The Drum Awards Festival - Official Deadline

-d -h -min -sec

Artificial Intelligence Connectivity Machine Learning

How data, machine learning and AI will perform magic for consumers


By Lisa Lacy, n/a

September 10, 2016 | 11 min read

Imagine wanting a cup of coffee and suddenly finding it before you, freshly prepared to your exacting standards.

Consumer data is key to frictionless experiences.

Consumer data is key to frictionless experiences.

This isn’t a fantasy from the mind of JK Rowling. In the not-so-distant future, this will be reality for Muggles, too.

In fact, thanks to a surge in consumer data, brands and marketers can already make better inferences about consumer wants and needs, but as AI and machine learning are more deftly integrated, insights will only get better, as will the ability to anticipate consumer needs – and to even make decisions on behalf of consumers without any input from them whatsoever. Like, say, ordering a cup of coffee.

The right pillow

As it stands, digital enables brands to customize offers for specific users rather than provide generic solutions. For example, a hotel can use data to greet a guest by name and have a room ready with either a soft or a firm pillow, depending on that guest’s preference, said Aaron Shapiro, CEO of digital agency Huge.

“Technology exists today…to identify small ways that companies can do a better job of servicing users and start to implement and test and evolve from there,” Shapiro said. “At the end of the day, companies use technology to help users solve problems. Users will love it and have a better relationship and drive more business performance, but now we have amazing tools in the form of cloud computing, machine learning and AI to do a better job to meet user needs.”

One heck of a pregnancy reveal

But we haven’t quite worked out all the kinks yet.

Case in point: There’s the oft-cited example of Target, which mines data in part to assign pregnancy prediction scores to its customers, and the poor Minneapolis-area man who reportedly learned this the hard way, complaining the retailer was sending his teenage daughter coupons for baby clothes and cribs only to learn later she was, in fact, pregnant.

The Target-knew-his-daughter-was-pregnant-before-he-did example illustrates how quickly acting on consumer insights can shift from useful to creepy and, as it happens, “creepy” is a word we hear often in conversations about data in marketing.

However, as RP Kumar, executive vice president and director of international research, insights and planning at marketing agency Ketchum, observed, younger generations seem less concerned by privacy as they grew up in a digital era, well aware Google and the like are keeping tabs on them, so this will perhaps be less of an issue moving forward. Time will tell.

Good old-fashioned deduction

In the meantime, with just a bit of analysis, marketers can make useful inferences like, say, consumer behavior changes at different times of day, said John Stewart, senior vice president of strategy and analysis at marketing and technology agency DigitasLBi. In other words, a consumer is likely doing different things at 10:00 AM on a Tuesday than 7:00 PM on a Saturday night.

The trigger before the trigger

And that means brands looking for clever ways to use data and to connect at the right moment — preferably without crossing any lines — should look for what Stewart called 'the trigger before the trigger'. For instance, a brand that sells appliances might think, “Consumers who are moving are likely to be in the market for new appliances.” And that may very well be true. But targeting them when they are moving may be too late. Instead, Stewart said to look at what might precipitate a move, like a career change, which would be reflected in LinkedIn data.

“Forty per cent of career changers are in the market for a new appliance or insurance or cable, so we’re moving now to a market of data,” he said. “Instead of what was a trigger beforehand, now it’s the trigger before and the trigger before, so you can hone that and give the right message.”

But it doesn’t just have to be these anticipatory triggers. It can also include a little good old-fashioned deduction based on the data a brand has on hand. For example, if a brand knows parents are searching for adult acne products and have teenagers at home, marketers can intuit there’s a high probability their children have skin problems, too, said Chris Albert, senior vice president and director of digital/social research and analytics at Ketchum. And brands can use that inference to try to get their products in front of the teens, particularly at a time like back to school or school picture day.

Data patterns, self-learning algorithms and mimicking human behavior — oh my

And while these are certainly savvy ways to use data to target consumers with relevant products and services, the integration of AI and machine learning means a whole new ballgame in which marketers can punt some of the heavy lifting.

Per David Hewitt, vice president at agency SapientNitro, machine learning looks for patterns of data such as consumer behavior and via statistical inferences creates self-learning algorithms to make better decisions and optimize user experiences.

And Jon Suarez-Davis, CMO and CSO of data management platform Krux, noted machine learning can crunch data quickly, which marks a major shift from marketers combing through spreadsheets to unlock their own insights.

