Google’s U-Turn on Cookies: What Does It Mean for Advertisers

Episode Description

In this episode of PPC Town Hall, the discussion centered around Google’s recent decision to no longer deprecate third-party cookies in Google Chrome, despite previous announcements suggesting otherwise.

Ben Vigneron from Blackbird PPC discussed what this means for advertisers and the broader digital marketing landscape.

Episode Takeaways

Google’s Decision and Its Implications:

• Google has decided to delay the deprecation of third-party cookies in Chrome, which was initially driven by privacy concerns and regulatory pressures (e.g., GDPR, CCPA).
• This delay means that advertisers will continue to use third-party cookies for a while longer, but it doesn’t change the industry’s overall direction toward increased privacy and first-party data reliance.

Consumer Privacy and Choice:

• Despite the delay, Chrome will introduce a choice screen, similar to Apple’s, allowing users to opt out of tracking, which could lead to a significant reduction in available data for advertisers.

Shift to First-Party Data:

• Advertisers should focus on leveraging first-party data by integrating it back into ad platforms (e.g., Google Offline Conversion Tracking).
• UTM-based tracking remains relevant and should be utilized to measure performance across different platforms.

Advanced Measurement Techniques:

• Media Mix Modeling (MMM): A powerful method that correlates historical spend across channels with backend data (e.g., sales or subscriptions) to determine the true impact of various marketing initiatives.
• GeoLift Testing: Useful for isolating the impact of specific campaigns in certain geographic areas by comparing them to control groups.
• Both techniques help address the limitations of third-party cookies and provide a more holistic view of marketing performance.

Analytics Tools:

• Google Analytics (GA) remains useful but has limitations, especially with channels that don’t drive website sessions (e.g., LinkedIn lead forms, YouTube views).
• MMM can fill gaps left by traditional analytics by providing a broader picture of channel performance.

Future of Digital Marketing:

• As the industry moves towards more privacy-focused practices, advertisers must adapt by becoming more technical, particularly in data science and first-party data integration.
• Tools like R, Python, and open-source libraries (e.g., Meta’s Robin, Google’s Meridian) can help marketers implement advanced measurement techniques without needing to be data scientists.

Survival in the Age of AI:

• With the rise of Generative AI (GenAI), digital marketers should focus on leveraging these tools to augment their skills rather than replace them.
• Learning to integrate and use these tools effectively will be key to staying relevant in the industry.

Episode Transcript

Frederick Vallaeys: Hello and welcome to another episode of PPC Town Hall. My name is Fred Vallaeys. I’m your host. I’m also the CEO and co founder at , a PPC management software. Today we have an episode that’s hyper relevant given some of the recent news out of Google. Welcome Google Chrome was supposed to deprecate third party cookies, and they kept pushing it off year after year.

And finally, they announced that they will no longer deprecate the third party cookie. So the death of the third party cookie is overblown. But what does that really mean for advertisers? Will consumers continue to allow third party cookies? Or is it time we find new ways to do advertising? Measurement and all of the things we typically think a lot about in PPC and digital marketing.

So to talk about that and a lot more, we have a very analytical expert. His name is Ben Vigneron. He works for Blackbird PPC in San Francisco, and we’re going to go deep on cookies, media mix models, and a lot of other analytics for digital marketing. So let’s get rolling with this episode of PPC Town Hall.

All right, Ben, welcome to the show. Great to have you on.

Ben Vigneron: Thanks for having me. It’s a pleasure.

Frederick Vallaeys: All right. So Vigneron, where is that name from?

Ben Vigneron: Well from France it means winemaker. So it couldn’t really be more French really. And yeah, I’ve been you know, working in the U S for like 12 years now.

Prior to that, I was in Paris you know, in the, in the tech industry for, for years. And currently with Blackbird PPC, doing a lot of measurement for, for our clients. And, you know, obviously everybody, everybody’s been a little challenged by, you know, all the, you know, the back and forth about, you know, third party key duplication.

So, That’s what we’re going to be talking about today.

