You'll learn how to

Supermetrics check mark icon green

What contextual advertising is

Supermetrics check mark icon green

How it's different from behavioral advertising

Supermetrics check mark icon green

Some examples of contextual advertising

Supermetrics check mark icon green

How to create, run, and measure contextual advertising

Subscribe to the Marketing Analytics Show podcast

Learn everything you need to know about marketing analytics.

Super headphones

Transcript

Anna Shutko:
Hello, Ned, and welcome to the show.

Ned Dimitrov:
Hi, Anna. Thanks for having me.

Anna Shutko:
I’m very, very happy to have you. Today, we have a very, very interesting topic. It’s going to be all about contextual advertising. So my first question to you, Ned, is what is contextual advertising? Can you please explain it in a couple of words, and maybe you could also tell us more about w contextual advertising can benefit marketers, especially today?

Ned Dimitrov:
Yeah, I think in explaining contextual advertising, it’s best to explain the two main forms of digital advertising that are available today. So, digital advertising is basically split into two things. One is behavioral advertising, and the other is contextual advertising. Let me explain the difference between the two.

Behavioral advertising means an ad is shown because of information about the user. The user has done something in the past, for example, purchased some shoes, and then an athletic shoe company might show an ad to that person because they’ve purchased shoes in the past. So that’s behavioral advertising. It’s all based on past actions of the user, and the ad is showing up because of something the user did in the past.

Contextual advertising is different, on the other hand, because it doesn’t use any user information. It just uses information about the article next to which the advertisement is showing. So regardless of what user it is, if the webpage is a page about running shoes, it will show the athletic shoe advertisement. So regardless of whether the users read that page or purchased shoes in the past or not, we’re just showing the ad because of the content of the page. So that’s where the name contextual comes from because the ad is shown because of the context in which it appears.

Anna Shutko:
All right. That sounds awesome. And I really like how you outlined the differences between behavioral and contextual advertising. And especially now when marketers cannot rely on cookies anymore, this form of advertising is going to become more and more and more relevant. So let’s talk about the structure of these kinds of campaigns. And again, maybe you can mention how they’re different from behavioral ad campaigns. So how can you structure an effective contextual advertising campaign?

Ned Dimitrov:
So, I think in terms of talking about structuring a contextual campaign, we have to talk about the different ways in which contextual shows up in modern DSPs. So contextual advertising is actually very old. It predates online advertising entirely. So before digital advertising, the way that contextual advertising would show up is you would have a magazine about a particular topic like Runner’s World, and the advertisement for the athletic shoes would show up in Runner’s World magazine.

Again, this is a type of contextual advertising where we’re showing the ad because the content of the magazine has to do with athletics and with running. And when digital advertising came about, there’s also different that contextual advertising basically moved along to the digital realm where an advertiser would sort of purchase an entire website or a section of a website, like you might sort of purchase the lifestyle or athletic section of The New York Times to show your running shoe ad.

But that’s kind of a legacy way of running contextual advertising. Some more evolved ways of running contextual advertising, which is still legacy, is with keyword phrases or key phrases. So in this way of running contextual advertising, the advertiser basically selects a list of 100 or 400 key phrases like running, running shoes, athletics, and so forth. They might write it out in a little text box in the DSP. And the DSP looks for direct matches between the webpage text and those 400 phrases that the advertiser wrote. And if there’s a direct match, the ad will show up on that page. And there are various versions of this, but this is essentially key phrase matching.

The third legacy way of doing contextual advertising is with IAB categories. So here, the publisher places an article inside of a category. So The New York Times might label each of its articles like this article is about athletics. This article is about automobiles. And then, in the DSP, the advertiser would select a contextual category as defined by the IAB, and then all articles that publishers label with that contextual category would receive the advertisement.

So all of these are legacy ways that don’t work quite as well as some more modern ways of doing contextual that are AI-based. So let me explain why these legacy ways don’t work that well. So first, key phrase advertising puts a lot of weight on the advertiser. A lot of heavy research needs to be done to collect those list of 400 words, and then it might not scale well. So there might not be enough pages out there that match your list of 400 words to actually show and spend your advertising budget.

Hierarchical or IAB categories type of contextual advertising scales well, but it has disincentives because publishers tend to put their articles in many categories. And so your ad might show up on a page that’s not quite exactly the context that you wanted it to be in because the publisher placed that article in multiple categories. And it also doesn’t have the ability to sort of doing highly niche contextual advertising.

