Marketing to the Power of AI

Why are so many marketing directors losing their jobs?

The marketing world is being shaken up. On one side, companies are questioning (at times, simply eliminating) marketing directors. On the other, there’s a feeling of euphoria over technology and artificial intelligence.

Between the two extremes, the watershed criterion is surely the ability to deliver results. It doesn’t matter how charming the name of your virtual assistant is or how amazing your TV campaign looks, marketing has to deliver more.

To figure out how far marketing to the power of artificial intelligence (AI) can take us, let’s start with the exercise of a self-driving car — how do you teach a car to drive itself?

If you thought of teaching the car’s program by instructing it to “stop at the red light,” “go at the green light,” or “avoid hitting the tree,” you got it wrong. That’s not quite how it works.

Self-driving cars are equipped with hundreds of sensors that capture millions of data points per second. They absorb everything they see, hear, and feel while a flesh and blood person drives them. Hours and hours of “people driving” are recorded. This is used to produce an application, an algorithm that will reproduce that whole intricate relationship between stimuli and reactions in another car, a car that will drive itself. A car learns to drive like a child learns to talk — it observes, tries, and makes mistakes until it gets it right.

This method of learning is what we call machine learning, a discipline within the field of AI.

Applying artificial intelligence to marketing

The following stories are a good illustration of the past, present, and future of digital marketing:

  1. At the website of Bank A, every user is automatically classified in terms of propensity models (chance of purchasing something) based on their browsing behavior. Every piece of data available is used in the calculation: browser, speed, device, location, time, day, etc., just like the self-driving car uses information about speed, distances, street signs, and maps. By making the calculations itself, the program can classify a person as being more likely to use the call center. The reaction is fast, and the display size of the call center number on the website increases immediately. Since this model has been in use at this company, conversions both via the call center and sales on the website have exploded. There’s no need for marketing directors to gather around a conference table and discuss how men and women of different ages should be treated. In the meetings at this company, no one talks about age or gender, or other demographic information. Instead, they focus on propensity, statistics, and computation.
  2. At Bank B, the challenge is to “digitize clients.” So they created an “omnichannel” message system, in which every offline action at each of their branches generates a message. For example, somebody makes a bank transfer at an ATM and soon after receives a text message on their phone: “How about making your next wire transfer through the app?” At this company’s meetings, people discuss rules for selling different products, what the message should say, what the website layout should look like, and the necessary budget. The company is doing well. But that’s all.

Example A is typical of a marketing department that’s ready for AI. Using data input (big data) and comparing them with outputs (success with online purchasing or call center conversion), algorithms are created and replicated by applications. Experiments and tests are conducted frequently to improve the models, control groups lend them credibility, and the bank grows exponentially, swallowing its competitors.

Rule-based marketing is predominant in companies like the one in example B. Most marketing departments and their current digital strategies follow the logic of an adult learning a new language: grammar rules, subject, object, verb, predicate — which is very similar to the concept of teaching a car that a red light means stop. The bank grows slowly and is under constant pressure to cut costs because it is seeking efficiency using outdated formulas (and marketing).

Guess which marketing director has their job on the line?

How to do marketing to the power of AI

In a hypothetical digital marketing evolutionary scale (Figure 1), at the top of the pyramid we have marketing to the power of AI (like in Bank A), while the lines below include the old digital marketing and marketing automation.

Figure 1: The evolution of digital marketing.

How does marketing make it all the way to the top? Three dimensions come into play:

  1. Algorithms are propensity models that leverage an abundance of data and powerful tools can radically change how marketing works. This is excellent because people are different. Consumer classification criteria like “male, 40-50 years old” are antiquated and inefficient. I’m 43, and I don’t fit into any of the marketing “personas” I’ve seen recently.
  2. Personalization 1:1 means being capable of sending messages to each person in an individualized manner, one by one, regardless of channel (“omnichannel”). My bank, for example, doesn’t understand that. Last week, they sent me an email suggesting an investment. When I accessed their site, the banner invited me to make a loan.
  3. Real time means doing everything we discussed above instantly. For example, if the propensity model says that I’m a prospect for product A (algorithm), and now I’m talking to the call center about product B (omnichannel), the model should reclassify me and show me a different banner when I next visit the website seconds from now. In other words, it is constantly following me and adapting in real time.

Figure 2: Return On eXPerience.

Return on Experience (ROXp)

At the top is the expected result, the return on investment, which can be called Return on Experience (Return On eXperience – ROXp), since the goal of everything we’re talking about so far is to deliver the right message, at the right time, to the right person. Using the best channel, coordinating and orchestrating in real time, based on propensity models, we put the customer first. Now that’s a real “customer centricity.”

Bank A, for example, sees its investment in platforms and AI (Adobe Analytics + Audience Manager DMP + Adobe Target) pay for itself in a few months. Today, while people are discussing ROI (return on investment), their marketing department is celebrating an additional revenue that is more than 20 times greater than the investment.

Examples of ROXp in media purchasing (Advertising)

Another example of delivering results is known as “the last mile” in media purchasing. This strategy is based on the concept of data propensity: in this example, a large retailer runs its predictive models internally and, when it decides to use them (to personalize the experience), it does so with programmatic media and Facebook. However, instead of sending user data to the media companies (Facebook and programmatic players are media vendors), it sends only the banner and device information. In other words, the brand is telling the social network and the display banner provider to show this specific banner on that specific cell phone.

Do you know why that would be much more efficient? First, because the brand is not feeding the social network algorithm, preventing this knowledge from being used to its competitors’ benefit. The other day, I clicked on a banner for an online course and, ever since, I’ve only seen banners about that. The retailer skipped this step and is using the data only for itself. It is crazy (in addition to a violation of privacy laws) to provide the media company with consumers’ email data, the car they bought, etc., but there is a lot of digital marketing that still does that.

Second, because campaigns that use this strategy, when compared to parallel campaigns (control groups) on the same media channels, have two to three times higher conversion rates, the ROI is eight times greater than the amount invested.

Marc Pritchard, chief brand officer of P&G, recently talked in a CNBC interview about a very similar story. Using the data property and last-mile strategies, P&G has reduced their advertising expenses by more than $1 billion, while sales continue to rise, nonstop.

Marketing strategies using artificial intelligence are created and improved every day. It’s up to each marketing director to decide which one best fits the business needs and, consequently, which is best for the future of the company. These are examples I’ve given:

Figure 3: Map of examples along the evolutionary ladder of digital marketing.

The foundation for the development of this new kind of marketing includes a series of capabilities and processes that marketing professionals and companies need in order to improve:

  1. Statistical thinking: being truly “data-driven” requires changes in mindset.
  2. Marketing platforms: from cloud marketing, to cloud experience, to the cloud.
  3. Multidisciplinary teams: development of autonomous and independent teams.
  4. Data: ownership, governance, and data protection law.
  5. Hierarchy, departments other than marketing, goals, and incentives.

To become a marketing professional to the power of AI, one has to study and update oneself about all of that. There’s no other way. The good news is there’s still time.

In the upcoming articles, I’m going to address each one of these items in turn.

Best regards!