Frederick Vallaeys: Yeah. So and obviously there’s a lot of stuff that you write on search engine lens. So if people want to go read a little bit more about the topics we discussed, that’s a great place to start. You’ve also spoken at SMX advanced recently. That session is available for free.

On the SMX website, we’ll put a link in the show notes to that, but but we’ll be covering a few of those topics, but let’s jump right into it, right? Like we said, the most recent news out of Google is that the third party cookie will not be deprecated. So talk a little bit about that. What does it mean?

And, and does it actually mean that we should continue to rely on third party cookies?

Ben Vigneron: Yeah, that’s a good question. So, I mean, first of all, we rarely see Google change directions like that. They have rolled out controversial features in the past and they’ve stuck to it. Like, you know, enhanced campaigns.

And, you know, exact match type, not, not, no longer being exact search terms no longer being available in the reporting suite. So they, they have pushed features to end users and stuck to, to them before, and rarely do we see them, you know, like change courses. And so it’s interesting to, to, to see, I think really, you know, we should be thinking about, you know, advertisers, agencies, And obviously end users, when we think of the potential impact, I would say, long story short, we’re still going to be using third party cookies for a little longer.

It appears that there was some pushback, not from Google themselves, but from other antitrust committees. Thinking that, you know they would gain an edge over the rest of the industry by deprecating support for third party cookies before others. Right. So it’s kind of interesting how the whole thing is working.

Like focusing on data privacy is now giving an edge on the competition. And we’re seeing that with like Apple advertising on like Safari. Their browser, right? Being being, you know, data privacy focused, right? And that’s that’s Something that people care about these days like duck duck go release a browser that you can, you know Download and use and you know, no tracking no cookies It’s you know, it defaults to like denying any any any permission whatsoever so it seems to me that the overall direction the industry is taking has not changed with this decision However In the short term, things are slowing down.

Google is, is, you know, you know, pushing the brakes and saying, okay, we’re going to give it, you know, a few months, maybe a couple of years before we go you know, full speed with like no more third party cookies. However, they’re definitely ready. Technically, you know, GA4 is all about that. And then.

You know, they, they, they want to be ahead of the curve. They want to be leading the industry. They want to be ahead of GDPR and CCPA. And right now we’re seeing that, okay, the industry is not quite ready. Antitrust committees are saying, Hmm, it could, you know, give you an edge that you already have, like an extra edge.

So let’s, you know, let’s, let’s, let’s wait and see. I don’t think there’s going to be any drawback for you know, And users really. So I think they had more to, to lose pushing this for so they are just giving it some time.

Frederick Vallaeys: Yeah. A lot of very interesting stuff in that to unpack. So obviously this all derives from the push of GDPR and CCPA.

As far as giving users privacy choices. And so one of these, the fundamental thing here is choice, right? It’s not about saying you cannot deploy third party cookie technology. It’s about giving users the choice of whether that’s working for them or not. So I think one of the big things that we should not lose sight of is that while Google has said, well, technically we’re not going to change how the Chrome browser works, we are going to introduce a choice screen, which Much like we see on Apple devices today, that every time you install an app, the user gets a thing that says, Hey, would you like to allow tracking across all apps?

And the number that I heard is that about 70 ish percent of users reject that cross app tracking, right? So you lose a lot of data. And so what’s interesting here is Chrome is going to have a similar. Warning screen. We’re probably going to lose the data from the third party cookies, even though third party cookies technically will still exist.

So what do we do with that?

Ben Vigneron: So, I mean, we should be prepared. So, so we’ve seen that happen with you know, Apple iOS letting. Users opt out from cookies and so, or advertisers, what happened is they lost conversion tracking from iOS users for the most part, maybe, you know, 70 percent opted out roughly depending on the industry, but the vast majority.

So you need to find other ways to measure performance, whether it’s like online or offline, but you need to find some way to evaluate your, the efficiency of your program. So there are a couple of alternatives and I think, you know, there are new technologies out there that are focusing on just that, but long story short.

Without third party cookies, you need to use your first party data. And so it truly means that you need to first pipe down your first party data back to the ad platform. So we’re talking Google offline conversion tracking. The name has changed. I don’t remember exactly, but essentially passing back your backend data to the ad platforms.