A niche contextual advertising might be something like spray paint products. The IAB categories do not go into such detail that they would have a category for spray paint products where publishers are putting their articles in that category. It would be just general categories like athletics or auto.

Some of the newer ways of doing contextual advertising are AI-based. Let me contrast these with the legacy ways. With an AI-based technique of doing contextual advertising, the advertiser would place in just a few words. You might place in athletic running shoes, running shoes, and soles, maybe because you’re marketing the soles of your shoes. And so it would take just those three or four words, and the AI based on those three or four words would compute a whole bunch of relevant related words to that.

This is basically the AI doing the research for you, so you don’t have to place those 400 words. And all of those relevant related words they’re not equally relevant. Some are going to be more relevant than others, and so forth. And this allows the AI to essentially grade each page to say, how in context is this page? How much does it match that running shoe context that the advertiser wants? And the ad would get placed on the most in-context pages, the ones that match the context of the advertiser the most.

The advantage of these AI methods is that one, you don’t have to limit the advertising to a specific publisher that you have a specific contract. Two, you don’t have to do all of the research to find those 400 words for key phrase advertising. And three, you can have an arbitrarily niche context that the advertiser defines that fits that advertiser’s particular product or brand message.

And so, going back to your question in terms of how to structure contextual advertising today, for a modern contextual campaign, I would strongly suggest that advertisers use one of these AI-based methods of creating contextual campaigns where the AI does the heavy lifting in terms of determining whether an article is within the context or not within the context of your desired context in which to appear your ad.

Anna Shutko:
All right. Excellent, that makes a lot of sense. And thank you so much for sharing. This is actually super, super interesting to hear about all the categorization of the less relevant methods. And it makes a lot of sense that AI-based methods are now more and more effective for the marketers. And they’re working more and more effectively.

Now, let’s talk about this a little bit more. So you’ve mentioned that marketers just have to select a handful of keywords, and the AI would do the optimization for them. Can you also give any kind of recommendations on what else marketers should prepare in order to make the most out of this AI-based method? Should they consider something regarding their messaging, maybe creative, maybe the overall campaign structure, any tips regarding these?

Ned Dimitrov:
Yeah, definitely. So to give tips on structuring a contextual campaign and messaging around it, we have to go back to the difference between behavioral and contextual. So the main advantage or one thing that contextual has that behavioral does not, is that the brand message is hitting the user at the time that the user is consuming a particular type of content online.

The user’s frame of mind is in a particular frame of mind because they’re consuming that content currently. And because of that, the brand message can be highly effective in reaching the user’s mind. And so you can create your messaging to sort of fit the context in which your ads are appearing in. For example, you might be marketing your running shoes to elite runners, and then you could fit messaging that has to do with how this helps increase their distance or their stamina and so forth. Or you might be marketing your shoes to sort of a cross-training type of crowd, which is a different type of context. In which case, you could have a different messaging about how well your shoes work for sort of side support of stepping and so forth.

And so, the brand message can basically be highly tuned to the context in which the ad appears. And this makes the brand messages much, much more effective than in a sense, the standard behavioral advertising, because the behavioral advertising might appear in any context. But in contextual, the context is fixed, and so the brand message can play off of that context.

The second thing that I would say that advertisers should think about when they’re creating their contextual campaign is the frame of mind that they want the user to be in. So what do you want the user to be thinking about when they see your brand message? And sort of the same way that in behavioral advertising, there are different audiences that you could try targeting with your brand message.

I would place sort of similarity to that with different mindsets that you want your message to appear next to in contextual advertising. So try to brainstorm maybe three or four different mindsets that you want the user to be in when they see your brand message. And each of those mindsets would sort of translate into a different contextual campaign. So not all contextual campaigns are created equal. Different mindsets are going to have different effects in terms of receiving your brand message. And so try to brainstorm those different mindsets that you want the user to be in.

Anna Shutko:
All right, these are excellent tips. Thank you so much for sharing. And now, let’s talk about one very interesting thing you’ve done, namely, the experiments you’ve run. So you’ve mentioned that you have compared data based on different access. So you have run the vertical-based analysis, how contextual ads perform depending on the vertical, how they perform depend on different strategies compared to more traditional methods such as retargeting ads.
So can you please tell me more about these experiments and how marketers can test the effectiveness of a contextual advertising campaign? Maybe you have a couple of tips regarding that.

