Digital Advertising Needs an AI Paradigm Shift
If there is any aspect of marketing that can benefit from machine learning, it’s digital advertising. Digital advertising relies on wide swaths of online networks, retargeting empowered by cookies, and “precision” spends based on keyword performance and tightly-honed customer profiles.
When an ad achieves double-digit percentage rate of increased engagement, click-throughs or sales, brands celebrate a tremendous victory. Meanwhile, customers groan at the relentlessly targeted spam.
As each year of the Internet passes, the online medium becomes more intrusive. Online consumers are subjected to forced advertisement plays on YouTube, unrequested interstitials and offers for additional content and experiences, and pay-to-avoid advertising firewalls. The vast majority breeze by the ads, some not even seen, others willfully ignored. The overall click-through rate online is just 0.05 percent, according to SmartInsights.
For all of its impact on the brand and bottom line, advertising annoys customers. Most view it as spam.
The advent of data has made some personalized advertising advances possible, but brands seem to squander the opportunities. How many times have you bought a product, and then found your online experience was littered with advertisements from the brand? Even worse is when those advertisements are for the same product you just bought, a complete customer experience fail.
Brand advertisers need the promises of AI. They need the ability to use predictive click-throughs based on behavior, demographics, and optimized content. Perhaps even better will be the optimized expenditures on media. Wasted dollars previously spent on spray and pray networks to loosely defined audiences will get optimized by data infused machine learning.
The Big Two: Google and Facebook
It should come to no surprise that Google and Facebook are at the forefront of advertising AI. Both brands are massive companies that rely on advertising revenues. Increasing customer successes leads to more media dollars spent.
Google Ad Words first widely deployed AI recommends keywords to advertise against, ad headlines, descriptions, and extensions. The intent is to help advertisers build out and scale their campaigns in a more rapid fashion.
However, as much as Google’s AI is intended to help, it lacks context and needs guidance. Unfortunately, if your advertising manager isn’t watching Ad Words closely, Google will implement its AI recommendations as part of a brand’s user agreement.
Facebook advertising AI will let you know when a post is performing well and should be boosted, help you rank audience targets, and score your ads. Algorithm-based tweaks to Facebook’s advertising manager also flag political ads, over-laden text ads that won’t perform as well as others, and other forms of offensive content. In many cases, the attempts to catch bad actors have been too aggressive causing Facebook to receive complaints from its advertisers.
Like most marketers, I am suspicious of Google and Facebook’s AI attempts. After all, they intend to get you to spend more. And sometimes they do just that without delivering the brand meaningful results. That’s where AI-driven third party programmatic bidding comes in.
9 Million Bids Per Second
Imagine handling 9 million online advertising bid queries per second. It’s too much for even the best programmatic processes to manage in a completely optimized fashion. That’s why The Trade Desk built Koa, an artificial intelligence engine that utilizes real-time data analysis to make advertising recommendations.
“We look at 9 million queries per second,” explained Tim Sims, Senior Vice President, Inventory Partnerships, The Trade Desk in an interview with eMarketer. “If you put that in the context of other mature marketplaces like Nasdaq or the New York Stock Exchange, it’s astronomical. The sheer size of that bidding creates opportunities… The other opportunity we have is utilizing that data to create more AI decisions, where the platform takes control to find the best execution path for a particular campaign.”
It makes sense. Online advertising bids rely on so much data to make the right offer. There are incredible sources of data that vary from time of day, the day of the week, seasons, weather, customer profiles, incentives, campaigns and more. That includes bidding on the right page, not just lowering costs.
Auction AI will help advertisers keep up with changing policies and market fluctuations in real time, lowering research costs and by optimizing bids based on win probability. Further, AI can update the massive inbound flux of advertising data in a structured manner that ensures clean analysis.
IBM Watson has a Cognitive Bidder technology that analyzes factors like weather, the Internet of Things, and analysis of web pages to make better bidding decisions. Insights are used to fuel optimized bidding strategies, and inform buyers about messages that have a higher probability of producing outcomes.
We have discussed how Lucy is optimizing targeting in a prior article. Some brands are using that optimization to not only refine messaging but also utilizing their spend. BMW has seen a 10% increase in optimization by using Lucy not only to recommend who to target, but which social networks to spend their money.
In the advertising space, real progress is also coming from niche-specific third parties that are creating significant advantages in online media bid tools, targeting, and content optimization.
Advertising Job Impacts
Does that mean AI will replace humans on the advertising side? No. A human is absolutely necessary to oversee the granular recommendations made by the AI.
In fact, if you consider how machine learning works, humans are needed to guide the AI and help it hone its performance against data-rich tasks. It’s unreasonable to expect the AI to get bids right on the first attempt. What it can do is show correlations and possible answers.
Humans need to train the algorithm on how to select bids. Even after a programmatic bidding AI starts performing well, human guidance is required on changing budgets, new programs, and incentives, and creating and maintaining the guidelines rules for the AI to operate in.
What advertising managers should look for are increased efficiencies and stronger buys. For example, consider how much quicker an algorithmic application could tag and organize inventor once it is trained. Further, if the page content is tagged appropriately, the AI can find inventory that places ads against the right content or images in real time. Then it can flag new opportunities.
How powerful can programmatic AI become? One marketing AI platform Albert claims a 30x increase in leads, 50% lift in ROI and a 600% increase in conversions.
“Generali Global Assistance has a small digital marketing team,” said Tiffany Glass, Head of E-commerce at Generali Global Assistance in an Albert Case Study. “Albert improved our efficiency in executing and measuring campaigns across multi-channels via AI. It would have been difficult to achieve the volume of campaigns in the diversity of digital channels in such a short period of time without Albert’s optimization algorithm on our side.”
The possibilities are tantalizing.
Of course, like every other form of AI hyped promise versus reality, marketers should assume a steep curve as a machine learning algorithm comes to understand your business, brand, and department’s preferences. A period of training should be expected, and further DO NOT fire your advertising manager. Oversight is needed during the early phases of any marketing AI implementation.