Programmatic Media Buying 101: What is the Difference Between AI, Machine Learning & Programmatic?

The world of digital advertising and programmatic advertising has developed its own language in the last couple of years, full of terms that are commonly heard and used everywhere but mean something very specific when attached to the word advertising. Most recently it’s almost impossible to read an article or even talk about media buying without bringing up the terms Artificial Intelligence (AI) or Machine Learning. The terms AI and machine learning are often used interchangeably but they are different. What is the difference between the two and what should they mean to us or me as a marketer or CMO?

AI for Programmatic Buying

Artificial intelligence is the concept of reproducing human intelligence in machines so they can execute on activities that normally would require a human brain to be involved in, such as making data-based decisions.  By using AI-powered systems brands and advertisers case save money and time by completing tasks faster than us mere humans and make less mistakes.  When you apply this to the programmatic media buying industry, you bring efficiency to the media buying process, freeing people who’s job it is buy media from the more tedious and allowing them to focus on the strategic and creative elements of their jobs.

The reason digital media executives keep talking about AI technologies is that they allow us to have algorithms that analyze a user’s behavior, allowing for real time programmatic campaign optimizations towards consumers who are more likely to convert. Advertisers then have the ability to gather all this rich audience data to then use it to be more accurate with their media buys and overall targeting tactics – ultimately spending less money and time and bringing in a higher ROI.

Will Machine Learning Replace Media Buyers?

The words Machine Learning can conjure up images of old sci-fi movies in which someone develops an intelligent robot that then dominates its creator or destroys a large city… leading to many questions about how this technology could affect the digital media industry.
Machine learning is a type of Artificial Intelligence that provides computers or robots with the ability to learn things by being programmed specifically to take certain actions, improving their knowledge over time, much in the same way our brains do.
Computers using machine learning focus on imitating our own decision-making logic by training a machine to use data to learn more about how to perform a task.

Imagine you ride your bike to work every day. Over time, after trying different ways to get to work, you will learn which route is faster or maybe which road or path is better according to the day of the week or based on the weather outside. This is exactly how machine learning works. You feed the computer or algorithm with large amounts of data so it will analyze information from the past and learn from it to apply the learnings to any new data it receives in the future.

When applied to programmatic advertising, machine learning algorithms can analyze large volumes of data from difference sources and draw conclusions from it. It means you can almost replicate the brain of an experienced media buyer in a machine or algorithm so it becomes capable of  predicting, planning and optimizing media. Almost…. but not yet, though the machines can certainly make programmatic advertising more efficient, faster and easier to implement, there remain many factors which need human brains to input link the machine learning to an overall media buying strategy.

So How are AI and Machine Learning Connected to Programmatic Advertising?

Programmatic advertising is the automated process of buying and selling ad inventory through an exchange, connecting advertisers to publishers rather than having to make individual deals with each publisher. This process uses artificial intelligence technologies to improve efficiency and make better decisions for the advertisers with their budgets.
There is a lot of investment being made in marketing and ad buying technologies to leverage AI.  Companies like Xaxis, are betting heavy on AI for improving their future Programmatic Buying Platforms.  Fo right now marketers are using AI to stitch massive amounts of their data together, but it still hasn’t replaced human analysis.  For media agencies, Artificial Intelligence is still more a buzzword or a catchphrase to get peoples attention.

David Lee, programmatic lead at ad agency The Richards Group, said that he regularly gets pitches for AI-enabled products but the AI part of the products usually “doesn’t seem to affect performance outside of being a buzzword.”

You need Machine Learning to feed AI but you don’t need AI for Machine Learning. What that means is that machine learning is the technique — using algorithms to process data, learn from insights and make predictions for future programmatic campaigns which then trains the AI.
Both Machine Learning and AI are here to stay.  If you are a marketer or a media buyer, get familiar with these terms as they will continue to occupy the press and blogs like ours.  But for now they are not taking over for humans, that’s still in the sci-fi section of the video library.

The 7 Most Common Errors Programmatic Media Buyers Should Avoid

Many brands still don’t take advantage of all the possibilities offered by programmatic advertising, preventing them from increasing the profitability of their media buy and maximizing their ROI.

Instead, if brands leveraged the potential of programmatic advertising they could broaden their audience, reaching twice as many unique users, increasing conversions by more than 36%, and reducing CPA vs. traditional online media buying methods.

Analyzing Campaign Errors

Digilant analyzed nearly 500 programmatic advertising campaigns and identified the seven most common mistakes made by media buyers that hinder performance of their campaigns.

  1. Vague or overambitious campaign goals.

    Although digital marketing has become increasingly precise in its targeting, it’s still very common for advertisers to want to cover too many goals or KPIs at once with their programmatic investment. Advertisers should be clear in setting their KPIs to whether for example they are looking to increase brand awareness in a new market, drive online conversions or in-store traffic, or other goals.  That starting point is imperative, the advertiser’s target must be aligned with the most appropriate programmatic tactics, which will ultimately improve campaign performance and ROI.

