. So we’ve known for a while that third party cookies were going away. What we maybe didn’t know was exactly how those would be replaced. So earlier this year, Google announced they won’t build alternative identifiers to try to track individuals around the web.
And instead they introduced FLOC. FLOC stands for federated learning of cohorts. FLoC is designed to aggregate data about groups of people in order to continue to support advertisers through cohorts. So cohorts would be a variation of kind of what we would typically think of as audiences. And so they would create these cohorts while still protecting the privacy of the people in these cohorts while allowing monetization use cases such as Interest based advertising, ads measurement, reporting, and other scenarios like that, that we’ve come to rely on third party cookies for.
Frederick Vallaeys: Yeah, and so we’re going to spend a few minutes here setting a bit more context, but so, Nava, like, explain to us first party and third party cookie and, like, why, why do people hate third party cookies so much?
Navah Hopkins: Sure. So when we think about first party versus third party, first party is data that you own.
You have created an experience that the user is willing to give you that information. And also it is the user’s completed actions on your site. So the ability to have just analytics of did did a shopping cart belong to a user? And just being able to pick out that information. Did traffic happen on your site?
You will still be able to see. And if you were able to earn that email, you were able to create that, that bond, you will still be able to target through things like customer match and customer audiences on those third party sites and through the ad networks. Third party data is when a user does something, you used to be able to, and you still to a certain extent are able to as, as it’s being depreciated NAG.
The user as they would go to these other sites to remind them to go get your product to remind them to complete the purchase to remind them to go do X, Y, Z thing. So third party data was basically any data that you did not specifically own. So think about buying email lists. Think about publishers that you just You don’t own the channel, so you can’t control what’s what’s there.
So when we think about first party data versus third party data, a good way to think about it is first party data is that trusted relationship that person has opted in to talk to you. Third party data is kind of the Ticket walker trying to, or scalper trying to get you to buy the ticket.
Frederick Vallaeys: Right.
Cause the ticket scalper knows you just showed up to the arena. Exactly. Interested in taking it. And so the example in, in PPC would be. You know, I’ve been on Edmunds. com, I’ve been on Cars. com, so I’m clearly shopping for a new car. And then out of left field, some like Mercedes dealership comes in and says, Hey Fred, here’s an ad for you, do you want to buy this car?
And it’s like, whoa, like why does Mercedes know that I’ve been shopping for a car? Because I haven’t been to that website.
Navah Hopkins: Well, one, one thing, and I, I, I hope we’ll, we’ll all dive into this, and this might be kind of jumping ahead of time. The desired agenda, but there is this idea with with cohorts being very similar to In market audiences, custom intent audiences.
So the ability to, to still say, Hey, Fred, you seem like you’re interested in cars, but it won’t be, Hey, Fred, it’ll be Fred amongst a thousand, two thousand, 5, 000, 10, 000 other people that show similar browsing behavior. So you still will be able to have that contextual targeting. It’s just not going to be, you looked at that 2015, you seemed like you were really interested in financing options.
Here’s. A very specific offer to you based off of the exact things that you did.
Frederick Vallaeys: Right. And Amy, go ahead.
Amy Bishop: Oh, I was just going to say, I think maybe it might help to talk just about a little bit about how exactly Fluck replaces cookies. And maybe I should have covered this in kind of the initial piece that I gave, but the big difference is that cohorts are built within the browser using on device processing to keep each person’s History private on the browser.
So chrome browsers specifically. We’re specifically talking about chrome when we’re talking about FLoC. They will use algorithms to create a very large number of cohorts based upon people that kind of share the same interest. But each of that. Each person’s cohort information in their individual browsing history is kept private.
So only their browser itself within that specific device will be able to look at their history and then assign them to a cohort. So initially, your browser. Sorry. Essentially, your browser determines Which cohort corresponds to your recent web browsing history, grouping you with thousands of other people so that then when you’re visiting a site, they have the ability that site has the ability to ask your browser which cohort you belong to, and then your browser will give them an identification number of the cohort and then.
It’s essentially giving them only that little piece of information as opposed to third party cookies, which would have allowed companies to kind of follow you individually across different sites around the web. So FLoC essentially works on your device without your browsing history being shared. So to give a more specific example, like, for instance, when you’re browsing chrome.
It will tell that site that you’re part of this specific cohort number. And then it’s up to that website to actually know what that cohort number means. So for instance, if that cohort number means that I’m interested into your to your point, like Mercedes, then it’s going to give like an ID, like 1, and it’s going to tell the website.
This person is part of cohort one, two, three, four, five, six. And then that website has to take that information and determine what ad to show based upon that, but they’re not going to receive any more context about me or my browsing habits or what else that I’ve looked at. So it’s a lot more private than what third party cookies used to be and what used to be shared as part of that.