That’s, that’s become huge on Facebook offers the same feature. LinkedIn. Is working on the same feature. They will all have similar names, but surely that’s number one. I would say just don’t trust ad platform data anymore. Be skeptical, challenge the numbers and use your own data and try to pipe it down to the ad platform.

So to me, that’s number one. Number two is UTM based tracking is not that. Right? If you’re tracking parameters, observe all sorts of data privacy concerns, right? Which you should, right? So if all the data you’re passing in your URLs and your UTMs is anonymized or just anonymous, like we’re talking, you know, a campaign name, maybe an ad ID, stuff like that, we can’t tell who the person is.

So it is absolutely 100 percent compliant. And those tracking parameters is. On that, they will still work fine, and that will help you, you know, measure your performance in your HubSpot, in your GA, in your Salesforce, in your Shopify instance. So, I think for those who are not using UTM based tracking, that’s definitely something to consider.

The last piece, which is dear to me, is, you know, advanced measurements. So, essentially, using data science. To really get to the bottom of the true influence, the true contribution or the true incrementality of any initiative. So that means using methods such as or media mix modeling, where essentially we’re trying to correlate all your spending across all your initiatives.

Against your first party back end numbers, whether it’s your Salesforce opportunities, your Shopify subscriptions, whatever the outcome may be. This is what is about and it’s extremely powerful and it will really Bypass any click based third party cookie based tracking challenges, right? So it’s, it’s super robust in that regard.

And then there are other methods where, you know, you would just like run a test and see what happens in your backend still. So, so really first party data is, is the future, if not already the present. And there are ways to pipe down that data. So that’s called markups or marketing operations. So you’re going to need someone to.

Get this done and then you’re going to be able to have all those very, you know, high quality reliable signals used for algorithmic bidding in the Google, Facebook, LinkedIn platforms and then outside of that try to take a step back and look at whether Your search campaigns are doing anything for you as a business Is your social media doing anything?

You can look at paid, you can look at organic, you can look at your, the effect of promotions so that’s what is all about trying to get a holistic picture of what is moving the needle for you. Again, we’re not using any third party cookie data. We’re really just using your backend numbers, which we try to correlate with your historical spend for the most part.

All the signals can be used as well. That’s, that’s pretty much it. And that’s really what I love to do. I do this for, for, for my clients. So and Geolift testing are the two things we use a lot. Either to have a holistic picture of what’s moving the needle or to run a test in a silo and say like, for instance, YouTube, we’ve never, what we hear a lot is we’ve never seen any value from YouTube.

We don’t see conversions coming from YouTube. We have UTMs in place. We don’t see any data categorized as coming from YouTube in our back end. Right. So we don’t think YouTube is doing anything for us. And then if you run a GeoLift test, which essentially. Think of it as you run your YouTube ads in some locations, not others.

And you’re going to look at the delta between your groups. You’ll pretty much, every single time, you’re going to see some lift in those locations where you’re running YouTube ads. As long as you’re running, you know, spending enough money, like your test budget should be large enough. And as long as you know which method to use to measure that data, and then you’re always going to see an uptick in sessions, new users, and, and potentially purchases, subscriptions, MQLs, opportunities, right?

Depending on latency. So there are a number of things to account for, but. Jewish testing is, is a great tool to look at one individual channel or sub channel in a cell and, and measure its true impact. Is all about getting the overall picture.

Frederick Vallaeys: Okay. So yeah, a lot of stuff in there, right? So you talked about piping the data through OCI, setting up UTMs, and that’s the more basic stuff.

And there’s plenty of blog posts that we’ve done on and that you’ve done where you can learn more about that. So let’s go a little bit deeper. on those two areas that you really are passionate about. So the GeoLift test and MediaMix models. So one thing you just said was with GeoLift test, you have to have enough data.

So let’s talk a little bit more about specifics. Like How big of an advertiser do you have to be? What kind of data might you need on the back end? And, and, and generally debated. I’ve always thought about these split tests as you want to control as many variables as possible. So doing a split test between say, California and.

