The digital marketing industry has undergone extensive changes over the last decade, especially in the way it handles media buying and selling. The Google Display Network (GDN) has contributed enormously to these changes, as it has become a major player as a demand-side platform, allowing media buyers to execute programmatic advertising. However, GDN also has some huge limitations, such as the amount of publishers and the level of insight it presents into audience segments.
In comparison to the traditional concept of media transmission, where the message from the advertiser was paramount, digital media has brought the audience to the forefront. In digital media, audiences are often grouped into different segments based on certain characteristics. In performance media, data allows advertisers to create these segments and optimise their media buying accordingly. The theory is that the more deeply you are able to segment your users and test and refine your message, the more effective your programmatic advertising campaign will be. The ultimate goal of performance media, and programmatic advertising in general, is to find the exact person who is ready to purchase your product at the exact right time.
Audience segmentation is now serviced by a large number of suppliers. Third-party data is provided by Demand-side Platforms (DSPs), who give access to ad networks with an added layer of technology that can interpret patterns in interaction and buying. Often DSPs connect to tens of ad networks, each of which has their own layer of data and audience segmentation. Figuring out which of these segments is performing best can be difficult, because you must set specific goals and optimise towards segments that are completing these goals. Algorithms are used to optimise towards goals such as click-through rate, impressions, sales, or cost-per-sale. Segments that are completing each of these goals in the highest volume should be amplified with more budget, and those that are not performing should be stripped of budget. This is the core of optimisation.
As an example, lets use a more local scenario to prove the point. Lets consider a local whisky bar holding a live jazz event. The bar wants to find locals that like jazz and whisky. If we were trying to find their audience using performance advertising, we would input into our systems that the event is a music event, and that the primary goal was to sell tickets and whisky on the night. Then, the system uses a mathematical process to create some segments to examine. For example, the system might test people in a certain age category, of a certain gender, and in a certain location. It may also identify some more granular segments to test, but at the start, it is better to cast a wide net with your targeting, so that you can let the data speak for itself. Depending on which fish take the bait, the net will be cast in different areas. Often new and interesting information will be found about an advertiser’s audience, and they will find that their customers have certain characteristics that they hadn’t considered before. Maybe you thought your audience was generally right-wing English speaking men, but with data we find that, in fact, left-wing French speaking women love you.
The great thing about this type of data is that it can help you to form new creative that has these audience segments at its core.
Another way to use audience segmentation in programmatic advertising to increase conversions is to use third-party data to find lookalike audiences. There are a few ways to create lookalike audiences. You could use your first-party data (i.e. the information you already have about your current audience from your Google Analytics or bespoke admin system) or you could create an audience with wide media targeting to find a new audience. A performance media buying system can then use this information to select new users that have similar characteristics (I.e. they ‘look like’ people who have already purchased your product, and so, in theory, should be more likely to buy your product too).
Lookalike audiences work very well for ecommerce, as well as many other types of websites. If we think back to the whiskey bar in the context of online, it is possible to segment those who looked at the marketing material (the ad), those who got to the website of the bar, those who got to the page of the jazz event, and those who purchased a ticket to the event after looking at these pages. The best lookalike audiences are obviously based on those users who purchased your product or converted on your website – the segments are created by looking at the online behaviour of your customers and replicating it.
Creating lookalike audiences is a different process depending on the platform. On some platforms, an initial audience of thousands is required to create a lookalike audience that is a big enough sample size to make assumptions on trends. On other platforms though, you may need as little as a few hundred people, but it is more beneficial in the long run to have the biggest sample size possible.
While remarketing implies marketing again to the same audience, in fact, it has more to do with acquiring customers, as it involves using third-party data. For example, remarketing through Search involves targeting groups based on what they’re searching on Bing, Google and other search engines/portals. Initially, this involves creating a list of any and all keywords that may relate to business’ products, much like what is required in SEM campaigns. Audience groups are divided into segments based on these keywords, and depending on performance, optimised accordingly. This is a popular method of targeting in industries known for being highly competitive in search advertising. It is also an effective way of getting past the expensive bids that flow on from the high levels of competition on a particular search term. Beyond this, it’s also a method of targeting that, in many ways, flies under the radar, as competing businesses are none the wiser about the fact that their users are being targeted post search. This offers another route for marketing to your competition’s audience, and, more importantly, one that doesn’t require you to bid on highly competitive keywords that you mightn’t be able to afford.
Bringing it back to the bar example, a way to use keywords in your targeting is to make assumptions about themes that your target audience may be searching, and in this scenario in might be “jazz”, “whiskey”, “live music” or perhaps “whiskey-tasting”. Using all four of these keyword assumptions in the initial stage of your programmatic advertising campaigns, and with equal budget, will allow you after some time to eliminate the words that aren’t performing and increase your spend into those that are.
In addition to targeting your web traffic, email lists offer an important resource for targeting. At Benchmarketing, we have third-party partners who we work with to gather these email audiences. With access to expansive third-party databases, we can make meaningful comparisons between these lists to build large databases to boost a campaign’s reach, and ultimately its performance. We can access this type of remarketing on the Facebook platform and GDN. One point to keep in mind, is that you should ideally have access to a broad email database to successfully run a programmatic advertising campaign of this nature
Every time someone comes to your site, a cookie records the IP address in a database. With this data you are able target every user connected to the same router. This means you can target on more than one device, such as the smartphone, tablet or computer. Ads can be adjusted according to device, which should improve the results of your programmatic advertising campaign.
In the context of the whiskey bar, first you would use the IP addresses to build an audience of users who had previously visited the website of the whiskey bar. It is possible that anyone else living with the people who visited the site might accompany them to the bar. Soyou might want to push your ads to this broader user network. Another option is to remarket to the IP addresses of visitors who arrived to the shopping cart but didn’t convert. While this audience would be smaller, it may also have a higher chance of converting.
With so many devices and screens, it is inevitable that newer and more discreet targeting techniques will be employed. Programmatic Advertising is only beginning, and with Smart TVs, Smart Watches, Smart Cars with GPS tracking, and so on, programmatic TV and radio will soon become reality. If you are looking to invest into programmatic advertising and need the edge on your retargeting campaigns, get in touch with Benchmarketing on 1300 049 498.