Nerd Alert: Why Your Hyper-Targeted Ads Might Be a Waste

Summary

Episode Summary: "Nerd Alert: Why Your Hyper-Targeted Ads Might Be a Waste"

Release Date: August 7, 2025
Hosts: Leonard Jasper and Rob Demars
Podcast: The Marketing Architects

---

Introduction to Nerd Alert

The episode kicks off with Leonard Jasper and Rob Demars diving into the "Nerd Alert" segment, where they explore complex marketing research and distill it into actionable insights. Leonard sets the stage by introducing a study that scrutinizes the effectiveness of third-party consumer profiling and audience targeting in digital advertising.

Understanding Data Types: First-Party vs. Third-Party Data

Timestamp: [02:24]

Rob Demars humorously explains the difference between first-party and third-party data using a cookie analogy:

"First party data is when you have the information in house... Third party data would be like buying cookies from a guy in the back of a Toyota Corolla... you might get something unreliable."
— Rob Demars [03:02]

Leonard emphasizes the importance of distinguishing between data sources, clarifying that first-party data pertains to information directly collected from your own customers, while third-party data is acquired from external sources, often leading to inaccuracies.

The Study: Assessing Third-Party Data Accuracy

Timestamp: [02:43]

Leonard introduces the study titled "How Effective is Third Party Consumer Profiling and Audience Delivery?" by Nico Newman, Katherine Tucker, and Timothy Whitfield (2018). The research aims to determine the actual performance of third-party data in targeting specific demographics and interests.

Study One: Programmatic Campaign Simulation

Timestamp: [03:17]

• Method: Simulated a typical programmatic campaign targeting men aged 25 to 54 using six major Demand Side Platforms (DSPs) and 19 data brokers.
• Findings: On average, 59% of ad impressions correctly reached the intended demographic. While better than the random baseline of 26.5%, the accuracy varied significantly across different DSPs, ranging from 40% to 72%.

Leonard notes the modest improvement over random targeting but underscores the variability and potential inefficiency.

Study Two: Raw Accuracy of Data Segments

Timestamp: [04:33]

• Method: Assessed the accuracy of demographic labels (age and gender) from third-party brokers against verified self-reported data from consumer panels.
• Findings: The average accuracy for identifying men aged 25 to 54 plummeted to just 24.4%, worse than random gender guessing. Gender accuracy averaged 42%, and age accuracy ranged between 10% to 32%, highlighting significant inaccuracies.

Rob reacts with surprise:

"Wow."
— Rob Demars [05:27]

Study Three: Interest-Based Targeting

Timestamp: [05:28]

• Method: Shifted focus from demographics to interests such as sports, fitness, and travel.
• Findings: Interest-based targeting showed improved accuracy, climbing to the 70-80% range. While still imperfect, this method outperformed demographic targeting.

Cost-Benefit Analysis of Third-Party Targeting

Timestamp: [06:04]

Leonard discusses the economic aspect, revealing that third-party audience targeting often doesn't justify its cost unless applied to higher-priced media. For lower CPM (Cost Per Thousand Impressions) platforms like display advertising, the performance gains do not outweigh the additional costs.

A standout quote from the study:

"Third party audiences are often economically unattractive."
— Study Highlight

Leonard interprets this as third-party data frequently not offering sufficient value for its price, leading to mismatched impressions and wasted advertising spend.

Practical Implications for Marketers

Timestamp: [07:30]

Leonard advises brand managers and media planners to critically evaluate their use of third-party data. Key considerations include:

• Accuracy Verification: Ensuring that the targeted audience truly aligns with the intended demographic.
• Cost Justification: Weighing the additional costs against the performance benefits.
• Testing and Skepticism: Experimenting with and without third-party data to assess its real-world effectiveness for their specific brand.

Rob adds a relatable perspective:

"Buying a third party data set is like paying for a VIP concert ticket and ending up in the parking lot."
— Rob Demars [08:54]

He underscores the frustration of investing in data that fails to deliver the desired audience, likening it to a disappointing concert experience.