Ned Dimitrov:
Yeah, absolutely. I can talk too about the performance of contextual advertising campaigns. So I’d like to put this in a little bit of context, so to speak. So those legacy methods of doing contextual that I mentioned, the key phrase targeting and the IAB categories, have been around for decades. And historically, they’ve seen relatively low performance.

And that’s why advertisers when they think about creating a media plan, might not include contextual in their media plan because they think of contextual as these legacy methods, the non-AI-based methods of specifying context. And historically, they haven’t performed that well. But at StackAdapt, we ran a very large study including thousands of campaign comparisons that actually ran live on the internet to compare the performance of new AI-based contextual methods to more standard behavioral-based methods.

So in terms of creating these comparisons, one comparison that we did was compare the AI-based contextual to retargeting. And the reason that we picked retargeting is because for most advertisers, they’ll recognize retargeting as sort of the gold standard in terms of conversion performance. Retargeting often delivers the best CPAs, the best cost per acquisition for a marketer, because you’re delivering the ads to users who already know about the brand. That’s why we are retargeting those users because they already know about the brand. And so retargeting often delivers the best CPAs.

Now, in these thousands of comparisons between a retargeting campaign and a contextual campaign run by the same brand with the same conversion point and so forth, about 22% of the time, the contextual campaign outperformed the retargeting campaign in terms of CPA. So, let me put that into perspective. About one in five times, the AI contextual campaign did better than the retargeting campaign in terms of CPA.

I think these are fantastic results, especially considering that retargeting is the gold standard in terms of performance, in terms of conversions. And so just that statistic alone should make most advertisers think of including at the very least trying an AI-based contextual campaign in their marketing plan to see if they happen to be one of those five people for which contextual will outperform retargeting in terms of conversions.

The second thing that I would say that you alluded to, Anna, is that not all verticals are created the same. So for some verticals, contextual works a lot better than others, and there’s many reasons for that. So let’s just take a look at one vertical, which in our study showed excellent performance for contextual advertising, which is the healthcare vertical. The healthcare vertical comparing retargeting to AI contextual campaigns, we saw that AI contextual campaigns outperform retargeting more than 50% of the time. In more than one in two comparisons, AI contextual did better than retargeting for healthcare clients.

And so that’s a fantastic thing to consider in terms of that vertical and contextual. But another thing to consider in terms of vertical and contextual is that behavioral targeting in healthcare oftentimes has a lot of ethical implications to it. You don’t want to sort of following a user around the internet with a certain medical advertisement because you’ve identified them as having a medical condition. That’s generally not ethical, and it’s not something that an advertiser would really want to do to sort of gain audience.

However, when you’re doing a contextual campaign in that vertical, you’re targeting just the sites that discuss those medical conditions using no user information at all. And so, while the user is reading that site, discussing the medical condition, they see your brand message next to it. And when they’re off that site, it’s gone. There’s no tracking. There’s no user information involved.

And so, there are multiple reasons why contextual advertising and AI-based contextual advertising work particularly well for the healthcare vertical. But the same is true for other verticals. There are certain arts verticals where contextual advertising performs very well for similar reasons that the user’s mind is particularly open to the brand message when they see the advertisement.

Anna Shutko:
These are some incredible insights you shared, especially the healthcare vertical experiment you were talking about. So I think a lot of marketers will learn a lot from this answer. Thank you so much for sharing.
:
And now, let’s talk about the tests may be a little bit more and then the reports marketers have to create in order to see how their contextual advertising performs. So, first of all, what kind of reports have you created to test the performance of these contextual ads you were talking about? And another question here is what kind of metrics should marketers pay attention to when they’re evaluating the results of their campaigns?

Ned Dimitrov:
That’s a fantastic question, Anna. And it brings up a point that I want to discuss, which is what contextual is good and what contextual is not good. So as many of the listeners know, just different advertisers have different KPIs that they’re looking at when they look at advertising campaigns. Some might look at clicks or CPCs. Others might look at click conversions. Others might look at the view-through conversions. There are various attribution models for conversions to consider in terms of evaluating the effectiveness of an advertising campaign.

So, let me just discuss a little bit about the things that context is not good. Contextual in AI-based contextual advertising is not good at clicks and collecting clicks. So you’ll see significantly lower click-through rates and significantly higher CPCs than you will in behavioral advertising. And there’s a very good reason for this. The reason for this is that the ad is shown to the user when the user is consuming some sort of great content.

