Trust is the foundation of all relationships, particularly in media. If you don’t trust your agency, you shouldn’t be using them — especially related to advertising. There’s a lot of talk in the marketing and media domain at the moment regarding transparency related to media buying. The infamous ANA report on media transparency in the ad industry has done its part in stirring up this debate, which has been brewing under the surface for decades.
When I say decades, I really do mean decades. The fundamental business model of media agencies and consequently its issue with transparency has been around since the 1960s. This model has been (and to some degree still is) a brilliant, well-lubed money-making machine that guarantees revenue.
The whole idea is that advertisers want to run campaigns to promote and distribute their offers and messages, but lack the knowledge of how to actually purchase the insertions in the different media types. Enter the media agencies that are experts in negotiating, buying, planning, and tracking media across all media channels. This setup was beneficial for the advertisers — they didn’t need to learn all the tricky, messy, and complicated ways of buying media, but could instead focus on the messaging and how much money to spend. A clear case for outsourcing, right? Maybe not, since the media agencies were the ones developing the relationships with the media, and the advertisers were wedged out and didn’t talk to the actual media where their ads ran. That meant that they had no means of knowing the actual media prices for each single insertion. I’m sure you can realize how many possibilities this gives the “man-in-the-middle” for business optimization.
This is the core problem with the business: lack of transparency. You as an advertiser drops a bag of money into a bucket and you are later informed that this money has been distributed to run your ads on different media. But you don’t really know how much you paid for what, and worse yet, you have no idea if there was any effect on your business KPIs at all! But transparency is not the problem — it’s a symptom of lack of insights into the effect of advertisement.
The age of AI
We have entered the age of artificial intelligence (AI). With massive amounts of data, measurements, and algorithms to support its growth, AI is now finally in a stage where it can solve real-world specific problems much better than humans can across many domains. Media buying is no exception.
Is media buying really that hard? Yes, it is. Consider the following scenario. You are given $10 million to invest into media. How should you distribute this? You have your options of major channels: TV, radio, cinema, display, out-of-home, social, search, newspaper, magazines, etc. Within these groups, there are specific media choices to make, including network, timing, pressure, format, buy-type, geographical location, and audience target. If we limit the decisions needed to TV alone, then, naively calculated, you will arrive at more than 6,000 possible ways you can allocate your money for a given campaign. The 6,000 comes from approximately 25 networks, 4 day parts, 10 possible weeks, 3 different spot lengths, and 2 different creative executions.
Imagine what happens when we throw in the other media channels. There are tools within each specific media channel that handle this today to optimize media metrics like GRP, affinity, or number of impressions. However, it does not solve the advertiser’s transparency problem, since you still don’t know how it will affect your business KPIs. This can be solved by letting AI learn from the past what media combinations affect you and by how much. This AI can then distribute your money towards the most effective combinations every time you run a campaign and consequently directly optimize your business KPIs. Better yet, it will use the new data that the future campaign generates to learn even more than it did before.
How AI learns and acts
AI learns differently from humans. It represents the world as a set of parameters and interactions between parameters and data. One of these parameters could, for example, represent the media elasticity of a 30-second TV spot on NBC with respect to store transactions. These parameters are then updated every time new data arrives, allowing the AI to learn continuously. There’s an ongoing debate on whether these parameters are stochastic, but the basic principle of AI is still the same.
The process for a marketing AI looks like this:
- The AI is born, given prior knowledge about the world. This knowledge can come from human media planners and/or previous studies.
- The AI is fed a batch of historic data to update its priors into posterior knowledge.
- The AI is now ready to use for planning, execution, and retrospective reporting of effectiveness.
- The AI plans a campaign, which gets booked and executed
- The AI receives new data each day, updates its posterior knowledge, and reports findings of effectiveness.
- The AI returns to step 3.
The road advertisers need to take
The reason for introducing AI into the mix is that it can provide us with an unbiased view on how media is actually affecting our business. It will report what it has learned from data and utilize that knowledge every time there’s a new campaign to plan.
This of course is easier said than done. A successful AI needs a continuous feed of data and the ability to communicate its findings with the right “people” (which in some media are actually computers in the form of booking systems).
However, the reward for the CMO and the rest of the business is a continuously learning entity that always allocates the money to the right media, providing the most bang for the buck and thus allowing decision makers to make accurate, fast, and optimal decisions with confidence.
AI can transform the way media agencies buy digital media for their clients — but only if marketers are willing to make the change.