Counterarguments and Final Thoughts

Timestamp: [10:31]

Leonard acknowledges that while the study presents a compelling case against third-party data, some scenarios may still warrant its use, particularly when targeting high-cost media where the additional expense is offset by the media spend. He advocates for a balanced approach, suggesting that marketers remain open to testing third-party data sets but maintain a healthy skepticism regarding their effectiveness.

Rob concludes with a succinct takeaway:

"Not all third party data is created equal."
— Rob Demars [11:46]

Leonard echoes this sentiment, encouraging listeners to evaluate the quality of their data sources critically.

Closing Remarks

The episode wraps up with acknowledgments to the production team and invitations for listeners to connect on LinkedIn and leave reviews. Leonard and Rob sign off by reiterating the importance of building effective marketing strategies based on reliable data insights.

---

Key Takeaways

1. Third-Party Data Accuracy: Significant inaccuracies exist in third-party demographic targeting, often performing worse than random targeting.
2. Interest-Based Targeting: Offers better accuracy but still falls short of perfection.
3. Economic Viability: Third-party data targeting is generally not cost-effective for lower-priced media but may justify costs in high-value advertising scenarios.
4. Strategic Evaluation: Marketers should critically assess the value of third-party data, considering both accuracy and cost, and remain open to testing its effectiveness within their specific contexts.

Notable Quotes

• Rob Demars: "Buying a third party data set is like paying for a VIP concert ticket and ending up in the parking lot." — [08:54]
• Rob Demars: "Not all third party data is created equal." — [11:46]

---

For more insights and detailed discussions on the latest marketing trends backed by research, tune into future episodes of The Marketing Architects.

Transcript

[Rob Demars]

Nerd Alert. Learning is important, right?

[Leonard Jasper]

Yes, exactly. What a bunch of nerds.

[Rob Demars]

Nerd alert, Right?

[Leonard Jasper]

Marketing Architects. Hello and welcome to the Marketing Architects, a research first podcast dedicated to answering your toughest marketing questions. I'm Leonard Jasper on the marketing team here at Marketing Architects, and I'm joined by my co host, Rob demars, the chief product architect at Misfits and Machines.

[Rob Demars]

Hello.

[Leonard Jasper]

Hello. We're back with your weekly Nerd Alert. Every week I'll take a deep dive into academic marketing research and translate its complex ideas into simple, understandable language for Rob and of course, for all of you. Are you ready to nerd out, Rob?

[Rob Demars]

Elena, my third party data says I'm 87% ready, but my first party data says I'm at a full 100.

[Leonard Jasper]

Oh, nice. Let's take that first party data answer and go forward. Love it. Relevant. Let's get into it. As always, I'll link the research we cover in the episode notes. This week I read a study titled how effective is Third Party Consumer Profiling and Audience Delivery? This is by Nico Newman, Katherine Tucker and Timothy Whitfield, published in 2018. But before I get to the findings, Rob, I wanted to ask you, when is the last time you encountered an ad on the Internet and thought, I don't know if this was meant for me.

[Rob Demars]

Pretty much every time I go on to Amazon for one because I share an account with my wife and two kids in their 20s and I would just say in general all the time, just very frequently they just. I don't know what it is if I'm just some weird anomaly, but it's. There's just all kinds of weird stuff being served my way.

[Leonard Jasper]

Yeah, same. I think my, the one that comes to mind for me recently is I was looking up like pictures of haircuts to bring to my next haircut and for some reason I ended up being retargeted on Amazon for wigs. And I don't know exactly what happened, but yeah, some kind of, I don't know if they. Some kind of data set got shared or what happened or maybe like I, I looked into an image that was from Amazon. Maybe I was thinking that could be.

[Rob Demars]

It as a bald man, I. That one I would actually accept if I was being targeted wix.

[Leonard Jasper]

But I mean it probably make more sense than me at this point in time. All right, well, we've all probably encountered this almost on a daily basis. You see an ad, it just doesn't seem like it was meant for you. And sometimes this is due to a marketer like us using third party data to target our potential customers. But before we get too far into the study, Rob, can you help us out and can you explain the difference between first party and third party data?

