AI for media buying isn’t necessarily a far-off dream
Artificial intelligence is now a concept that we interact with almost everywhere we go. It’s the technology that drives autonomous cars, suggests what we should buy and automates our homes. So why is it taking so long for media businesses to embrace this concept?
Digital media campaign execution still isn’t easy. Marketers are faced with never-ending technology options for their marketing programs and messages, all of which need to be bolted together somehow. And for any marketer considering the DIY route, there are some important considerations to note before diving in. The reality is that the tech stacks required for truly personalized messaging aren’t off-the-shelf solutions.
Just imagine the setup requirements, technical coordination and processing power required to dynamically construct an ad, tailor it to an individual’s preferences, determine how much to bid on it and then serve the ad across multiple devices. In today’s real-time-bidding ecosystem, a marketer has less than 20 milliseconds—per ad—to do all of those things. It’s not easy, and bringing in AI to aid that decision-making process makes this even more complex.
AI tools exist that allow marketers to analyze historical performance of campaigns and make recommendations for the future. These engines can potentially determine the most valuable people for a given campaign, but that also raises some questions and issues:
- How much time does that process actually take?
- How much does the engine know about those people and are the recommendations statistically sound?
- How does the engine know if these valuable people can be found again, and how hard will it be to find them?
- What’s the opportunity cost of doing a post-campaign analysis and then having the data and recommendations go stale by the time you’re ready to launch your next campaign?
The solve for these problems is real-time optimization, and we’ve been doing this for years. Our CORE platform crunches more than 7,000 variables across our 200 million unique profiles to determine an individual’s likelihood to take an action, whether that action would be an in-store sale, loyalty program sign-up or online purchase. In addition to all of the typical ad serving necessities, our ad stack provides unique decisions on more than 150 billion messaging opportunities per day—all in less than 14 milliseconds each.
Our SVP of decision sciences recently commented in the Wall Street Journal, where he noted that we can dig into our logs and understand exactly what variables were considered in every decision and why each decision was made. Thankfully, though, we don’t need to do that—the machines do that for us (with our oversight, of course) and then further optimize those decisions each time they’re made—which is the beauty of AI and machine learning.
The sheer number of considerations our platform makes is simply mind-boggling. When our clients share their data with us, we can make correlations between client site behavior, purchase patterns, historical campaign messaging activity and many other variables to determine a person’s propensity to take an action.
And here’s the kicker: we re-analyze all of those variables every time we have an opportunity to message someone. This is real-time optimization at its finest, and our algorithms have led to an average 10x incremental return on clients’ ad spend, all while ensuring a privacy-compliant, brand-safe and fraud-free ad environment. Smart decisioning also helps ensure that the conversation with the consumer remains personal and purposeful, preventing irrelevant or wasteful messaging to people that have already taken the desired action.
When considering using AI and real-time decisioning within marketing programs, there are a number of factors to consider:
- Know your own data. What types of customer data do you have on-hand? How often is that refreshed? How are all of those data points connected to an individual (both current customers and prospects)? How “clean” is your data?
- Build in complimentary data. Which data sets should be prioritized for incorporation into decisioning models? What other data assets will compliment your first-party information?
- Understand how it all works together. Can you construct a feedback loop between business outcomes and decisioning models? Can it be done in real-time or is there a time lag?
- Know what you need to measure and how. How will the overall impact of decisioning/AI technology, both on campaign performance as well as bottom-line business results, be measured?
Persistent identity resolution and data hygiene are needed for AI to work at its best, and then the data needs to be stored in the right environment for analysis. This requires coordination between all stakeholders across marketing, IT, finance and others.
Conversant has been ahead of the advertising AI curve for quite some time now, but I do believe we’re only at the beginning of understanding what’s possible. Technology will only improve; algorithms will only get better; and our knowledge of consumer behavior will only be more enhanced by these advancements.
Both the marketer and the consumer continue to win as AI better understands peoples’ needs and desires. Consumers get more tailored and timely messages while marketers minimize wasted ad spend and drive the best possible outcomes for their businesses. Having the ability to optimize in real-time makes this happen.