Frederick Vallaeys: So that raises. The big question. Okay. So I’m an advertiser. I go into Google. I’m like, okay, I want in market car buyers for a specific brand. That’s my audience. Go and show my ads. Easy. And now you’re kind of saying that as an advertiser, I’m going to have to be like, I want to target one, two, three. A w nine to six.
Amy Bishop: Yeah, and so I think from an advertiser standpoint especially since we’ll be For many of us, that are working through the google ads platform. It’s going to probably be made simple for us I would say Where some of the pain points are going to lie are going to be on some of the other ad serving platforms that are going to have to Figure out how to decode this and make something of it so I think from an advertiser standpoint, I actually believe that the impact is Less for the advertisers as it is for some of the other folks that are involved in accustomed to using third party cookies.
Frederick Vallaeys: And does that perhaps start touching a little bit about Bernanke? And I don’t know if we want to talk about that here for a second, right? But so Bernanke is this thing that came out in the past week or so through a Texas lawsuit against Google, where it sounds like Google is giving itself an unfair advantage in the auction for Display Network. And so you’re kind of saying the same thing might happen here in the roundabout way. So if you’re on Google, because it’s Google having built these cohorts, they can tell you maybe what’s in these cohorts, whereas if you’re off Google, it’s, you don’t have quite the same information. And
Amy Bishop: it’s hard to say, you know, I don’t want to make accusations because there’s just too little information for us to really know, but just to kind of, I do think there’s potentially some things that are.
Coincidentally close here. So just for like a little bit of background Texas is leading a multi state lawsuit against Google and the original complaint that was filed in December claims that Google uses its monopoly power to control pricing and engage market collusion to rig auctions. That same filing mentions a project which was codenamed Jet Ed Blue, and that was an agreement between Google and Facebook where Essentially, Facebook agreed not to compete too heavily with Google, but in exchange for preferential treatment on Google’s advertising network, and I believe there was also a minimum spend that Facebook also agreed to hit in order to make it worth Google as well.
But anyway, as part of as part of this ongoing lawsuit, Google filed documents presumably accidentally that were unredacted and they detailed project for Nancy, which is. What we’re talking about here. And project for Nancy was said to be an internal Google program, which allegedly made use of data from publishers as well as competitor bidding models to try to determine their bids for specific placements. And initially it really made it sound like it was kind of for Google’s own promotion of Google, but I really pretty sure that it’s more of Google’s promotion of its ad network. In kind of comparison to other either other networks or other, sorry, like in comparison to other exchanges, essentially, they were using this data to make sure that clients on their exchanges.
We’re getting these placements at the best prices compared to what other what other competitors.
Frederick Vallaeys: Interesting. I mean, I think it’s again to the point that things are becoming a bit more complicated and obviously like online advertising is all about big data and how you use that data. That’s what gives you your advantage.
And that data has been relatively easy and free flowing. But now some of the data is going away. Some of the data that we still will have is becoming come maybe a little bit harder to use. So I think that’s what makes all of this so interesting. We actually have a good question that I want to dive into here a bit.
But Vince is asking a specific question about how these cohorts work, right? So say In my example, and Vince, thank you very much for assuming I’m 35. I’m actually a bit older than that. But yes, I’m , and I do wear a red hat. So no, actually I don’t wear a red hat. It’s not, so just understood the red hats might mean something completely different here.
So, but yes, say you have these attributes of a user. How do you figure that they’re in a certain cohort, not a different cohort? And this is actually through a mechanism Google calls SimHash. What’s nice about FLoC and TurtleDove is that all of these are open source systems. So you can actually look at the code, you can see the implementations and there’s a lot of technical detail about it.
But what Google is fundamentally doing is it’s saying, well, here’s the behaviors that this person has taken within their browser, and that’s fed into a SimHash algorithm and SimHash is basically saying, okay, let’s take these behaviors and, and, and put that into a unique key, but it’s not a unique key for every single behavior.
It’s called simhash for similar hashing, right? So two users who do kind of the same thing, but not exactly the same thing, would hash to the same key and the same cohort. And so that’s where you’re going to start seeing you know, perhaps somebody wearing a red hat and being 35 years old and visiting Mercedes might also fall in the same cohort as somebody visiting to be on W site.
By the way, the demographic aspect and weigh in Nav and Amy, but I don’t think the fact that you’re 35, that’s not part of SimHash, right? I mean, what we’re talking about here is really what websites have you visited and what has your behavior been?
Navah Hopkins: Yeah. Yeah.
Frederick Vallaeys: So, I mean, and that’s where it comes down to. So with SimHash, I don’t have to visit exactly admins. com and cars. com to fall into the same simhash as somebody who’s visited similar sites about cars. And that’s kind of the unknown. And I can’t remember exactly how many cohorts would be in this first iteration, but it is actually limited.
And so in simhash, you can set the binary length. And so that controls how many, cohorts you can have. And I thought that the number in the initial phase was going to be relatively small. So that’s kind of bad from a targeting perspective, right? Because a lot of people doing very different things are going to come in into the same cohorts.
