Artificial intelligence empowers marketers and insights professionals like never before. The seamless combination of disparate and observed customer data points, often at the scale of billions of daily events, analyzed through the lens of supervised and unsupervised machine learning was once considered science fiction (or perhaps marketing fiction in this case). But the industry has made tremendous advances, and now marketers can easily put the most sophisticated machine learning technologies to work for them through fairly simple interfaces and tools.
Assumption vs Reality in Consumer Interest
As a marketer, having a deep understanding of your customer is a prerequisite for the job. However, I would challenge marketers to really think about what informed their perception of their target audience. Did this knowledge stem from some past experience with the customers? Did it stem from some unconscious mental image of the ideal customer? Was the market research describing your customer derived from a ‘representative sample’ of the world? There is really only one way to remove the human bias that often clouds one’s ability to deeply understand their customers – remove the humans.
There are now sophisticated machine-driven techniques that deploy clustering algorithms to understand the nuances of customer segments and micro-audiences. Techniques that are completely data-driven and don’t rely on human inputs. By understanding the behaviors of these micro-audiences, marketers are able to create customized messaging, data-informed media plans, personalized online and offline engagements – all of which ultimately grow your business.
A leading yogurt brand assumed its customers were health and fitness enthusiasts based on their aspirational image of the brand, and crafted their messaging accordingly. However, our unsupervised model, free of any preconceived notions, discovered another avid fan base: people who frequented fast-food restaurants. It also found that these consumers were highly responsive to taste-focused messaging rather than health-focused ones – a profound insight that not only shifted the brand’s marketing strategies, but also resulted in a significant lift in yogurt sales.
Understanding Physical Behaviors From Digital Signals
Tobler’s first law of geography states “everything is related to everything else, but near things are more related than distant things.”
In terms of marketing, consumers who work and live in the same designated marketing area (DMA) are more similar to one another than they are to consumers across the country. But the truth is, DMAs are huge, with hundreds of thousands of people in them. ZIP codes aren’t really that much better. In Manhattan, for example, ZIP codes go across the island from east to west, incorporating a wide range of neighborhoods, income levels, housing status – just to name a few.
AI allows us to look at the world in smaller segments, say 100 meters by 100 meters, which can often be the difference between a successful and unsuccessful marketing campaign. Instead of relying on an overview report of consumer behavior in an entire city from six months ago, AI allows you to examine behavior down to single city blocks as recent as yesterday.
The so-called ‘retail apocalypse’ has made brands painfully aware of the pitfalls of choosing the wrong physical locations. With AI, you can discover unique insights into which regions are more likely to succeed based on that area’s interest in your offering. So if you are an arts and crafts store, we can understand where your potential customers are and what your competitors may or may not already be doing in that region. This data informs where you should open your physical locations, if anywhere.
The Always-on Approach to Media
Traditionally, advertising relied on humans to make important decisions about a campaign, such as who to target, when and where to target them, and how frequently they should be targeted. Human optimization might be okay if campaign managers only have one campaign to manage, but most often they have hundreds, which means any real optimization falls by the wayside. Fortunately, media optimization is an ideal use case for AI.
Many programmatic advertising platforms apply machine learning to all the levers that have a direct impact on campaign results– such as audience segmentation, scale vs precision, frequency and traffic quality. The machine tries a tactic, notes the results, and then improves the criteria in real time.
For instance, let’s look at how AI optimizes audience segmentation. If a marketer wants to launch a customer acquisition campaign, an AI-driven platform can examine all of the online and offline behaviors of your existing customers and build a brand new audience segment that not only looks like but acts like the original customer group. And as the machine is exposed to more positive and negative outcomes (see Chapter 2 for more details), the model will automatically tunes itself and get more accurate every single day.
It gets really interesting when you look under the hood of these models and see the features that score highly. You are able to surface behaviors that are often validating for a marketer but sometimes nuggets of information emerge that are non-intuitive. For example, an internet service provider was prospecting for new customers. The customer acquisition model discovered a significant number of people looking for schools well outside of their current districts. The brand was then able to infer that these people were showing a high probability of moving to a new residence, making them prime candidates for new service messaging.
These models also know which data is important to look at and respond to. Let’s say I go to a surfing site, then to a sporting goods store site, then a swimwear site. A model would classify me as a Surfer. Now let’s say I decide to move to a new, landlocked city and I stop visiting some or all these websites. A finely tuned machine learning model would remove me from the Surfer segment. Data freshness is a huge issue since you don’t want to waste your media budget targeting consumers who have already converted or have already left a segment. Machine learning allows us to update models multiple times per minute, and drive efficiency in media spend.
When digital behavior translates to real-world results, marketers win. Artificial Intelligence gives unprecedented insight into what consumers are looking for, where they’re looking for it and how to ultimately interact with them. This can only happen if you have good quality data, a topic Dstillery will explore in Chapter 4.