Minnesota might not be the best because we have completely different climates, different parts of the country. So more often I hear it’s like Seattle versus Portland cities in close proximity, but largely same demographics. So that controls for those variables. So talk a bit more about those specifics.

If someone wanted to do a geo lift test.

Ben Vigneron: Yes. Good, good question. So yes, like you said you want to make sure your tests locations are carefully selected. So that they are similar to your control locations, right? You want to, to, to groups, test and control to be similar statistically speaking, and that means geographically as well.

And I mean, in lots of, you know, every, every single aspect of it. So overall, if you’re, if you’re running a test in the U S there are 50 States. So potentially, and most advertisers, they will they will randomize the groups and they will pick 25 States. If you do this, you’re likely to fail. Randomizing your groups might not work.

For instance, you could have California and New York in the same group. You probably don’t want that. You probably want California and New York in control or the other way around. Then if you randomize your groups, you could very well have more States from the West coast, not so much from the East coast.

So it’s tricky. You could also use DMAs instead of. States, although state is more common as a dimension in any data source. So typically state is good for that reason. Like GA, Shopify, Salesforce, they all have state. They don’t all have DMA level. Otherwise I would use DMAs more often because it’s more granular, but yeah.

So if we’re, if we’re thinking state, the way you want to pick your states is first of all, your test group should account for at least 20 percent of the country in whatever metric you’re looking at if it’s you know, we’re talking like purchases, right? So you want to make sure that going in your test group represents roughly 20 percent or more between 20 and 50 percent of the overall country.

Now that’s not it, right? Now you need to make sure those potential test locations have behaved similarly to your control locations historically. So what you want to do is measure the existing data between your test and control group pre intervention. We haven’t run the test yet. We just want to make sure there is no difference between our groups going in.

Right? So that any to be detected effect can effectively be attributed to the test. To our intervention, right? So I’ve put together a script in R that will do that for me, where essentially I’m going to randomize the states and I’m going to run a statistical test to minimize the variance across groups.

And I’m going to select those test locations that, you know, drive the lowest variance. Which means my test and control group, day over day, they tend to behave similarly, historically. Which means in, in the future, when I run my, my test, my intervention, I know that if I see some kind of delta, I can attribute it to my, my intervention, my, my experiment, right?

So it’s all about looking at your X Circle data and minimizing the variance, which is essentially the difference in performance across groups. So it’s a little technical and it’s super cool. If you don’t do it, you might see something. You might say, okay, I’m picking that one state. I’m running my YouTube ads in that one state.

And I think I’m seeing something you won’t be able to establish causality. You’ll only be able to kind of. inform you and your team that, yeah, it looks good, but you won’t be able to, to confidently say this is working. And I can confidently attribute that uptake to that experiment. And with the method I’m talking about, and by the way, I’m using Google’s own causal impact method for this, which was It’s an open source library that was put together exactly for this to to measure the impact of a marketing intervention as a time series right across two groups.

So that’s just one use case, and that will tell you exactly what the lift is like and how confident you can be with the results, right? And so that’s that’s what I’m using. Pre intervention to make sure my groups are similar going in. So that’s, that’s, that’s it. There’s, that’s why I mentioned about data science.

It is a little advanced and you know, once you have the script put together, it’s no longer, you can have the whole team. Put it together themselves without getting into the off script, but it’s, it’s definitely a little advanced. It took me some time to put together in a way that works every single time.

But now what background do you have? Are you technical by education or how did you learn how to program? So not really. I mean, I did study mathematics, but I really learned how to code at Adobe. So I joined Adobe without any SQL or R Knowledge. And you know, the first two years at Adobe, I really focused on that.

I was a business analyst at the time and we started working on a, on a, on a platform that would actually analyze a bunch of data and spit out a few, like a bunch of like essentially PowerPoint presentations for. The customer success team to use and share with their clients. So that’s, that’s how I learned, you know, SQL and all.

And I really, I think I met the right, the right people who really got me passionate about it. And, and I really got into it. So I’ve started reading books about it and You know, I’m learning by doing still like to this day, I’m learning new functions, new, new visualizations almost every day. And it’s, it’s really, really cool.