  2. Failure to segment audience data using programmatic technology.

When provided with large volumes of user data, the possibilities of different types of audience segmentation are endless. There are about 200 individual data points associated with each online user, and by using dynamic programmatic reporting, marketers can create profiles that allow for real-time segmentation and thus increased performance.  To capitalize on this enhanced campaign performance, the audience must be segmented at several levels. With each layer, the objective is to filter and eliminate users that do not fit the target audience for that brand.

  1. Ranking users without considering their value.

By applying machine learning and using data from advertisers and third party data providers, it’s possible to determine the appropriate user profiles for the advertiser to target in real time that are most likely to convert. Skipping this step puts campaigns at risk for failure. After identifying users’ behaviors, predictive algorithms can be applied to determine the value of each profile and user in real time. Knowing the value of the user will allow the audience to be segmented efficiently and effectively, by focusing the campaign on the right users and increasing the investment on users who will be more prone to make a purchase.

After executing a campaign it’s important to reexamine consumer conversion data to optimize the effectiveness of future media buying actions, as brands can exponentially enhance the returns on their programmatic campaigns by knowing more about their user behaviors and attributes.

  1. Delivering the same creatives to customers and leads.

One of the great strengths of programmatic advertising is its predictive ability. It is possible to apply data science algorithms to find potential “new consumers”, not just recycle the same users gained through retargeting.

But it would not make sense to send the same message to the every user. It is necessary to personalize the messages directed to the different profiles that the campaign wants to impact, using technologies like Dynamic Creative Optimization (DCO) to optimize the ad investment. This level of customization is not done as often as it could be for programmatic campaigns, which can negatively impact performance.

  1. Low investment in attribution.

Insights gleaned from programmatic KPI metrics allow marketers to understand campaign performance at a level that is unmatched by other traditional channels such as print advertising or television. The added invested in attribution gives media buyers the opportunity to analyze the results beyond last click, which is a one dimensional view of online marketing and doesn’t allow for full funnel analysis.

Attribution allows you to understand how the media really affect results. For example, actions in the media may be linked to loyalty data or to credit card transactions; So by using attribution technology it is possible to measure the impact of a campaign or a channel on the final conversion of a new customer. In addition, advertisers can also analyze the impact of a campaign on the brand and the perception of users.

  1. Campaign reports are not optimized for future strategies.

Programmatic ad buying provides more metrics, information and data than any other advertising medium. Taking advantage of these real-time stats can help brands and agencies discover ideas that are not always intuitive to them and guide the strategy of their next campaign.

For example, a sportswear retailer may be focussed on targeting a totally male audience. However, a programmatic campaign using intelligence gained through data science could reveal that its highest performing audience is actually in the segment of women aged 25-34.

  1. Using the wrong marketing channels.

There are many ways to reach an audience programmatically — desktop, mobile, apps, video, native advertising, audio and traditional television, for example.

Each channel offers potential advantages and drawbacks that marketers need to carefully weigh when deciding where to allocate their ad spend. If the priority is to take a low-cost action with a quick return on advertising investment, it’s best to invest your budget in display. Video and audio justify the highest CPM if you pursue better brand recognition.

It is also important to keep cross-device segmentation in mind, as the average consumer connects to the Internet through five or more devices daily.

Programmatic ad buying relies on advanced data science solutions to provide marketers with a comprehensive understanding of their respective marketplace and at the same time gives them the tools they need to set out more precise guidelines for optimize advertising campaigns and increasing their ROI. However, many companies still treat their target audience as one large segment, often employing obsolete tactics without analyzing the consumer’s behaviors, interests and attitudes, to find the right segments within that large audience to target.

Advanced segmentation, especially adaptive segmentation allows you to identify the most essential existing audiences for a brand and uncover new key segments. It is as important to spend time with your media buyer to find the right tactics and channels for a programmatic campaign, as it is to learn from the results. The flexibility provided by programmatic advertising allows a continuous optimization during and after a campaign. The analysis and strategy prior, during and after the campaign will ensure that future media buys will have better results for the investment made

Summary 

  • Too many campaigns are executed without having properly analyzed the value of each user, which is essential to effectively segment the audience, thus improving performance: investment should be increased in clients more prone to conversion.
  • The second most costly error: do not apply algorithms or look alike models to find potential “new consumers” by recycling users gained through retargeting. The messages are not targeted to the different profiles that the campaign wants to impact, and the investment is therefore not optimized.
  • Unclear objectives, mistaken marketing channels, inability to identify adequate data layers, poor measurement of objectives and not optimizing the information obtained are other frequent mistakes.
  • Properly using the potential of programmatic advertising allows advertisers to broaden their audience, reaching twice as many unique users, increasing conversions by more than 36%, and reducing CPA versus traditional online methods.
Like what you see? Join the 500+ clients that have partnered with Digilant.