[Rob Demars]

Yeah, I'll try. First party data is when you have the information in house. So as an example would be you're going to bake some cookies and you've got this classic family recipe that you pass on from generation to generation and all the ingredients are there because it's your grandma's recipe and you know, it's all good. Third party data would be like buying cookies from a guy in the back of a Toyota Corolla and he's just saying you're thinking you're going to get the chocolate chip cookie, maybe there's going to be some raisins in there, maybe it's going to be delicious or it might give you intestinal issues. Is that pretty good?

[Leonard Jasper]

Yeah, I wasn't expecting a cookie example, but yeah, first party data is data you have on your own customers versus third party data, which is something you're purchasing from and not to be confused.

[Rob Demars]

With the other kind of cookies in marketing. So I suppose that's, yeah, maybe not.

[Leonard Jasper]

The best example, but no, it's great. Thank you. I want to make sure I get our terms straight before we get too far into this because we are going to be talking a lot about third party data segments. The authors wanted to answer what seems like a straightforward question. When we pay to target specific audiences, so people with particular demographics or interests, how often are we actually reaching them? What they did was they ran three large scale field tests across 19 different data brokers. They tested over 90 audience segments and worked with six major DSPs. So this is a really comprehensive look at how well third party targeting actually performs. It wasn't just one data set or one dsp. They looked at a lot of different things. Let's talk about what they found. Study one simulated a typical programmatic campaign. This could be something that your brand is buying digitally. They use demand side platforms to deliver ads targeting men age 25 to 54. Then they use Nielsen's digital ad ratings to verify who actually saw the ads. And the results showed that on average 59% of the impressions reached that correct demographic. So that might sound decent, but if you just served ads randomly, you'd hit that target 26.5% of the time. So yes, it's better than random, but only because the baseline is so low and it came, you're already laughing, it's going to get worse.

[Rob Demars]

That's hilarious. Okay, keep going.

[Leonard Jasper]

That targeting came with a cost, right? Extra targeting doesn't come cheap. They also tracked brand safety and bot traffic. And performance varied significantly across DSP. So it was anywhere from 40% to 72% accuracy. So that's just saying how accurate the DSPs in predicting brand safe inventory. Are they serving them to bots or not? And then in study two, they went further. Instead of running full campaigns, they looked at the raw accuracy of the data segments themselves. Seems important. They compared the broker's demographic labels, age and gender, against self reported verified data from a consumer panel. And when they looked at it this way, the average accuracy for identifying men aged 25 to 54 dropped to just 24.4%. That's worse than random gender. Accuracy alone averaged 42%. Not great. Age to accuracy hovered around 10 to 32%, depending on the range. Crazy. I mean, that's really, really bad.

[Rob Demars]

Wow.

[Leonard Jasper]

So Rob, why do you think it's so hard to get age and gender right when we're using targeting? When we're using third party data to target.

[Rob Demars]

I'm a 51 year old male who likes to watch the Bachelor. I don't watch football, and my favorite food is sushi and White Castle hamburgers. And I listen to Public Enemy and Taylor Swift. So how does that fit into a box? You know, I just think there's probably a level of precision that is trying to be attempted here that isn't really capturing the true uniqueness of the people that they're trying to capture in. Into bucket.

[Leonard Jasper]

Yeah. No. And you also have a very small online footprint. I would think. So it's probably pretty hard, right, to get data on Rob demars, to even see that you're interested in those things. How much of that is obvious online?

[Rob Demars]

Yeah, it would. It would definitely not be obvious.

[Leonard Jasper]