And, you know,

Frederick Vallaeys: how do you learn about new things?

Ben Vigneron: So typically it starts with a use case. I’m trying to solve a problem. And then I’m going to, you know, search for it. So I have a couple of books at home, but then I just use Google. And then those websites I know about already. So I’m going to tend to trust those websites more.

And typically, there are multiple ways to solve the same problem. Which is really the beauty of it. Because then you can wrap your head around like, Why is this method relevant for to that case? And that’s what I do. Like, if you think of like first party data and problem I’ve had not too long ago was trying to figure what are the main personas for our cons, right?

So essentially, Profiling or clustering, if you will. And it’s, it’s a very complex issue. And so R and Python also, they have many ways to solve that problem. And which tells me one method is, is the best one, right? They can’t all be equal. There must be one method that, so I run all of them. And I, I, so I get into the, you know, the scripts.

I load my data, I run it in different ways, and then I’m going to pick the one that. I feel more comfortable with that in terms of the insights and my ability to explain to a team how it all works, but yeah, it’s, it’s, you know, R is an open source programming language. It’s very easy to find some great documentation especially for remote workers.

It’s very easy to, to learn about it. And it, you know, it can. Come across as highly technical in the first place, and it is but, you know, after a few hours into it, like, you know, you’re like, it’s, it’s, it’s single nature.

Frederick Vallaeys: What about code interpreter from GPT? Are you using that at all to write Python scripts?

Ben Vigneron: I do not I have tried and I was unimpressed. To be honest, like most of the time. So I use this with R mostly and it did help me a little bit. But for the most part, it just wouldn’t work. So I had to fix it, right? So it does 80 percent of the job and then you have to get into the script and fix it yourself, which is still a time saver, but I would still use it.

Like if I’m looking for, if I’m confident with my prompt, like I know this is basically what I’m looking for. So I’m, I can I can enter that in GPT and I think it should be useful. But for, for the most part, I would not trust it too much. It, it, it just, it might help me figure out like how to make a visualization look better, you can just say.

Can you make this plus look like this and that and it’s going to, it’s going to kind of happen, but yeah, from my experience, it, it really still requires a lot of manual input, which I’m okay with. I

Frederick Vallaeys: agree with that. I I was recently thinking about, I talk a lot about Google ad scripts and how GPT can write them for you.

And. It’ll, you’ll copy and paste the script into Google ads. It’ll come back with errors. You give those errors back to GPT and it’ll fix it. And after a couple of iterations, you get something that produces output. But the thing is you really have to check the output because it may have had errors along the way.

And so my, my premise is that if you have the capability to do it already. GPT can make you faster at doing it, and you can take it like you said, the last 20%, you can make those fixes manually, or you know what it’s doing wrong, you can tell it what the mistakes are, and then it’ll actually fix them. I also think that if you’re willing to learn R or Python or statistics, GPT can be a great teacher.

And you can ask it, what is the right model? Why do you think this is the right model? Where can I learn more about this model? And I think where it starts to fail is in the things where you have really little interest or capability to learn it yourself. If you’re just going to trust it to do a hundred percent good job, that’s where you’re going to fail.

Ben Vigneron: Yeah, it will augment your existing skills, but you still need the skills to begin with, right? Because you’re going to need to fix the output anyways, right? So I totally agree with you. And I would add also, like, YouTube is a great place to learn anything. It’s crazy. There are tutorials about, like, how to set up Python on your machine and get started.

And it’s amazing and it’s free. So that’s also a good, you know, starting point. Obviously it won’t go very deep, but that’s a good starting point. And then you can get into like the Coursera, LinkedIn, whatever, like get into that, buy some books and, and just, and just learn by doing. So there’s so much knowledge out there.

It’s, it’s very exciting times. If you were going to start in a digital marketing career today. What would you learn?

So I guess, I mean, it kind of depends what you like to do, I would say, but for me I have enjoyed getting more technical all the time. So I would definitely study data science from the get go in school.