“Marketing is an art and a science. The art is about connecting with humans. The science is spinning up all these insights we could never do on our own and allowing us to ask smarter questions and see these patterns — and now I can activate all these events and start to predict what [consumer] behavior is,” Suarez-Davis said. “These are all elements we could only dream about a couple of years ago.”

AI, on the other hand, uses complex programming and software to mimic how humans think, act and respond to a situation or in a conversation. This, in turn, helps brands find meaning in data and make relevant recommendations, Hewitt said.

Per Michael Horn, managing director of data science at Huge, Netflix is a great example of AI and machine learning in action today.

“Netflix tries to predict the thing you want to watch next. The more you watch, the more valuable the service,” he said. “How do they predict what shows you will like? In their case, when you’ve been a member for a long time, they have lots of data on your viewing habits and can say with 85 to 90 per cent confidence, you will like this thing and you will like this other thing. It’s like Amazon product recommendations, too. There’s a long history of using data mining and machine learning to predict your next purchase.”

Per Kumar, the use of AI and machine learning is still rudimentary, but he, too, pointed to Netflix recommendations that are based on previous choices and are continually refined as a result of viewing behavior.

And it’s that awareness of consumer preferences that tailors available options that we are also starting to see from players like Stitch Fix and Spotify, which look at past purchase behavior, lifestyle and affinity data to help narrow the choices they offer consumers, Horn said. In other words, rather than forcing consumers to pore over a huge range of options, Netflix, Amazon, Stitch Fix and Spotify use data about consumers to make educated guesses about what they will probably like and weed out what they won’t.

“You still have control, but it doesn’t require half an hour to check boxes,” Horn added. “It can make better guesses. It’s predictive.”


And as this evolves, consumers will provide less and less input — until their needs are met as if by magic.

Jeremy Lockhorn, vice president of emerging media and mobile at interactive agency Razorfish, noted that we are on the precipice of a shift in consumer expectation in which digital, but especially mobile, is getting so good at knowing a consumer’s context that it is already starting to deliver what people want before they ask for it — like Waze telling a consumer it’s time to leave for the airport because it sees the flight on that consumer’s calendar and knows the traffic conditions.

“This type of magic will be a novelty at first, but will quickly become the expectation of every interaction on a person’s device,” Lockhorn said.

Building profiles

For his part, Hewitt said the marketing experience of the future will be more about integrating with services that recommend a brand’s products as part of a larger solution brought together for the consumer, such as a virtual assistant booking a flight to San Francisco for a consumer who also wants to eat at a four-star vegetarian restaurant when she/he lands without spending more than $25 — and making the latter reservation, too.

“We’ll see a lot more energy put into building profiles so personal assistants can better serve us, but it’s a totally new model for how brands and marketers build awareness and become part of the equation and can be more anticipatory and save me time and energy,” Hewitt added.

Back to that cup of coffee…

Shapiro used the example of a coffee shop on the ground floor of Huge’s office building in Atlanta, which Huge is using for what he called “a test bed of retail innovation.”

Thanks to an app, the coffee shop knows when employees are nearby and can place and fill their regular orders automatically so the coffee is waiting when said employee walks in the door in the morning.

“As a consumer, I never had to do anything to get my coffee — I didn’t have to order, wait or pay. It’s completely frictionless,” Shapiro said. “Because of data, we can predict the order you’re likely to make and if we make a mistake, it’s not a big deal. We just wasted a cup of coffee. But I think this will happen more and more as part of a lot of experiences.”

And that’s because brands/marketers can learn certain people place certain orders repeatedly and are highly likely to request a regular order, so it’s simply about building a data profile so that the data collected can start making decisions on behalf of consumers, Shapiro said.

If you think about it, Amazon Dash really is labor intensive

“Amazon Dash is great, but what’s even more innovative is Amazon has showcased a new version of Whirlpool washing machines with a built-in scale and measuring solution as part of the actual machine,” Shapiro said. “You put in detergent and run out and it knows and an order is automatically placed and delivered. It’s so frictionless, you don’t have to push the button. If you think about, it, when you wake up, why shouldn’t my alarm talk to my coffee machine? And when I’m cooking, why shouldn’t my pot talk to my oven so if I’m burning something, it can communicate to the oven to turn down the temperature. It’s that kind of connectivity that is the way data can connect disparate systems that is the next generation of digital and can profoundly impact people’s lives.”

Artificial Intelligence Connectivity Machine Learning

More from Artificial Intelligence

View all


Industry insights

View all
Add your own content +