Focus on these four areas to improve your marketing using AI

Focus on these four areas to improve your marketing using AI Brian Baumgart is the co-founder/CEO of Conversion Logic and has over 19 years of experience in advertising technology. Much has been written about the significant impact that artificial intelligence is having across the business spectrum, and marketing is no exception. In this article, Brian points out the important distinction that marketing is a skill, while AI is a tool. He outlines the parameters and limitations of current AI technology, and offer tips and tools to marketing professionals on how to effectively balance AI with their human judgement to achieve true marketing success.

When it comes to marketing, the goal of this game is not to win; it's to keep playing and to leverage data to improve your play over time. 

Great marketers use new tools to enhance their ability to do this efficiently and effectively, and the world is changing faster and faster. No matter how much technology evolves, marketing will always have a human element. Artificial intelligence (AI) might seem like a silver bullet or an elusive capability, but if you're trying to influence people, you have to understand people. The day AI can understand people better than people, there will be nothing left for us to do.

Right now, only 15 percent of organisations are using AI, but 31 percent plan on utilising it more over the next 12 months. And of the companies already taking advantage of AI, 28 percent say they use it for marketing purposes.

Looking into the future

Channels, publishers, formats, tactics, and promotions are expanding, and marketers are taking notice. In fact, 61 percent of marketers in companies of all sizes say that AI will be their biggest data initiative for the next year. So in order to be able to analyse data effectively, we need more powerful data tools.

Over the past 10 or 15 years, AI has exploded in terms of capability. Initially, it was finding paths, faces, and other objects in photos without any supervision. Then, it was able to turn speech into text. Now, there are self-driving cars using AI.

AI is and will be useful in situations with well-defined structures, outcomes, and parameters. So driving is a high-variability exercise, but it's extremely structured. You've got roads, you've got laws, you've got rules, and you've got a clear objective: Get from point A to point B safely.

But things like strategy are not very well-defined. It's not well-structured, and it has a huge human component. So the best use for AI in marketing is identifying patterns and data that are too large for human cognition — because it would take humans a year to pore over a hundred million rows of data ourselves.

Critically, what AI can't do right now is understand the results in context. Most marketers have relationships with different vendors, different agencies, different budget constraints, and different contracts. While AI is moving in the direction of generally making recommendations — taking a distribution of possibilities and simulating possible recommendations for marketers — having AI try to make decisions in context is just not reasonable yet. Empower the human to do the human thing.

Decision-making and artificial intelligence

Ultimately, marketing comes down to decision-making. You've got to make decisions about how to strategise, where to spend, how to leverage data and technology, where to dedicate resources, how to adapt, and how to contribute to revenue.

At the core of this decision-making is the idea of looking backward and forward. So looking backward is the analysis part: What did I do? What happened? What can I learn from? How can I grow? How can I adapt? How can I get better at this skill that is marketing?

On the forward end is action, when you have to start making decisions. Spend more money here and less money there. Improve or remove the creative. If there's a loop between these two, your decision-making and your skill actually improve.

In marketing, you need a complete perspective of things that are going on. You need mostly complete data – or the best you can get. That gives you a full view of reality and your decisions that need to be analysed.

You also need consistent metrics. If you change the goal every week, you're never going to get the results you're looking for. So you need consistency in how you measure your decisions and how you measure your performance as a marketer.

And then you need accurate attribution. You need to know what your decisions are doing and which factors are driving the outcome.

The four pillars of marketing success

If we're going to treat marketing as the skill that we want to get better at and AI as the tool or set of tools that we're going to use to improve that skill, there are generally four areas we need to focus on:

1. Data: You need to be able to aggregate and normalise the data into one context. What that means is that everything is a line probability, all the time variables are the same, and everything has generally a comparable meaning of time and place.

All these different dimensions need to make sense in the same context. Different data comes from different places in different context. Everything needs to be in the same context, with the same intent, and with the same challenges aligning everything.

2. Systems: Second, it's important that marketers — and the people or vendors who are supporting them — have flexible architecture. You need to be able to add new data sources from different places easily, quickly, and readily. If you're changing your stack every year, there's something wrong.

A huge component of this is feature engineering. You need industry experts who understand what these data and variables mean in context and then build an archive of new, interesting features to power these AI algorithms to do more powerful things.

3. Algorithms: The system as a whole should be modular and scalable. That means you can easily plug in different algorithms and different data sources to do different things. So if you've got logistic regression hard-coded into your system, that's probably not a good starting point. You should be able to drop in whatever new thing Google releases next year and have it ready to go quickly.

Everything also needs to be connected by functionality. If you're doing prediction modeling, you need to connect that based on general inputs and then the output, which is a prediction function. If you're doing clustering or segmentation, have the same inputs and outputs structure so you can easily change the data or the algorithms themselves to be adaptable.

4. People: It's important to have the right people for the job — after all, they are the ones driving all of this. Marketers who are trying to improve their skills, are using effective tools, and feel empowered to make decisions are critical.

Marketers are not tools. Marketers are people who have skills, and as the tools evolve, the people need to become increasingly capable. Use the tools to do more of what you do, which is the skill of marketing.

Interested in hearing leading global brands discuss subjects like this in person?

Find out more about the Digital Marketing World Forum (#DMWF) international event series, arriving in Amsterdam from September 19-20 and New York from November 7-8.

View Comments
Leave a comment

Leave a Reply

Your email address will not be published. Required fields are marked *