And so statistics. Or Python, I wouldn’t, you know, a more formal education around that, I think is, is, is really powerful. SQL as well, although SQL is fairly easy, but if you want to reach a more advanced level that you could, you could use some, some training. So those are the things that, you know, I’ve learned.

Mostly at Adobe, and it was great, I think, because, you know, we had, like, business cases to, to solve. We, I was not alone, I had a team to, to work with, so that was great, but I think if you are interested in marketing and you’re still in school, go as technical as you can. Those technical skills will pay off over time because most marketers, they don’t know as much about, you know, track how tracking works, how to look at the data, how to run a test.

And those are the areas where you can really make a difference. And again, build trust around your results and not only say, Oh, My CPA is down in Google ads, but also consider like marginal returns, diminishing returns which is a concept we haven’t touched on yet, but it’s, it’s very important. Then the concept of incrementality.

Is it all those conversions real would they have occurred anyways type of stuff? Those are the things that truly matter at the end of the day for the cmo Yeah,

Frederick Vallaeys: and all of these things depend on having the right data the right first party data Which you’ve kind of mentioned several times. You also mentioned there’s a data ops team.

So talk a little bit about What that should look like if someone’s not really thinking about first party data a lot. How do you bring it in? How do you start it? Is there a tool where you collect it? How do you pipe it back into google ads? What you do?

Ben Vigneron: Yeah. Yeah, absolutely. So so over the years i’ve seen a lot of Of, of, of changes there with, for instance, our clients using like Zapier to Zapier is an amazing tool to connect different data sources together.

And essentially that’s, you know, for instance, for B2B advertisers LinkedIn is a powerful platform and, you know, people like submitted their email. Maybe their job title, maybe something else. And then the data would typically get kind of lost into the abyss. It would stay in LinkedIn and not necessarily shared with the back end data, right?

So that’s, that’s definitely something that is challenging because in that case, your best data is not in your own back end. Your best data is owned by LinkedIn. Right. So that’s, that’s something you definitely want to tackle and you want to make sure you are capturing all that data. And that’s something that a tool like Zapier, I don’t work with Zapier.

I, there might be other tools out there that do the exact same thing, but I know Zapier is fairly easy to use and works. It’s pretty robust from what I understand. And you can just capture all the details from your lead from LinkedIn to your backend. And then you can use that data internally for whatever purpose, right?

So this is a role that didn’t really exist when I first started, you know, we would have campaign managers managing campaigns in Google and Facebook and LinkedIn, perhaps, you know, all of them, but they would really trust in platform data and they would not even have access to the backend. And now we have teams that’s really bridged the gap between, you know, backend first party data and at platforms.

Newsletters, anything, right? And so this is a role that has become more prevalent because you need to break that gap both ways. You need to capture all the data from the ad platform to your back end. And you need to also pipe your back end data, perhaps enriched, right? You can qualify your lead. You can do all sorts of things back to the ad platforms so that the ad platform can, you know, ingest the signals and find users for you in a, in a meaningful way.

So this is definitely something that I’ve seen happen over the past few years. Marketing operations are definitely. Something that is fairly new. And it’s, you know, I think it’s great to have people like help you bridge that gap. Now, I would say from my standpoint what we’re doing here is all the data we’re passing around is mostly click based.

And so that doesn’t solve for the underlying issue of incrementality. What is the true influence of whatever initiative? And that’s where GeoLith testing are relevant because they solve that. So let’s talk more

Frederick Vallaeys: about right? So now you have all this great data you put in the right places and we’ve lost.

cookie data so we can’t necessarily know on an individual basis which click led to what result And that’s where comes in. It gives you the bigger picture. So what does that really look like? How does that work?

Ben Vigneron: So typically you you’re gonna want to have two years of data and by data, I mean historical spend across channels.

So say you’ve been running Google, Facebook, and LinkedIn, right? So you have three data points, your historical spend for Google, Facebook, and LinkedIn, and then you have your backend, you know, subscription volume, right? So all of this by week. So you have three inputs, your spend for across three channels and you have one output, your.

numbers, for instance overall, right? Not attributed to a single channel. We are, we don’t know that yet. That’s what we want to find out. And so what we’re going to do is what’s called a response decomposition where we’re going to. Essentially correlate your weekly subscription volume against your historical spend across those three channels, right?