So you lose a lot of precision. But the trade off and the benefit is that you get a lot more privacy because it’s unlikely that somebody would fall into a cohort by themselves. The other technical detail here, and again, please weigh in on this if you have any thoughts on it, but there’s fundamentally a two level system.
So the first one is the assignment into a FLoC cohort. The second one is for that cohort to be able to be used, it has to pass a certain number of users. So even in the situation where there’s only five people going into a cohort, Google has put in place a mechanism that says, well, Even though you asked to show ads to this cohort, we’re disallowing it because we haven’t reached the minimum threshold for how many people are in it.
So it’s, it’s too likely you could actually figure out who these five people might be.
Navah Hopkins: One thing and we, we kind of touched on this on the Twitter thread leading up to this town hall, is that Ad platform performance is limited. If you try to target too small a group, there is just a threshold where there just isn’t enough data, there isn’t a search volume, there just isn’t gonna be enough.
So when we think about cohorts number one it’s important to remember that the cohort is still gonna compete in an auction with other audience types. So the, the basic premise of of do is that it’s, it’s going to check to see, what is the better match? What is the better bid? What is the what ad should win between one of these cohort audiences and just an in market audience and what I like to call native audiences?
The other thing to think about is just we’re already used to from a search standpoint, keywords with low search volume. There are just inherently going to be certain ideas, certain groups of folks that if we try to target them specifically, there just isn’t enough for either the ad platform to justify the, the, the auction, because it’s just, That there isn’t enough so now it’s not going to serve the other piece is When you think about the persona behind the cohort I actually don’t mind the loss of precision because if I get too precise I can pigeonhole myself into individual sales And individual sales while useful Don’t have scale.
So as we’re thinking about targeting and how we position our creative, how we position our campaigns this might actually be a very good kind of kick in the pants to think about how what kind of audiences are we targeting? What kind of personas are we building? And how can our creative better serve a persona as opposed to an individual person and nagging them into giving them giving us our money?
Or their money.
Frederick Vallaeys: Yeah, that makes a lot of sense. And that sounds like a, a good strategy.
Amy Bishop: Yeah, and I would add to that from an audience perspective too, we’re already accustomed to, even if you upload an audience of email addresses, if it, if it doesn’t match to enough people you can’t use it. So I think we’re all kind of.
Already accustomed to ensuring that audience sizes are big enough. So this, I don’t think, I don’t think that particular part of it is going to be hugely impactful because it shouldn’t be really, really different from what we’re already used to.
Frederick Vallaeys: And then here we have a question. So if a cohort is not large enough what about its availability?
So again, let’s dive into some of the technical details here to help you understand that. So the cohorting or the FLoC. Scheme is currently set to renew every week. So every week Google would recalculate what cohort you fall into. Right. So it’s, it’s more likely that this would be treated like a low search volume keyword, which means.
that you can still have the keyword. And at some point it just not serving, but you don’t really know that. Google just gives you a flag that maybe it’s not eligible, but then the next day, all of a sudden it could become eligible because more people are searching for that keyword. Same thing with these the FLoC mechanism, right?
So there’s a weekly recalculation of what cohort you fall into.
I also want to bring up Turtle Dove, right? So Nava, you kind of mentioned it, and I think we focused on FLoC a little bit so far. So FLoC, again, it is about what behaviors have you done, what websites have you visited, and that assigns you to a cohort. There’s a separate initiative, which is called Turtle Dove.
And this is actually many iterations in, and all the iterations have been called a species of bird. But this is for remarketing. Okay, so two different types of audiences. They have different names. So kind of confusing, but turtle dog is how do you remarket to someone? This is actually really interesting to me because we’re, we’re talking broadly about third party cookies going away.
Right. So but if somebody has been to my website and I want to remarket to them, is that not a first party cookie? Like, why do we have to do something different here?
Amy Bishop: So Oh God, Amy. Oh, I was just going to say, so within Google’s ecosystem, it should still be considered first party because they’re also considering all of that logged in data.
So basically they’re walled garden. They’re considering first party data as well, where it starts to cross the lines with third party is when you start delivering ads on other publisher sites that you don’t own because third party tracking is a big part of how those users are tracked. So that’s where Fledge comes in.
And essentially what that strives to do. So fledge is an iteration of turtledove. Turtledove was proposed to allow advertisers or ad tech companies to create audiences or sorry, cohorts based upon their own first party data. To try to inform what adds that particular consumer or group of consumers might see fledge was built upon that was built upon turtle dove and basically it expands on turtle dove and they’re still continuing to iterate.
I still expect to see more iterations of this, but it expands on it to include a way for on device bidding algorithms to use additional information from a new trusted server. They’re kind of vague on what a Trusted server is but it would be defined by compliance with certain principles and policies designed for this purpose.