The reason that the ad is there is that there’s great content next to it about a specific topic, and the user is consuming that content right now. Because they’re consuming that great content, they do not want at this moment in time to go and click on that ad, and go and convert. And so, when you look at the metrics and the reports that come out of contextual advertising, advertisers should be prepared for relatively low click-based metrics. So low CTRs, high CPCs, and so forth.

However, what contextual advertising and AI contextual advertising is great in as I’ve alluded to before, is getting the brand message into the user’s mind. They’re consuming great content, but their mindset is in a particular mindset when they see your advertisement. And because of that, your brand message has the ability to enter their mind much easier than it would if they saw that advertisement in sort of an arbitrary different place on the internet because they have a particular mindset.

And so, the metric that AI contextual advertising is fantastic in is view-through conversions. And what that metric measure is basically the ability of the advertisement and the brand message to make it into the user’s mind. And then, the user will go and perform the conversion action on their own later. It doesn’t have to be an immediate click when seeing the ad. They can type in the domain. They can search for the domain and the brand name on their own afterward to perform the conversion action. And we’ve seen this across multiple advertisers across multiple verticals.

So that’s how it would break down in terms of the reports that you would expect contextual to be not good in, and you would expect the contextual to be good in. So I would strongly encourage listeners to look at view-through conversions for AI contextual and compare that to view-through conversions for things like retargeting various audiences that they might get from DMPs and so forth.

Anna Shutko:
Okay, fantastic. Thank you so much for such a detailed answer. I think it’s very, very clear for our listeners what they have to pay attention to. So once again, it’s great that you have identified these differences.

And probably one of my last questions would be what are the typical mistakes marketers make when they start interpreting the results of their contextual ad campaign? So you mentioned the data they should be paying attention to. But how should they approach experimenting, maybe analyzing data depending on different audiences, depending on different verticals? And especially if we compare contextual advertising to something like retargeting, are there any other core differences they should note when they’re comparing to different data sets?

Ned Dimitrov:
Yes. I think a common mistake that I see, especially with AI-based contextual, is that advertisers often think very carefully about those five words that they put in the context. And they’ll hyper optimize like, should it be this word? Should it be that word? Should it be a plural? Should it be not a plural? And in general, that’s a too low level of an optimization to be effective in your advertising campaigns.

In terms of optimizing advertising campaigns, what I would suggest paying attention to is the mindset that you want the user to be in when they see your brand message. And so that goes back to my point a couple of questions ago where I said the advertiser should brainstorm maybe three or four different mindsets that they want the user to be in, which correspond to three or four different contexts that the user to be in.

And so those different contexts wouldn’t be entirely different sets of words that you type into the AI contextual to specify the context, as opposed to micro-optimizing each of the sets of words within the AI contextual. So the AI contextual is fantastic at taking those small sets of words and extrapolating them into all their different variants in all their related words and so forth. And that’s why it’s made so that you don’t have to hyper-optimize the individual words inside of the context that you specify.

The other sort of common mistake that advertisers make is they expect sort of a key phrase or word level report when it comes to AI contextual. So AI contextual is not a key phrase matching engine. It doesn’t work like those legacy methods where there’s just a key phrase, and if the key phrase is on the page, your ad appears there. If it is one of those legacy methods, it is possible to expect a key phrase level report, which shows you for this key phrase, there are this many impressions, this many clicks, and so forth.

But AI contextual fundamentally doesn’t work this way. It takes your words. And from that, it uses various AI techniques to extrapolate to many related words and many related topics online and places your ad on the most in-topic, in-context pages that appear online. And as such, it simply doesn’t have the possibility of presenting a key phrase level report.

And so that’s a common mistake that I also see is approaching it from the key phrase level and thinking that the basic optimization is the key phrase level, whereas the basic optimization should be the mindset level. What mindset should the user be in when they consume the brand message?

Anna Shutko:
Great, fantastic tips. And thank you so much for sharing. Now, I think we are coming to the end of the interview. Ned, thank you so much for coming to the show. And if the audience would love to learn more about you, and I’m pretty sure they will because you’ve shared so many useful tips, where can they find you?

Ned Dimitrov:
The audience can find out more about me and some of the AI-based contextual things that StackAdapt has been doing at StackAdapt.com. That’s the DSP for which I worked, and that’s where we did all of the experiments comparing AI contextual to more standard behavioral techniques.

Anna Shutko:
Fantastic. And once again, thank you so much for coming to the show.

Ned Dimitrov:
Thank you very much for having me.

Try Supermetrics for free

Get full access to Supermetrics with a 14-day free trial.
No credit card required.