Display marketing is evolving fast with the ascension of programmatic buying. Google contributed to this growth with the Google Display Network, but the platform is still limited in terms of control, targeting opportunities, and the variety of sites available.
In digital media, audience is key. In this context, audience refers to segments of users who are potentially available online to receive our marketing messages. Ultimately, in a performance-based model, we would use all the tools and user data at our disposal in order to show our ads exclusively to those most likely to be interested in them. The more segmentation you do, and the more tactics you test, the better equipped you’ll be to maximize the effectiveness of your campaigns.
By using the five tactics below, you’ll only ever pay to reach users with real potential to convert!
There are currently a large number of third-party data suppliers out there. Some DSPs are connected to more than 30 suppliers, which represents tens of thousands of audience segments at your disposal. The sheer amount of selection can be overwhelming to navigate, and identifying the highest performing segments isn’t always easy. With predictive audiences, an algorithm is used to test different segments based on predetermined objectives. Optimization can be focused on clicks and CTR, or on conversions and CPA. The algorithm identifies the highest performing segments and increases the budget allocated to them, while reducing the spend on segments that perform poorly.
Let’s take the example of a museum launching an exhibition of historical war photographs. The museum wants to reach local visitors with an interest in the exhibit’s theme. So, we tell the system that this is an exhibition, and we’re aiming to maximize online ticket sales. The algorithm will identify a set of segments to test, then adapt spending depending on the performance of each. In the end the results can be surprising, your strongest segment may end up being foodies, or jazz aficionados. You’ll usually see a correlation between your target audience and these segments that aren’t necessarily the direct focus of your campaign, and you can use this new information to adjust your creative and test new directions that might connect better with your audience’s interests.
In the same vein, here’s a second tactic to help you objectively select third-party segments. To use lookalike audiences, first you need to build an audience of your own. It could be your general website audience, or a more specific segment of your users. The system would then compare your selected audience with third-party segments, and identify those with the strongest correlation. From there you can identify lookalike segments, and target users whose online behaviour most closely resembles that of your users.
This type of strategy applies to e-commerce sites, among others. In the case of the museum, you could segment your site’s visitors by identifying the users who browsed your site, who looked at the page describing the exhibition you want to promote, and who actually bought a ticket online after visiting the exhibit page. The ideal is always to build your lookalike audience based on users who generated a conversion – meaning those who bought a ticket. The lookalike tool would then identify user segments whose behaviour is similar to your users’. Then all that’s left is to choose who you want to target.
Lookalike modelling requires different prerequisites depending on the platform you’re using. With Facebook Marketplace, a lookalike segment can be created from a sample of a few hundred, whereas on some RTB platforms, the minimum requirement might be upwards of ten thousand. There’s no question that the bigger the initial audience, the more precise the lookalike model will be.
Although the term “remarketing” might suggest otherwise, this is actually an acquisition tactic, since it’s exclusively based on third-party data. Search remarketing consists of targeting users who use specific keywords in search engines like Google and Yahoo!, as well as portals like MSN, About, WikiHow, etc. The first thing the advertiser has to do is build a list of keywords and keyword groups, as you would for an SEM campaign. Audiences can be segmented based on keyword category, and optimized based on performance. This type of targeting is particularly effective in industries where search marketing is very competitive. It’s an useful way to bypass the inflated bids generated by high competition. And it’s a subtle form of targeting, since competitors have no way to know that you’re targeting their users after a search. Essentially, this means it’s possible to target your competition’s users without necessarily buying their branded keywords and risking a bidding war that might sink you.
In the case of the photo exhibit, you might build categories of keywords related to photography, war, outing ideas, and museums in general. You would launch your campaign with all four keyword categories, then after a few days, you would identify the one or two categories that generate the more interest in your ads. At that point, you’d adjust your budget to maximize conversion.
Beyond remarketing to users who visited your website, you can also target users through their email addresses or mailing addresses. We partner with third-party suppliers for tactics like this. Our suppliers have huge databases that they use to compare with the lists we send them. Although it’s not always possible to identify every user on our lists, we can create a sizable audience, and enhance the performance of a campaign.
Museums usually have a list of newsletter subscribers, to which you could add the emails and addresses of people who might have taken part in a contest the last time they visited. Because not every subscriber will open email from the museum, and you know that contest visitors have already been to the museum, you’ll have created audiences with a strong potential for conversion. You could even create separate audiences from each of the lists, and maximize spending on the one that performs best.
It’s important to note that list-based remarketing is currently available on the Facebook platform, and the functionality should be available on the Google Display Network in the coming year. You generally need a fairly large bank of emails to launch a campaign like this though, since match rates range from 30-50%.
When a user visits your website, their IP address is registered in a database. The lists of your users’ IP addresses are saved, and you can then target all the users connected to the same networks as your website visitors. This is particularly useful for homes with more than one computer, and for cross-device retargeting. You can then adapt your messaging depending on whether a user is connected to the network on a smartphone, tablet, or computer.
For the museum, you would first have to create an audience made up of the IP addresses of the visitors who saw the page for the new exhibition. We can hypothesize that the people living in the same homes as these visitors would most likely either go with them to the exhibit, or at least hear about it. It therefore makes sense to support that behaviour by promoting the ads to the wider network of users. You could also create a retargeting campaign aimed at the IP addresses of users who made it as far as the shopping cart page of the site. This would be a smaller audience than the first, but with greater potential for conversion.
With the explosion of screens all around us, there’s no question that these targeting tactics will continue to grow and evolve. The popularity of Smart TVs, for example, provides us with a perfect opportunity to start adapting TV ads to consumers’ online behaviour. And that’s not all! More and more cars are now connected to the Internet, and drivers’ destinations are often saved in their GPS. So why not adapt radio ads to take into account the driver’s destination? Even when we travel on foot, smartphones and smart watches follow our movements and save our actions. In short, the era of Big Data is only getting started, and marketers who jump into the fray early on will have a significant advantage in the years to come.