And so, one, there are two ways that are accessible to the public currently that are open source libraries. There is metas. robin. Library available in R and Python, and you have a Google’s Meridian. So it’s funny to see the publishers are giving access to those amazing resources and there’s a ton of documentation.

So I’ve been using Robin for a long time, which I’ve customized over the years. So that, you know, the response curves look a certain way that I know, or, you know, I, so I kind of supervise the model if you will, but essentially it’s, it’s pretty simple. You don’t need to run a test to run all you need to do is pull your spend numbers of the past two years across all meaningful channels you’ve run.

Could go beyond Google Facebook, LinkedIn, maybe emailing was a top priority last year. You should include that. And then your backend numbers, which you should have, right? So, so typically it’s, it’s really a matter of just putting everything together. And then you can run your analysis and you’re going to be able to say, when I get a harder subscription, really 10 come from Google, 20 come from Facebook, five from come from LinkedIn, 20 come from seasonal trends, and some are unattributed, we don’t know.

Obviously you want to tackle that unattributed bucket as much as possible, but that’s what, that’s what there is to it. And it’s, it’s, it’s great because. Again, spend is always there. That’s, that’s not debatable. You know how much you’ve spent, right? It’s not like tracking. And your backend numbers, they should also be there and available for you to, to use.

Next, what you want to do is compare what you found through against your traditional reporting. Okay, over the past few months, we thought, you know, our, our Google ROI was X, but truly is telling us it’s Y. Is it higher? Is it lower? That’s something you want to look into and maybe some maybe run a geolist lift test if you want to establish causality.

Those are two methods that really work well together. What I would say is Attribution is a zero sum game and using third party cookie data to measure performance is mostly incorrect. Some channels will be overvalued, some channels will be undervalued. So the name of the game is to, you know, identify which channels are being undervalued or overvalued.

And that’s what is about. And I would, you know, advise anybody. In the marketing world to get into this. Cause it’s, it’s really what matters at the end of the day.

Frederick Vallaeys: And a great advice. I think a lot of people don’t use media mix models quite yet, and it sounds like it’s actually not that difficult to implement using the open source systems like Robin, and if you have the data, but you should already just put that in.

So. But that said, most people, most advertisers do have Google analytics set up that is shifting much more to a modeled view of what’s happening in the world. And one thing that is nice is that it does actually, in many cases, deduplicate, right? Because it does look cross channel. It’s not like Facebook and Google and Microsoft will each take credit for that same conversion and triple counted in that case.

But what is your take? Where does analytics still fit in given that it’s really changed from what it used to be? Is it still useful? How do you use it?

Ben Vigneron: It is useful. I think it’s useful for any initiative that is, that will drive a session to your website, right? Which is not always the case, right? So if you think of those LinkedIn lead forms, users never go to your website, so that’s not going to show in GA.

If you think of YouTube video views, those will not show in GA. You might, if they click, you will, but they mostly don’t click, right? Your TV ads. Your radio commercial radio ads out of home, all the traditional media, none of it will show in your GA. So depending on what your mix looks like, GA might be a great tool or not so great tool.

From my experience like B2B advertisers that use a lot of LinkedIn lead forms, they don’t choose GA as much. They don’t rely on GA as much. They will use GA for their Google campaigns, maybe their Facebook campaigns if they run any, but. Yeah they don’t work too well together linkedin and gm

Frederick Vallaeys: and so to deposit that really briefly.

So your point is that in Because you have the linkedin cost data associated with those On linkedin form fills and you have your back end data on how many new customers you got The can actually attribute that back in a way that ga could ever

Ben Vigneron: Yes, exactly. Because your input is only the, the, the spend numbers, right?

So how much you spend on LinkedIn and your output is only the overall KPIs or subscription purchases, whatever. It’s not attributed yet. Right. So if LinkedIn played a role, you’re going to see an uptick during those weeks when you spend the most on LinkedIn typically. Right. So yeah, you’re, you’re just skipping the whole the whole challenge by, by just looking at straight spend and straight backend numbers.