So they’re still trying to be really careful with how that audience data is used. All
Frederick Vallaeys: right. And one interesting point is that some of these initiatives have already been blocked in Europe under a lack of GDPR compliance. So right now these tests can only happen. Mostly in the United States. But yeah, it is fascinating because the, the other.
Point that you just made, Amy, is that the auction is partially shifting to the publisher’s website. And, and this is by the way, the whole thing, right? So to reframe the whole conversation a little bit, the debate that Google has operated for a long time in the way that the internet has operated is there’s the cloud.
Okay. And the cloud is basically Google’s infrastructure of servers and everything we’ve done, everything cool that we’ve been able to do in the last decade. That’s based on machine learning and artificial intelligence has relied on that supercomputer that happens to live in the cloud for us to be able to leverage that we’ve had to send data.
into it. So even if it was my data, my first party data, I had to send it to someone else to do something with it, and then they could send it somewhere else. Right. And that’s where the privacy generally tends to break down. Now the reason that this happened was because The only supercomputer that could do these machine learning things was the one in the cloud, but the world has evolved.
And now we have supercomputers in our pockets. We have iPhones and they’re much, much faster than anything people had obviously 10 years ago. So these devices can now do a lot of this processing, a lot of this classification and very simple machine learning. On the device itself. So it’s the shifting of all the machine learning having to happen in the cloud to some of it can happen on the device, right?
So that’s how you can build cohorts. But now it also sounds like some of that is shifting to the browser where the browser might be running on the computer and the browser itself is powerful enough to actually do some of the computation to make that auction work. So it’s a fundamental redistribution of where work is done in the internet that’s teeing up these proposals.
Navah Hopkins: One other thing I would just throw out there is that This is a really good time if you’re a publisher to take stock of your ad tech partners and your ad tech suite there is absolutely an opportunity to build your own solutions To serve to serve those ads and to engage with those cohorts That might actually be a more profitable equation and have better interface for you.
So I mean the native ad platforms Anytime you’re working with a Google, or if you’re talking about the Facebook, iOS ridiculousness, it, every single ad platform is going to make it very easy to buy ads through them. It’s, this is a time for you to disrupt potentially on the native side what those equations look like and what kind of relationships you’re willing to have serve.
One other note. I, I did want to touch base on when it comes to first party data when we think about domain structure. The, the, the main takeaway that, that I’ve, I’ve have from, from this whole equation is how many vanity domains, how many country domains, how many subdomains do we have and how many of those are actually gonna be considered in our first party data set.
So Jenny Marvin was very gracious. To let us know that we have five per first party data set but that that is another consideration as you’re thinking about how you’re serving ads and kind of what kind of relationships you want to build is how well you are equipping yourself to to receive those First party data sets,
Frederick Vallaeys: right?
So this is an article a post you did on search engine journal. So Let’s talk about that a bit more right? So you’re talking about subdomains and and really what qualifies as first party data And so your advice is basically if you’re the PPC team and you need to have your own landing pages because otherwise the SEO team gets pissed off that you’re breaking everything for them, it sounds like that’s completely fine.
However, there seems to be a limit, like Ginny Marvin from Google now is saying you can only have five subdomains connected to be like your first. So I,
Navah Hopkins: to be clear, And, and I am hoping Ginny will respond to this video and, and, and clear up if, if this, if this was misrepresented based off of the tweet.
And my advice is that there you can have up to five vanity domains. It does appear like the country specific domains and the subdomains you can have as many as you want. It’s just, you know, The, the core domain has, you can only have five. So why that matters is if you have, for example, your domain structure as bows USA and bows Canada, but all of them are.
coms, but the root domain structure is as opposed to, Just co. uk or ca or so on so forth that is going to require a migration that is going to require some some heavy lifting if however your your usa. bose. com or bose. com slash canada slash uk slash what have you that that also is is fine it’s just it’s the specific vanity domains that that can be an issue.
Frederick Vallaeys: Yeah, and then thanks for clarifying that distinction. It’s a pretty important one, right? The other point you made in the article, and I want to show this on screen here. You know, what is the. Value of audiences, I think. So I think, was this your research? This? Yes.
Navah Hopkins: This, this was a data set. I, I put together back when I was at WordStream based off of 21,000 Google accounts, at least a dollar of spend.
And a one thing that broke my heart was that 14% of those that were in. The cohort that I was looking at had no audiences associated and they had terrible performance. So it was just the, the, the main moral of the story was if you have no audiences attached, your cost per acquisition is going to be high.
Frederick Vallaeys: It’s a little bit hard to see, but the purple bar really high cost per acquisition. If you don’t do anything with audience.
Navah Hopkins: Whereas if you use customer lists, what I found really interesting was that. That first party data where you’ve earned the email you’ve earned that ability to communicate with the user beyond just following them around and nagging them to look at the ad like they they’ve actually given you that that email Had a better cost per acquisition than website visitors the similar audiences which we all tend to be oh poo poo on google Google doesn’t know how to discover good people for us or even facebook.