No. So obviously there's a lot of different reasons why this happens. I mean, it can be a technical limitation, privacy challenges, even like incentive misalignment. A lot of this data that you're buying from one party was actually bought from another party and another. Right. And you're probably, you have to kind of track the incentives. Right. You want to give more data versus making sure it's entirely accurate. So whenever you're buying from a third party, it's bound to get a little sketchy. There's one part of the study that really caught my attention and that's when they looked at households with kids, accuracy dropped even more. And you just mentioned, Rob, that you have an Amazon account with multiple people in your family, which is probably explains why this is happening. The researchers think it's because multiple people using the same device confuses the algorithms. So say your kid's watching Bluey on your iPad. Congratulations, you're now a four year old cartoon puppy enthusiast at least maybe from a data broker's perspective. Then in study three, we pivot from demographics to interest based targeting. And finally, we have some slightly better news. When you targeted interests like sports, feedback, fitness or travel, the brokers did a little better. Accuracy climbed to the 70 to 80% range. Still not perfect. But this kind of contextual targeting was noticeably stronger than the demographic attempts. Now this study also gets very practical. They ran a cost benefit analysis and found that third party audience targeting often just doesn't justify the cost unless you're using it for higher priced media. And that's what we found too. If you're already paying a lot on a CPM basis, then that increased premium won't be as much. For something like display, where CPMs are lower, the performance just isn't worth the premium. And I'll also add here that you should be trying to acquire your media for as low a cost as possible without sacrificing quality. So if you're doing that well, I think it's just going to become harder for those kinds of brands to justify the cost of third party targeting in most cases. One of my favorite takeaways from the study was this quote. Third party audiences are often economically unattractive, which is a very academic way of saying this probably isn't worth what you're paying for. It's and the study shows how risky it is to blindly trust your segments. If you're buying men 35 to 44 who like golf, there's a good chance you're paying for mismatched impressions, which could lead to a lot of waste. So if you're a brand manager or a media planner, it might be time to ask yourself, do I actually know who I'm reaching? And is it worth the extra cost to try? Because if your third party audience is less accurate than random guessing, it might be time to rethink your targeting strategy. We'll wrap up with a robgpt. Buying a third party data set is like paying for a VIP concert ticket and ending up in the parking lot. It's expensive, noisy, and you're nowhere near the audience you asked for.

[Rob Demars]

Not as. How did you say that again? Not as good as random. That's how I felt when I took the act. I would have been better just picking it random.

[Leonard Jasper]

You really think so? Yeah. I mean, that's crazy that. I mean age is definitely one of those things. Less accurate than just a random guess again, I think. I know we've talked about this a lot. I think that we have such an addiction to certainty and just we've marketers now, modern marketers have been trained make sure you're reaching your exact target. And so we'll pursue that even when it's taking us down avenues that are leading to the direct opposite of what you're trying to do.

[Rob Demars]

Well, it shows you how. How good of a job. We've branded digital as a channel. Right. Because you just assume accurate. You don't even drill into it. And whenever you say something like third party data, it sounds so smart. Right. Because it has the word data in it. But is the data any good?

[Leonard Jasper]

Exactly. Yeah. And I know we've covered studies like this before on the podcast, but I just think it's obviously still such a giant challenge that anytime we could bring this up. I know Nico Newman has a lot more other other studies. He's kind of the expert on third party data targeting. So I encourage people to go and look into his work if you're interested in it. But definitely something to consider testing, turning off and seeing what happens.

[Rob Demars]

I'd love to hear the other side of this if there is one, because there's enough people out there selling third party data to go. How do you dispute. Because it's such great. It's such a great study that you shared. So what's the counter to it?

[Leonard Jasper]

Well, they mentioned it a little bit in the study, which is if you're buying expensive media, then the cost of targeting on top of the CPM isn't going to be as significant. So again, I think really this all just comes down to cost because I'm fine paying extra to make sure I reach my right target. It's like I'm going to pay more to advertise at an event with marketers in it than I would advertising on a bench near my house. As long as you can confirm the know who you're reaching or at least the cost of it is enough to justify their performance. So I think it probably comes down to performance is a big thing. But again, I think this debate can get sort of personal. But if we can focus on cost, I think that's the best way to make the case for this is if it's just too expensive, it's not going to be worth it, let alone that it's inaccurate. But yeah, I guess some. The case for using it I suppose would be testing and seeing certain. I'm so sure that there's, I mean even the study found there's a big range of how good or poor a third party data set is so you probably want to test it out for your brand. I know for some of our clients when we've tested it we've consistently found that it's not as effective but I do think we have a use case or two where a certain third party data set was helpful for a client. So I think you just need to be skeptical.

[Rob Demars]

So not all third party data is created equal. Just leave it at that.

[Leonard Jasper]

Perfect. That's it for this episode of the Marketing Architects. We'd like to thank Taylor De Los Reyes for producing the show. You can connect with us on LinkedIn and if you like the podcast please leave us a review. Now go forth and build great marketing marketing architects.