But yeah, I think GA is still great though for lots of reasons, but yeah, like especially, you know, the integration with Google ads, obviously. Is native. That’s, that’s pretty great. But otherwise you know, if you’re running a lot of initiatives that don’t drive sessions to your website, then GA is just irrelevant.

So that’s, that’s, yeah.

Frederick Vallaeys: Great. Well amazing stuff. Thank you for showing us a little bit about and GeoLift testing and how that can be applied. Also some hopeful words there that you don’t have to be a data scientist, but there’s lots of great materials on YouTube and Udemy and other places to really learn how to use R.

Personally I’m a big fan of GPT Code interpreter, and I’ve certainly done some statistical analysis that would’ve been beyond my capabilities. So all of that said, Sam Altman said 95% of digital marketers will lose their jobs due to gen ai. The the core idea here is how do we, how are we to 5%, not to 95%?

How do we keep our jobs? Any piece of advice there for how to remain relevant.

Ben Vigneron: I’m pretty optimistic, to be honest. I think Gen AI is really about filling the gaps and not necessarily doing the work that humans do. You want to know how to use those tools. If you’re the one using the tool, you’re not going to lose your job.

Someone has to use the tool. And then it’s all about integrations because typically, you know, you have you’re working in like RStudio you have ChatGPT on the side. But how do you integrate both? Right? So if you focus on like using all the tools that are available and relevant to you and you learn the skills to integrate those tools into your daily workflow, you’re definitely going to keep your job.

I’m not, I’m not worried about that at all. Yeah, I think it’s going to be a slower transition than some would think. You know, we didn’t have Excel and we did, we didn’t lose our jobs. You know, see what I’m saying? Like SQL was not around. We could still find out a few things now. I mean, I’m not very worried about it and it’s going to be a slower transition.

However, focusing on technical skills, I think is, is definitely. Relevance using those tools, learning how they work under the hood is, is great to know. And then again, integrations.

Frederick Vallaeys: Yeah. Yeah. You’re not competing against a tool, you’re competing against other people who know how to use those tools.

Ben Vigneron: Exactly. That’s what I’m trying to say. Exactly. Yeah. I agree.

Frederick Vallaeys: All right. This has been amazing, Ben. If people want to connect with you LinkedIn, any other places?

Ben Vigneron: So you can email me at ben@blackbirdppc or you know, find us online blackbirdppc.com. But yeah, my LinkedIn is fine. So Ben Vigneron and yeah, I’ll be happy to connect for sure.

Frederick Vallaeys: Yeah, especially if people love the ideas of and maybe don’t want to do it themselves and need some expertise. What does Blackbird PPC focus on?

Ben Vigneron: So we do historically direct response marketing. So very our eye focused for so B2B and B2C, we’ve been over indexed toward B2B over the past few years, but we, we can do a B2C as well.

I would say we do some campaign campaign management. But also obviously through me and my team you know, advanced analytics and measurements. So we’ve talked about GeoLift testing, but there are a number of other analysis we run for our clients. So defining personas that drive high lifetime value things of that nature, finding out what are.

The creative attributes that drive above average CTR or commotion rates. Those are things that we can run internally. So I would say you know, campaign management plus advanced analytics how to apply data science to marketing really. And so that’s, that’s what we do where in San Francisco, the team is great.

Frederick Vallaeys: Great. So, yeah. And again, if you want to see more of what Ben has been working on, check out his posts on search engine land, go and check out his SMX advanced session that is available for free, just a simple registration to get access to that. If you’ve enjoyed this episode of PPC town hall, and you want to see more of them, please use the subscribe button at the bottom and we’ll notify you when the next one comes out, you can also go to ppctownhall.

com and sign up for our newsletter, and And then if you want to be more efficient as a PPC manager, we have lots of great automations and we have first party data integrations and as well. We’ve got a free trial. So check that out. Thank you for watching Ben. Thank you for being on and we’ll see you for the next episode.