Poo poo poo. They they don’t know how to find us good folks but Those similar audiences actually performed from when you prospect off of that seed of the customer list or even the website visitors better than website visitors. So the, the moral of the story from, from this data, at least the lesson I took away use audiences whenever possible, but the best possible audiences that, that you will be able to have are, are, are, Those that you’ve earned to have that continued conversation.
Frederick Vallaeys: And there’s a question that relates to this, right? So the blue box in that graph right there is for similar audiences which also performs really well on a cost per acquisition basis. But Merlane is asking what happens with similar audiences, right? And so his question was a little bit if you have, I think it’s a cohort, but it’s too small.
Can you like add some similar audiences to that?
Navah Hopkins: So similar, similar audiences and, and this may change, similar audiences are going to take that seed audience and find folks that are as similar as possible based off of their behaviors. So, As a FLoC and cookies as the dust settles, we may see similar audiences improve and it will enlarge that that group.
So you’ll be able to target more effectively off of that similar off of the cohort. You may also see those change. And it’s you really would only want to do similar audiences off of your first party data exclusively because there’s a question of whether you’re The Google’s able to get enough signals with these changes.
So yes, root answer. Yes, similar audiences expand your cohort. But the caveat is that we won’t know whether they will maintain their performance level until after a couple of months where we’re fully immersed in this new world.
Frederick Vallaeys: A couple more questions. Amy, I think. Maybe you’ll have an answer for this one. Bastou is hopeful, I think, that for healthcare remarketing might become available because the data is more privatized, so Google might have fewer concerns. What’s your take on this one?
Amy Bishop: That’s a really good question. From a remarketing standpoint, it’s really hard to say. I think that they’ll still probably be really careful around sensitive categories.
But one of the things that really interests me is that beyond advertisers, the pledge documentation actually highlights that third party ad tech companies and Publishers could create their own segments that they could then make available to advertisers, and the documentation suggests even that some businesses could have the ability to create and monetize some of that data.
So one of the questions that I had asked Jenny Marvin in an interview that I did with her a couple weeks ago was whether she sees Google offering a third party audience exchange in the future. Just given that the documentation highlights some of those things, I was just under, I was just wondering if If Google ads are potentially broker some of that functionality.
And her response was that they plan to deepen their support for solutions that build on direct relationships between consumers in the brands and publishers that they engage with. So to me, that sounds pretty promising. That even for some of those categories that can’t currently remarket, there may be more audiences available to them.
Through publishers or potentially other ad tech which may be monetized. So that could be good or bad. But I think that there will be a lot of audience opportunities for us in the future, maybe even more than what we have now.
Navah Hopkins: One conjecture point. I don’t know that we will have access to that improved remarketing and targeting functionality until the platforms past the regulatory scrutiny that they’re currently in s O.
I don’t necessarily see any industries that are currently limited or not eligible to target magically being eligible to target because of FLoC in the privacy associated. I. A thousand percent agree that it paves the way for for a potential. I just don’t think it’s realistic to see those sorts of solutions until we’re a it’s it’s fully rolled out.
And be Congress isn’t and the legislation in in the in the EU breathing down the platforms next, right?
Frederick Vallaeys: So let’s talk about timeline. I mean, FLoC is happening now. It’s being developed, but when are we losing third party cookies and when should we expect GDPR compliance? Let’s conject.
Navah Hopkins: Well, I mean, 2022 that’s they, they said 2022. So I I’m planning on having all ducks in order and pushing all of the folks in my sphere to have their ducks in order by December 31st, 2021, because It is, I mean, it’s already getting rolled out to a certain extent. So.
Frederick Vallaeys: Right now is the time to start testing, start playing with it, start to understand it.
Yeah. Faces delays, well, great. And at least we’re ready ahead of time, but we can still use the old thing until it really goes away. But, but that said, I mean, so Google’s not the only player, right? So a lot of this you know, iOS is changing the rules and the apps are changing the rules. So in some way we need these solutions sooner.
Rather than later because we’re already being blocked on big parts of our traffic.
Amy Bishop: Yeah. And we have to also remember that FLoC is only specific to Google Chrome. So at least I haven’t heard anything about any other browsers adopting it at this point. So when you think about external ad networks outside of Google ads, This creates some complexity for them because they have to figure out how do they adopt FLoC, and then if other browsers don’t adopt FLoC, do they use different identifiers with different browsers, and then just from a GDPR standpoint the concern right now, I think, with FLoC is that what Google was trying to do is replace third party cookies with a way that Publishers and advertisers could still work together because they knew that if they didn’t create a way for advertisers and publishers to work together, then somebody else would, and that could be more invasive, and it could be more damaging, potentially even in third party cookies.
But the critics of FLoC currently are saying right now you still could do a fingerprint of somebody, and not only would you have their fingerprint, which already narrows down who they potentially are, but now you combine that with their FLoC. And now it makes it a lot easier to figure out exactly who they are.
So they’re still working out ways to make sure that it’s really protecting people’s privacy as opposed to doing the opposite.
Frederick Vallaeys: Yeah. So a lot of this is going to evolve. And then also. The FLoC mechanism. Again, it’s described, it’s an open API, so there’s no restriction that it can only be used by Google.
It’s just a matter of duty. Other browsers decide to adopt it. And I think there was just a story about what was it? Talk to God. It’s something
doesn’t ring about. I read something about somebody saying we’re not going to participate.
Amy Bishop: Yeah. That was that duck go. Yeah.
Frederick Vallaeys: And how does that actually matter? I mean, they’re. Through a search engine and they’re already like the private search engine.
Amy Bishop: Doesn’t really change much for me. I don’t think I guess I wasn’t really surprised by it.
They’re already, you know, so they already are so focused on privacy that I wasn’t surprised to hear them say they weren’t going to support it. But just given that they’re a search engine, I’m not exactly sure what signals I, I’m not exactly sure how the role of a search engine at this point,
Frederick Vallaeys: maybe, maybe a cynic, get some free press.
Navah Hopkins: That was my cynical response is that they’re, they’re just looking like when they, I don’t know if folks on that Twitter thread, that duck, duck go put out of, here’s all the, the things that Google was collecting on you. Look, look at them. Look how terrible they’re being. So I, I feel like a lot of the content that’s coming out right now that isn’t.
Concrete here’s what to do. Here’s how to do it is very PR oriented. Like as much as I’m very excited about the new attribution visibility that we’re going to have for display in YouTube there’s an email that went out to today. I feel like the timing of it was very. Pointed especially with facebook stuff.
So My advice is take any any communication any stance that an ad platform or browser takes as pr First unless it’s it’s very specific. Here’s what to do. Here’s why to do it so on and so forth Yeah,
Frederick Vallaeys: okay. Let’s take one more question. That’s not tactical and now we’re going to jump into exactly what to do.
Sorry, they just changed the comment on the screen. But but the question was so yes, could congress say we don’t like this yeah, I guess congress can do that.
Navah Hopkins: So net neutrality is a good example here. We all Feared or at least I feared net neutrality when when it was introduced when it was going to be taken away and and it like the fact that The government has such control over what is basically a human need at this point?
It was ended up being more fear than actual negative outcomes. So I, I don’t actually think it, the fear of Congress saying nope is warranted in this case especially because this is actually very compliant with. They like the angle that they’re pushing for. But yeah, that’s my two cents.
Frederick Vallaeys: Right. And I suppose that’s part of this whole conversation. And why I want to have this conversation is there’s a lot of unknowns, but more than at least in the PPC community, we can sort of understand what’s happening and communicate it clearly to others who are not in the PPC community. You know, let’s, let’s have a conversation that’s based on facts as opposed to sentiments, because eventually we know what congressman is not.
A PPC expert, nor can we expect them to be, but we want them to. You know, get balanced information that helps to make the right decision.
All right. Let’s get a little bit specific here and we’ve got an article from Michelle Morgan that we’re going to show in a minute but how will companies target ads to cohorts, right? So if somebody visits their site, do they have to target those ideas? And I think that’s actually the distinction here between remarketing versus audience targeting.
So for the remarketing to someone who’s been to your site, you’d use Fledge and Turtle Dove. And for finding people who haven’t yet been to your site, but seem to be interested and might be good visitors, you’d be looking at figuring out the cohort ID. And that’s the thing we don’t quite know yet, and that Amy was talking about before, so I don’t know if you want to say it again, but how do we target a cohort ID?
Amy Bishop: Yeah, so I think so essentially what’s going to happen is your browser is going to assign you into a cohort based upon your browsing history and that that browsing history will not be shared back. It will stay within your browser. And then as you visit different sites, if they request That cohort ID.
Your browser will then give them that cohort ID. But it sounds like it will be up to those websites themselves to determine exactly what that cohort ID means. And this is another thing that has I’ve seen it spark a little bit of discussion just around the fact that the sites will be able to acquire those cohort IDs, and they may be able to match eventually your name to that cohort ID.
So that’s There are some privacy concerns with that as well, but it sounds like at this point, they won’t be really receiving much information with that cohort ID. I’m sure over time, they’ll be able to learn things about those cohort IDs. If they see a lot of people coming from one, two, three, four are going to the SUV page, then they’ll start to understand that that’s a cohort of people that are interested in SUVs.
But it sounds like kind of out of the gate. There’s not going to be a lot of information provided in that sense. I have a a suspicion that on the advertiser end of things, does the ad platforms are going to step in and make some of that a little bit more simple simplified so that we as advertisers are not having to keep track of series of codes and numbers.
So I expect that that will be translated before it comes back to the advertiser, but that’s what the exchange between the browser and the, Publisher looks like but what’s so weird about that to me and like the unknown is how do you name or identify a cohort? when it could be based on such a variety of Jumbled together signals right because because all the examples I’m giving is like it’s an automotive buyer, right?
Frederick Vallaeys: But but that’s not me like I’m in or a car, but I’m also i really got into archery lately It’s really fun. It’s a great way to go into nature with the kids, but I’m like, Hey, maybe I need a range finder, right? So that’s another thing I’m shopping for. And then I really like certain sports teams. So I’m looking at that.
So how does all of that, that entirety of my behavior puts me into a cohort? Right. And you can’t see me anymore. So let’s put some other people on the screen. But you know, that’s the entirety of the cohort, but how do you specifically go after someone who is just looking for cars, right? And then you don’t care about the fact that they’re looking for a range finder to do archery, for example.
Navah Hopkins: So one, one thing to think about in market audiences have trained us by and large how to think about those general groupings. But in. This is where I think a lot of the lead gen folks should have a conversation with the e commerce folks. E commerce is used to thinking about product groups and product types.
And my guess, and again, this is a guess, is that these cohorts are going to behave in a very similar way of Getting tagged. And I in browsers understanding that something goes into XYZ product group or product type. And, and as we know, depending on the product group and product type that you designate for your product, your product will either end up on exactly the right search result page or exactly the wrong search result page.
So. Maybe this is me being naively hopeful but everything that’s been released, it, it feels like it follows patterns that we’re familiar with the idea of tagging something according to the, the largest common denominator the idea of Okay. This, this person is expressed by the places they’ve gone by the, the interactions that they’ve done various interests.
These people also express those various interests cohort. So, but again, this might be extremely naive and it could be far more technical than it, than that, but it’s based off of everything that’s been released. It does feel like it’s following the pattern of, of feed management and progress and product types.
And, and just traditional audience targeting.
Frederick Vallaeys: Yeah. And I don’t know the answer. I mean, some of the documentation I’ve read seems to be that Google is pushing the classification or the, the naming of the cohorts onto third parties. So I’m like, Hey, sweet Optmyzr can do that. We can help our customers with it.
In fact, I’d love it if Google didn’t do it at all. I give people another reason to stay with Optmyzr. But it doesn’t sound like that would be the reasonable thing to do. So but again, A lot of guessing at this point. And these are also good questions to ask, right? And I think, to be fair, a lot of PMs within Google don’t necessarily have answers at this point.
This is a big topic of discussion and they’re trying to figure it out. And, you know, as things evolve and as regulatory bodies respond, things are constantly shifting. But voice these concerns and, you know, make sure you’re using things in the right way. It was sort of funny, and this is unrelated, but the broad match modified.
Sunsetting like we did a study and we looked at we saw that 95 percent of advertisers Didn’t use it as it was intended, right? And then Google pulls it away So it’s like well, of course Google is going to pull it away because you didn’t use it the way it was supposed to be Used. I think with audiences clearly we all love them.
We love remarketing. We’re using it. Well but communicate it to Google, like make sure everybody’s on the same page and knows how important it is to you. So for example, the data that Navah had, like, what was it? Two or three X improvement in cost per acquisition. When you actually layer an audience on top of something like that’s meaningful and make sure Google understands like that, if, if you took away audiences altogether, right.
And my cost goes up 3x, well, maybe I don’t have money to put it in Google anymore. Maybe I have to put it somewhere else. Yeah, share all information was one of the principles of innovation at google because that makes people do the right thing Maybe
Amy Bishop: and one other thing I wanted to mention too is in my conversation with jenny marvin.
I had asked about All of the audience reporting that we’re accustomed to. So knowing how different audiences are performing within our search campaigns, knowing that which types of audiences are coming to our website, even if we’re not advertising on those audiences. And it sounds like that type of data will still be available when you, so when you think about trying to decide which cohorts to target, we should have some level of data at our disposal to understand Who’s coming to our site currently and how they’re behaving.
At least I hope so. That’s what it sounds like. So It won’t be kind of like completely playing in the dark, but I definitely think there could be a role for ad tech to your point fred to step in and help make that easier for advertisers as well
Frederick Vallaeys: Okay, we only have a few minutes left But what are you doing to get ready and any new ways that you’re building your first party data?
For when third party data becomes inaccessible.
Amy Bishop: So I wanted to just throw this out there. I, I know building up first party data is probably the highest priority on everybody’s mind right now. But kind of one sidestep from that, that I just wanted to mention that nobody’s really talking about.
But in terms of getting ready is one of the last big shift that we saw in privacy was when Google started mapping search terms and all of the documentation around FLoC talks about how we will no longer be able to see what types of sites that people visited or their browser history. And we know that as advertisers as our placement tracking.
So nobody is really mentioning this, but it really sounds to me like placement tracking is going away. So I’m kind of thinking of it as if you knew that search terms were going away. You know, just having all of that. Data that you had and making sure that you’re adding just copious exclusions so that in the future You already have all those negatives in place I don’t know how negative placements will act in the future But presuming that you can keep the negative placements that you have I would just say Go through your placements and make sure that you’re adding a lot of negatives Nobody has confirmed that they’re going away But all of the documentation says that we won’t be able to see negative placements People’s browsing history.
So to me, that says that they are. So that’s one thing that I would just say, I’m not seeing a lot of talk about it, but that would be a big focus for me, in addition to making sure that we’re building up our first party audiences.
Frederick Vallaeys: That’s great advice. And that also, I think, is another push towards more bidding, right?
We don’t need placement exclusions. We don’t need keywords even if we’ve correctly communicated the value of a conversion to Google. Right. And so that, that’s one of the things that I’m , I keep advocating. Automation is here to stay. It’s only getting bigger. The way that you get it to do the job correctly is give it all the information and that means correctly telling it what is a conversion and not just what’s the lead, but like, did that lead actually convert?
And when that lead converted, was it a good long term lead? high lifetime value, low lifetime value, like start to make those distinctions because that does feed back into all of this machine learning. And if it knows that a certain placement, even if it can’t tell you what that placement is, but it knows it’s a bad placement, it’s gonna be done.
It’s gonna be the correct value, right? So our focus has to be on managing the communication of information into Google as opposed to trying to micromanage everything.
Amy Bishop: Yeah, and I think Navah made a great point earlier, which is that automation tends to perform best when you give it a lot of data. So moving into larger cohorts versus some of the smaller audiences that we may be used to may actually benefit automation in that way.
Frederick Vallaeys: Quick question here. Could one consumer have many cohort IDs? So the answer to that, I believe, is no.
Amy Bishop: I don’t think so.
Frederick Vallaeys: And also think about one consumer is not really one consumer, right? So it is based on your browser. So it is that browsers, everyone using it. And probably in the logged in state. So the Google Chrome would know that, okay, here on this browser version, I’m logged in as myself, this is my work profile.
So that’d be two browsers, each having their own cohort IDs. But it would be unique and it would be refreshed initially once per week, eventually, probably more quickly. Nava, what are you doing to get ready?
Navah Hopkins: Well, number one I, I went and joined Justino so pretty much my, my biggest advice Is get, get your on page CRO house intact regardless of whether you use a software, regardless of whether you do this in house, regardless of how you do it.
I think a lot of folks make the mistake of only thinking about the campaign and the targeting and what they can do at the account level. And it, that, and not thinking about What does the individual user look like to Fred’s point lifetime customer value? What kind of messaging are you putting on on the page?
What kind of Cross department initiatives are you working on? So layering in your tech seo and your content seo teams your social teams so the the biggest advice I have is a get everyone together in a room and come out with a migration plan. If you’re a big company, I think the smaller companies are honestly going to be fine.
It like this will be one of those things that two, two Kings are, are going to, to battle and, and the peasants don’t, don’t really care. Don’t really notice is it could be one rule or another. They don’t really care. SMB is to be Frank. I don’t think we’ll be as impacted by this. Particularly since app platforms have built in so many amazing resources. It’s the larger folks particularly international folks that will want to get that game plan in place. So you’re getting your landing page URL structure intact, making sure that your email lists are GDPR compliant and that you have an easy way to feed them back into the app platforms for cross channel targeting, but also attribution.
Frederick Vallaeys: Makes a lot of sense. So kind of get GDPR compliant even if you’re not in Europe because everybody eventually is going to have to play by rules kind of like that amy, what about you any final thoughts?
Amy Bishop: Yeah, I would say like I mentioned adding negative placements, but then also just looking at all of the different ways that you can Start to build up those email lists.
I think are going to be really really valuable and to to nava’s point. Making sure that your CRO is top notch as part of that. And then also just potentially also looking at other channels that you may not have tested in the past to see if there are different lead gen opportunities for you that might allow you to generate leads at a super low cost that then could be leveraging other channels, both including email as well as paid ads in the future.
Frederick Vallaeys: Great. Final point for me. You’ve got our virtual conference on level May 18th at the bottom of the screen. That’s where you can sign up for it. So we’d love to have you join us, whether you’re a customer of Optmyzrs or not. Going to have great educational tracks as well as some tracks specific to Optmyzr product and roadmap. But yeah, thanks everyone for joining us. This has been a really fun conversation. Amy, Navah, it has been great to have you on again. Hopefully we can do this again when we know the next set of things about FLoC or whatever it’s called at that point. But yeah, we got a couple of months here to get ready.
Focus on your first party data. You know, get ready for a more private web because it is happening. And the ones of us who move quickly and are ready to go. When that happens, we’re going to be the ones doing really well in advertising. Competitive advantage. So thanks for participating in the session today, and we’ll see you for the next one.
Take care.
Navah Hopkins: Thank you.